dataconvert updates

Converts CSV, xlsx/xls/ods, directories, ZIPs and globs into one INSERT
file per table. Producer-specific layouts (header/data rows found by a
marker cell) and sheet naming live in a gitignored dataconvert.json, with
dataconvert-example.json as the template. --max-rows samples each table
and SCHEMA.md records columns, types, row counts and full sizes.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

.gitignore

@@ -40,7 +40,5 @@ cfg/dlt/
 # not land here. They are versioned in their own repo.
 #
 # Anchored at the ROOT on purpose: a copy is a SIBLING of rig/, so a rule inside
-# rig/.gitignore cannot see it. The negation must name the full path for the same
-# reason — `*-rig/` is unanchored and matches at any depth, including rig/sample-rig.
+# rig/.gitignore cannot see it.
 *-rig/
-!rig/sample-rig/

CLAUDE.md

@@ -129,7 +129,7 @@ Every script stays runnable on its own — the standalone rule holds:
 ```bash
 python build.py --cfg amar         # -> gen/amar/
 cd gen/standalone && python run.py # bare-metal
-./ctrl/kind-up.sh                  # still works directly
+./ctrl/cluster.sh up               # still runs directly; rig builds the cluster
 cd gen/<room> && ./ctrl/start.sh   # each room owns its lifecycle scripts
 ```
 

Makefile

@@ -13,7 +13,7 @@
 #   make component ARGS="publish soleprint-ui /tmp/out --dist"
 #   make deploy    ARGS="--build"
 #
-# Every script stays runnable on its own (./ctrl/kind-up.sh still works, and each
+# Every script stays runnable on its own (./ctrl/cluster.sh up still works, and each
 # built room keeps its own gen/<room>/ctrl/*.sh) — the standalone rule holds, and
 # this only saves typing.
 #
@@ -42,7 +42,7 @@ $(eval $(ARGS):;@:)
 endif
 
 .DEFAULT_GOAL := help
-.PHONY: help build start stop dist docs cluster deploy component
+.PHONY: help build start stop dist theme docs cluster deploy component
 
 help:                          ## list targets
 	@grep -hE '^[a-z]+:.*?##' $(MAKEFILE_LIST) | sed 's/:.*##/\t/' | expand -t16
@@ -61,6 +61,11 @@ stop:                          ## stop a running room  [<room>]
 dist:                          ## compile the plexus UIs to single files  [<room>]
 	bash ctrl/dist.sh $(or $(ARGS),$(ROOM))
 
+# ── theme ──────────────────────────────────────────────────────────────────
+
+theme:                         ## ad-hoc pages: scaffold, add parts, bake  [new|parts|bake|check|export]
+	bash ctrl/theme.sh $(or $(ARGS),bake)
+
 # ── docs ───────────────────────────────────────────────────────────────────
 
 docs:                          ## documentation  [serve [port]|graphs [theme]]  (default serve)

berth/.gitignore

@@ -0,0 +1,21 @@
+# Machine-local config and credentials. Never committed.
+ctrl/.env
+
+# Rendered output. Regenerate with `make services render <target>`.
+# Generated config is an artifact, not source — the estate file is the source.
+#
+# The path is ctrl/render/out/, NOT render/out/. A pattern containing a slash is
+# anchored to the directory holding this .gitignore, so `render/out/` would mean
+# berth/render/out/ — which does not exist, and the real output would have been
+# committed. Caught by `git check-ignore -v`, which is the only way to be sure.
+ctrl/render/out/
+
+# The "default" scratch bucket: always gitignored, never versioned.
+def/
+
+# Key material. NEVER committed.
+#
+# Anchored to ctrl/ for the same reason as render/out/ above: a pattern with a
+# slash resolves against this file's own directory. `git check-ignore -v` is the
+# only way to confirm it, and ctrl/vpn.sh refuses to write a key until it does.
+ctrl/.secrets/

berth/Makefile

@@ -0,0 +1,78 @@
+# One target per ctrl/ script; the subcommand is an argument, not a second
+# target: `make estate show`, not `make estate-show`. The logic lives in the
+# scripts, never here.
+#
+# Config layers, weakest first: ctrl/versions.env < ctrl/env.d/<target>.env <
+# ctrl/.env < the environment. So `make estate plan TARGET=gcp` beats all.
+#
+# Every target defaults to its READ-ONLY verb, and the verbs that change a live
+# estate are not reachable by a bare word. Rationale: README.md.
+
+ESTATE := $(or $(shell sed -n 's/^ESTATE=//p' ctrl/.env 2>/dev/null),$(shell ls estate/*.json 2>/dev/null | head -1 | xargs -r basename | sed 's/\.json$$//'))
+TARGET := $(or $(shell sed -n 's/^TARGET=//p' ctrl/.env 2>/dev/null),aws)
+
+.PHONY: help check selftest estate services vpn dns certs host ports registry docs
+
+help:                          ## list targets
+	@grep -hE '^[a-z][a-z-]*:.*?##' $(MAKEFILE_LIST) | sed 's/:.*##/\t/' | expand -t16
+
+# ── preflight ──────────────────────────────────────────────────────────────
+
+check:                         ## is this estate coherent? reports, never fixes
+	bash ctrl/check.sh
+
+selftest:                      ## does berth still do what it says? exits 1 if not
+	bash ctrl/selftest.sh
+
+ports:                         ## port map  [show|verify]  (default show)
+	bash ctrl/ports.sh $(or $(ARGS),show)
+
+# ── the estate ─────────────────────────────────────────────────────────────
+
+estate:                        ## the estate  [show|list|plan|apply|destroy]  (default show)
+	bash ctrl/estate.sh $(or $(ARGS),show)
+
+services:                      ## gateway routes  [list|render <target>|deploy]  (default list)
+	bash ctrl/services.sh $(or $(ARGS),list)
+
+# ── the network ────────────────────────────────────────────────────────────
+
+vpn:                           ## overlays  [list|show <ov>|check|render|keygen]  (default list)
+	bash ctrl/vpn.sh $(or $(ARGS),list)
+
+# ── names and trust ────────────────────────────────────────────────────────
+
+dns:                           ## DNS records  [list|add|add-wildcard|remove]  (default list)
+	bash ctrl/dns.sh $(or $(ARGS),list)
+
+certs:                         ## TLS  [status|verify|renew|push]  (default status)
+	bash ctrl/certs.sh $(or $(ARGS),status)
+
+# ── the box ────────────────────────────────────────────────────────────────
+
+host:                          ## the remote box  [status|ports|services]  (default status)
+	bash ctrl/host.sh $(or $(ARGS),status)
+
+registry:                      ## the image registry  [status]  (default status)
+	bash ctrl/registry.sh $(or $(ARGS),status)
+
+# ── docs ───────────────────────────────────────────────────────────────────
+
+docs:                          ## documentation  [serve|graphs]  (default serve)
+	bash ctrl/docs.sh $(or $(ARGS),serve)
+
+# ── swallowing the argument words — MUST BE LAST IN THIS FILE ──────────────
+#
+# Words after the target are arguments, but make reads each as a goal, so each
+# gets a no-op rule. This block must come AFTER the real targets: when an
+# argument names one (`make host ports`, `make vpn check`), the last definition
+# wins, and it has to be the no-op. With it first, make ran both scripts.
+#
+# Make's "overriding recipe" warning is the swallow working as intended.
+ARGS := $(wordlist 2,$(words $(MAKECMDGOALS)),$(MAKECMDGOALS))
+ifneq ($(ARGS),)
+$(eval $(ARGS):;@:)
+# .PHONY too: some of those words name real directories (ctrl, estate, render),
+# and make treats an existing directory as already built.
+.PHONY: $(ARGS)
+endif

berth/README.md

@@ -0,0 +1,250 @@
+# berth
+
+spr's deploy half. A rig is a mobile installation; a **berth** is the allocated, paid-for
+place where it is moored and actually operates. Local rig → remote berth.
+
+```bash
+make check              # is this estate coherent? reports, never fixes
+make estate show        # the description, resolved
+make vpn check          # the overlay: addresses, routing, bindings, key hygiene
+make services render aws
+make ports verify       # does the local map still agree with rig?
+```
+
+`rm -rf berth/` is the uninstall.
+
+---
+
+## What berth is
+
+**One description of an estate, with a swappable executor.** The description is the artifact;
+the tool that runs it is a rendering.
+
+| layer | what berth uses | note |
+| --- | --- | --- |
+| overlay | **WireGuard**, config generated per peer | GPL-2.0, in-kernel, and **no coordination server** |
+| infra | **OpenTofu** — one executor, not a pair | plain `.tf`; `terraform` works identically |
+| gateway | Caddy locally, nginx on the box | a projection with per-target rules, not a format conversion |
+| pipeline | Woodpecker | Actions/GitLab reachable from the same description; not built |
+
+The estate's facts — domain, hosts, ports, instance size, firewall rules, services — live in
+**one file every rendering reads** (`estate/<name>.json`), rather than being restated in
+one tool's language and again in another's. **Swappability is bought by the description, not
+by maintaining two renderings** — a rendering you *can* produce, not one you *must* keep in
+step. Two live renderings cost you every resource twice, forever, with nothing enforcing that
+they agree.
+
+**The seam belongs in a script, not in a tool.** A tool's schema is a ceiling you do not
+control.
+
+*(This once leaned on a specific precedent, since withdrawn — `✖ B2` in [STALE.md](STALE.md).)*
+
+**OpenTofu, and only OpenTofu.** Terraform has been BUSL-licensed since 2023; OpenTofu is the
+CLI-compatible MPL-2.0 fork (Linux Foundation). Write plain `.tf` that runs under both; the
+scripts say `tofu`, and `terraform` works identically.
+
+**Pipelines are the second executor axis, and are not built.** Woodpecker is the
+self-hosted rendering; Actions and GitLab CI are the standards that must be reachable from
+the same description. Named here so the infra seam is not designed in a way that forecloses
+it.
+
+---
+
+## The overlay is berth's network layer
+
+Two instances in different clouds cannot share a VPC. A WireGuard overlay gives them one flat
+address space that berth owns and can reproduce on any provider — and, under a flaky
+environment, a second layer beneath whatever the provider offers.
+
+That inverts the usual cloud pattern. Instead of a VPC with security groups and private
+subnets, each instance gets a public IP, opens **only** the WireGuard port, and carries
+everything else inside the tunnel. **The security boundary moves out of the provider's VPC
+and into a layer that is identical on AWS, on GCP, and on a laptop behind NAT.**
+
+**Plain WireGuard, not Tailscale/Headscale/NetBird.** Tailscale's client is open but its
+coordination plane is proprietary SaaS; Headscale and NetBird are open but add a control plane
+to run. Plain WireGuard needs no server at all.
+
+**The cost is real and accepted:** no NAT traversal, no relay, no peer discovery. A peer behind
+NAT must dial one with a public endpoint. Fine here — the instances have public IPs and the dev
+box roams — but two roaming peers cannot reach each other. That is why `PersistentKeepalive` is
+a checked invariant rather than a detail.
+
+**Keys never enter the description.** Private keys are generated on the peer that owns them
+(`make vpn keygen`) into `ctrl/.secrets/` and injected only at render time; public keys live in
+the estate, because a config cannot be built without them. `vpn.sh` **refuses to write** either
+a key or a rendered config until `git check-ignore` confirms the path is ignored — this repo
+has already been bitten once by a `.gitignore` pattern anchoring to the wrong directory.
+
+This also decides something about the IaC layer: **OpenTofu must never generate a WireGuard
+private key**, because Terraform-lineage state stores every resource attribute in plaintext.
+
+---
+
+## Two rules that are not style preferences
+
+### 1. Every default is the read-only verb
+
+```
+make estate    -> show      make certs  -> status
+make dns       -> list      make host   -> status
+make estate apply / destroy -> print the plan, then refuse without --yes
+```
+
+rig's `make cluster` defaults to `up`, because every rig verb is safe — a kind cluster is
+disposable. berth's are not: `tofu destroy` costs money and takes live DNS with it. **A
+tool where every verb is safe must not grow verbs that are not.**
+
+The failure this prevents is not hypothetical. `ppl/ctrl/certs.sh:42` is `CMD="${1:-all}"`,
+so a bare `./ctrl/certs.sh` there issues a real Let's Encrypt cert, rsyncs it to the gateway,
+and reloads nginx. berth's `certs` defaults to `status`.
+
+### 2. berth and rig share a convention, not code
+
+Neither imports the other. They match on **shape** — the `make <noun> <verb>` dispatch, the
+four-source config layering, the key names — and consistency is verified by
+**recomputation**: `make ports verify` recomputes rig's `20000 + (cksum(name) % 200) * 10`
+to check the local map, rather than sourcing rig's `lib/config.sh`.
+
+Copying three stable lines is the whole cost of not coupling them. A shared library would
+put something outside `rig/` on rig's path, and rig's promise is that
+`grep -rIn -iE 'soleprint|\bspr\b'` across it returns nothing.
+
+**rig is also unaware that berth exists.** `ppl/local/Caddyfile` is berth's to generate; rig
+must not reference `local.ar` — its handover scrub refuses the string.
+
+---
+
+## The gateway doctrine
+
+Practised across this codebase for a long time and never written down, so: written down.
+
+- **Caddy where routing is dynamic and config-driven** — the in-cluster gateway that
+  multiplexes by Host header, and the host-side `.local.ar` name→port map.
+- **nginx where it is a static server or a plain long-running compose service on the box.**
+- **Envoy in `mpr`** — a deliberate one-off, not a third pattern.
+- **`ingress-nginx` only as a kind addon** — a different thing again from either gateway.
+
+### Rendering is a projection, not a format conversion
+
+The local Caddyfile and the box's nginx are not two spellings of the same content:
+
+| | local (Caddy) | cloud (nginx) |
+| --- | --- | --- |
+| granularity | one file | one file per vhost |
+| blocks per service | one | two (`:80` redirect + `:443` server) |
+| TLS | none; every address needs an explicit `:80` | one shared wildcard cert |
+| upstream | `localhost:<port>` | container name + `resolver 127.0.0.11` |
+| name depth | free | constrained by the cert |
+| ambiguity | most specific wins | exact, else `default_server` (= load order) |
+
+The `:80` is not decoration: without it Caddy 2 defaults each site to `:443` with auto-HTTPS,
+which on `*.local.ar` means cert provisioning that fails and breaks the listener. The
+variable upstream is not decoration either: naming the upstream in a variable forces runtime
+DNS resolution, so nginx **starts when the upstream container is absent** — which is what
+lets one nginx front a dozen independent compose stacks.
+
+Because the two disambiguate by **opposite** rules, a name set that is unambiguous locally
+can be ambiguous on the box. `make check` asserts against the projection, not the source.
+
+### Installing generated vhosts — an order that is not optional
+
+1. Generated config lands in `conf.d/generated/`, **not** `conf.d/`. `ppl/ctrl/deploy.sh`
+   rsyncs with `--delete`; sharing a directory means one set gets erased.
+2. `nginx.conf` needs a **third** include line — its `conf.d/*.conf` glob does not recurse,
+   which is why `conf.d/soleprint/*.conf` already needs its own.
+3. That include changes **load order**, and load order decides which `:443` block catches
+   unmatched names. So `default.conf`'s commented-out `:443 default_server` must be restored
+   **first**. `make check` fails on it deliberately: it is a gate, not a warning.
+
+---
+
+## What the checks assert, and why
+
+The scripts are deliberately thin on comment — the reasoning lives here. Every check below
+corresponds to something that is wrong, or was wrong, in a real estate.
+
+### `make check` — the estate
+
+| assertion | the failure it catches |
+| --- | --- |
+| every service name is covered by an issued cert SAN | **a wildcard matches exactly one label.** `*.d.com` covers `a.d.com` but not `a.b.d.com`, which needs its own SAN. Surfaces otherwise as a browser TLS warning, far from its cause |
+| a `:443 default_server` exists | DNS and the cert are wildcard but nginx matches `server_name` exactly, so without one the fallback for any unknown name is whichever vhost loads first — alphabetically, by accident |
+| firewall rules and listeners agree | a rule allowing a port nothing listens on is **dead config**; a service no compose file declares is **undocumented state**. Neither is visible from one side alone, which is why the inventory has two halves |
+| `HOST` is an ssh alias, never a hostname | there is no `Host <domain>` block, so a bare hostname falls through to the global defaults, ssh offers every key in the agent in turn, and `MaxAuthTries` (6) trips with *"Too many authentication failures"* before reaching the right one. The aliases set `IdentitiesOnly yes` |
+
+### `make vpn check` — the overlay
+
+| assertion | the failure it catches |
+| --- | --- |
+| peer addresses unique and inside the subnet | a duplicate is a silent misroute, never an error |
+| AllowedIPs do not overlap | AllowedIPs is **cryptokey routing** — the route table and the ACL at once. Overlapping ranges resolve to the last match, so an overlap is both a misroute and an unintended grant |
+| something carries `PersistentKeepalive` if anything roams | without it a NAT mapping expires and the tunnel works only while traffic flows outward — *"works sometimes"*, the hardest failure to read. Note it is **not** a property of the roaming peer's own entry: the roaming machine sets it on the entry for the peer it **dials** |
+| the listen port is in the firewall | otherwise no peer can be dialed at all |
+| no private key in the description | public keys are *also* 44-char base64, so the shape proves nothing. The real assertions are **no field named `priv*`** and **no key outside a `public_key` field** |
+| overlay-reached services bind a reachable address | **a tunnel cannot reach loopback.** A service on `127.0.0.1` is unreachable over the overlay; one on `0.0.0.0` is reachable but also exposed to the whole LAN |
+
+### Capturing the overlay
+
+```bash
+sudo wg show | make vpn capture --write
+```
+
+`wg show` has three forms and **only the first is safe**:
+
+| form | safe | why |
+| --- | --- | --- |
+| `wg show` | **yes** | prints `private key: (hidden)` |
+| `wg show <if> dump` | **no** | field 1 of the first line *is* the private key |
+| `wg showconf <if>` | **no** | prints `PrivateKey=` outright |
+
+`capture` refuses the latter two by shape. It matches peers by **allowed-ips address, not
+public key** — the keys are exactly what is missing at that point — and **drops a roaming
+peer's endpoint in the parser**, since that value is a home ISP address and a roaming peer
+has no stable endpoint anyway.
+
+## Layout
+
+```
+berth/
+├── Makefile              one target per ctrl/ script; the verb is an argument
+├── STALE.md              withdrawn assumptions, each with a check that runs
+├── estate/<name>.json    THE ARTIFACT — one description, many renderings
+└── ctrl/
+    ├── check.sh          reports and instructs; never fixes
+    ├── estate.sh         show | list | plan | apply --yes | destroy --yes
+    ├── services.sh       list | render <aws|gcp|local> | deploy
+    ├── ports.sh          show | verify   (the rig coincidence check)
+    ├── dns.sh certs.sh host.sh registry.sh docs.sh
+    ├── versions.env      pinned toolchain          (weakest layer)
+    ├── env.d/<target>.env  provider shape: aws | gcp
+    ├── .env.example      -> ctrl/.env, machine-local  (gitignored)
+    ├── lib/config.sh     the four-layer load, from rig
+    ├── lib/estate.sh     reading and projecting the estate
+    └── render/*.tmpl     nginx vhost shapes, substituted with sed
+```
+
+Config layers, weakest first: `versions.env` → `env.d/<target>.env` → `ctrl/.env` → the
+caller's environment. So `make estate plan TARGET=gcp` beats everything.
+
+**Identity is explicit — the inversion of rig.** rig derives its name from its folder so that
+copies never collide. berth refuses to guess, because a deployment has exactly one production
+and a wrong guess acts on the wrong estate. `ESTATE` names a file; the only convenience is
+that a single `estate/*.json` is used without being asked for.
+
+**`python3`, not `jq`.** rig ships a pinned static `jq` because its floor is "docker and
+nothing else" on a machine it does not control. berth's floor is already higher, so
+`python3` is a dependency it *has* rather than one it *adds* — the same reasoning by which
+rig chose `sed` over `envsubst`.
+
+---
+
+## Status
+
+**B0 (this) is the shape.** `estate/mcrn.json` is marked `UNVERIFIED`: it records what the
+repos *claim*, because `ppl/infra/` was written and never applied — no `~/.pulumi`, no
+`venv`, no stack state, files dated `mar 6`. B1's read-only inventory is what replaces those
+claims with observations. Until then, `estate plan` and `estate apply` refuse: there is
+nothing truthful to compare against yet.
+
+`make check` currently fails on two real things — see `def/plans/36.0/berth.md`.

berth/STALE.md

@@ -0,0 +1,104 @@
+# berth — withdrawn assumptions
+
+**Everything in this file is no longer true.**
+
+It exists so the live docs stay short and so a withdrawn assumption cannot quietly return:
+each entry carries a **check**, and `ctrl/selftest.sh` runs every one of them. A retraction
+that is only prose is a retraction nobody re-reads.
+
+Kept rather than deleted for the reason the docgen thread already wrote down —
+*a requirement that disappears without explanation comes back.*
+
+### For agents
+
+- **Do not restate these** in a plan, a README or a comment. One line pointing here is enough.
+- **Ids are stable.** Cite `✖ B3`; do not re-explain it.
+- **When you withdraw an assumption, move it here** — quoted claim, where it came from, what
+  superseded it and why, what changed in the code, and a check that proves it is gone.
+- **Only withdrawn things belong here.** A warning that is still actionable is a live rule,
+  however historical it sounds, and stays where it is.
+
+---
+
+**✖ B1 — "Pulumi is the source in spr, and Terraform must match the config."**
+*(INDEX §5, carried from 35.3)* Withdrawn 2026-09-12. berth uses **OpenTofu and only
+OpenTofu**. The rule assumed *open source* and *industry standard* pull apart — Terraform
+being BUSL, Pulumi being the open alternative. OpenTofu is both: the standard language, plain
+Terraform-compatible HCL, under MPL-2.0. Swappability was never bought by keeping two
+renderings; it is bought by `estate/*.json` being the artifact — a rendering you *can*
+produce, not one you *must* maintain.
+**Gone from:** `ctrl/versions.env` (no `PULUMI_VERSION`), `ctrl/estate.sh` (`plan` runs one
+executor), `ctrl/lib/config.sh` (`PULUMI_STACK` → `TOFU_WORKSPACE`), `ctrl/check.sh`
+(toolchain list).
+**Deliberately kept:** `README.md` and `estate/mcrn.json` both record that `ppl/infra/` was
+written and never applied — *"no `~/.pulumi`, no venv, no stack state"*. That is a historical
+fact about the estate, not a live dependency.
+**Check:** no `pulumi` in `ctrl/` or `Makefile`.
+
+**✖ B2 — "ctlptl's rejection is the precedent for *the seam belongs in a script, not a tool*."**
+*(INDEX §5, `berth/README.md`)* Withdrawn 2026-09-12. **ctlptl was reinstated** — pinned in
+`rig/ctrl/versions.env` at v0.9.4 — and had been removed for the wrong reason. The rule may
+still hold; it now has to stand on its own reasoning rather than that example.
+**Gone from:** `README.md` — the argument is stated directly, with no borrowed evidence.
+**Check:** `ctlptl` appears nowhere in berth.
+
+**✖ B3 — "`wg show` is safe; `wg showconf` is not."** *(my own note, 2026-09-12)* Incomplete,
+and the gap is the dangerous one. There are **three** forms, and `wg show <if> dump` puts the
+**private key in field 1 of the first line**. Stated as a two-way distinction, the `dump` form
+reads as safe.
+**Now:** `wg show` plain is safe; `dump` and `showconf` are not. `ctrl/vpn.sh capture` refuses
+the latter two **by shape**, rather than parsing around them.
+**Check:** `vpn.sh` names all three forms, and `capture` rejects both unsafe ones.
+
+**✖ B4 — "A roaming peer must set `PersistentKeepalive` on its own entry."**
+*(`ctrl/vpn.sh`, first draft)* Wrong side, and wrong in the direction that looks fine:
+`PersistentKeepalive` is set per-peer in a config, so the roaming machine sets it on the entry
+for the peer it **dials**. The original check would have **warned on a correctly configured
+overlay**.
+**Now:** checked once per overlay — if anything roams, some peer entry must carry a keepalive.
+**Check:** the invariant is not keyed on the roaming peer's own `keepalive` field.
+
+**✖ B5 — "The Makefile's pass-through block goes near the top, with the other variables."**
+*(`Makefile`, inherited from rig's layout)* Withdrawn 2026-09-12. When a subcommand **names a
+real target**, make has two recipes for it and the **last definition wins** — so with the block
+first, `make host ports` ran `ctrl/host.sh ports` *and* `ctrl/ports.sh ports`, the second
+failing because `ports` is not one of its verbs. Same for `make host services`, `make vpn
+check`, `make vpn show estate`.
+**Now:** the `$(eval $(ARGS):;@:)` block is **last in the file**, so the no-op wins and the
+word is swallowed — which is what an argument is. Make's *"overriding recipe"* warning is the
+swallow working.
+**Check:** every colliding invocation dispatches to exactly one script.
+
+**✖ B6 — "A base64 key in the description can be caught by its shape."**
+*(`ctrl/vpn.sh check`, first draft)* WireGuard **public** keys are also 44-char base64 and
+legitimately live in the estate, so shape alone proves nothing and would flag correct data.
+**Now:** two assertions instead — **no field named `priv*`**, and **no base64 key outside a
+`public_key` field**.
+**Check:** the estate's public keys do not trip the secret check.
+
+**✖ B7 — "`network.wireguard` is where the overlay is described."** *(`estate/mcrn.json`)*
+Superseded 2026-09-12: WireGuard is berth's network layer, not one service's transport, so it
+is a top-level `vpn` block with named overlays and peers. `ctrl/check.sh` and
+`ctrl/registry.sh` were repointed.
+**Gone from:** the estate schema — a `wireguard_moved` tombstone marks the old key.
+**Check:** nothing reads `network.wireguard.*`.
+
+**✖ B8 — "berth and rig are related through their first uses."** *(early framing)* Withdrawn:
+**rig and berth are peers — neither depends on the other.** They match on shape (dispatch,
+config layering, key names) and consistency is verified by **recomputation**, never by
+dependency. A shared library would put something outside `rig/` on rig's path.
+**Check:** berth imports nothing from rig; `ports.sh` recomputes the port formula and agrees
+with rig's golden values.
+
+**✖ B9 — "`langfuse.mcrn.ar` is an exception that cannot be generated."**
+*(`estate/mcrn.json`, `raw: true`)* Withdrawn 2026-09-14. It was filed as the one route a
+template could not express — a static `upstream{}` to a WireGuard address, with no `resolver`
+and no `set $var`. It was not an exception; it was **the first instance of the general case**.
+Those three properties are not three decisions, they are one: *this service is reached by
+address on the overlay, not by name on the docker network.* Naming that decision —
+`placement` — makes the file renderable.
+**Gone from:** `estate/mcrn.json` — `lng` and `langfuse` were **two entries for one socket**
+and are now one service with `placement: local`, `peer: nrft`, and a `local_host` for the
+name it answers to locally. `raw` is dropped.
+**Check:** the generated vhost matches the hand-written one, normalised for comments and
+whitespace — proof against a live route rather than an assertion.

berth/ctrl/.env.example

@@ -0,0 +1,18 @@
+# Machine-local config. Copy to ctrl/.env (gitignored) and edit.
+#
+# The estate's FACTS live in estate/<name>.json.
+# The provider's SHAPE lives in ctrl/env.d/<target>.env.
+# This file is only what differs between machines, plus credentials.
+
+# ESTATE is required when estate/ holds more than one file.
+# ESTATE=mcrn
+# TARGET=aws
+
+# ssh ALIASES, never hostnames — check.sh refuses a value containing a dot.
+# HOST=mcrn              # app user, no sudo
+# HOST_ADMIN=mcrn-admin  # sudo, only where genuinely required
+
+# Credentials: names only, never values. The secrets stay in ~/.aws and
+# ~/.config/gcloud where their own tooling manages them.
+# AWS_PROFILE=default
+# GCP_PROJECT=

berth/ctrl/certs.sh

@@ -0,0 +1,58 @@
+#!/usr/bin/env bash
+# The wildcard TLS cert for the gateway.
+#
+# Usage:
+#   ./certs.sh status     # SANs issued vs SANs the services need
+#   ./certs.sh verify     # inspect the cert served on :443
+#   ./certs.sh renew      # refuses: issues a real cert
+#   ./certs.sh push       # refuses: ships to a live gateway
+
+set -euo pipefail
+cd "$(dirname "$0")"
+
+source ./lib/config.sh
+source ./lib/estate.sh
+load_config
+
+status() {
+    local issued; issued=$(estate_get "certs.issued" | python3 -c 'import json,sys
+try: print("\n".join(json.load(sys.stdin)))
+except Exception: pass')
+    echo "issued SANs (estate/${ESTATE}.json: certs.issued):"
+    echo "$issued" | sed 's/^/  /'
+    echo
+    echo "SANs the services NEED (derived from services[]):"
+    estate_sans | sed 's/^/  /'
+    echo
+    local missing=0 s
+    while IFS= read -r s; do
+        [ -z "$s" ] && continue
+        grep -qxF "$s" <<< "$issued" || { echo "MISSING: $s"; missing=1; }
+    done < <(estate_sans)
+    [ "$missing" = 0 ] && echo "the issued cert covers every derived name."
+    echo
+    echo "certbot image: $(eval echo "\$$CERTBOT_IMAGE_VAR")   provider: $DNS_PROVIDER"
+}
+
+verify() {
+    echo "would run:"
+    echo "  echo | openssl s_client -connect ${DOMAIN}:443 -servername ${DOMAIN} 2>/dev/null \\"
+    echo "    | openssl x509 -noout -dates -ext subjectAltName"
+    echo
+    echo "read-only against a live host — announce and approve first (§7)."
+}
+
+refuse() {
+    echo "REFUSING: '$1' acts on a live cert and a live gateway." >&2
+    echo "  renew issues a real Let's Encrypt cert (rate-limited)." >&2
+    echo "  push rsyncs to the gateway and reloads nginx." >&2
+    echo "  Neither runs without explicit approval." >&2
+    exit 1
+}
+
+case "${1:-status}" in
+    status)      status ;;
+    verify)      verify ;;
+    renew|push)  refuse "$1" ;;
+    *) echo "usage: $0 [status|verify|renew|push]" >&2; exit 1 ;;
+esac

berth/ctrl/check.sh

@@ -0,0 +1,153 @@
+#!/usr/bin/env bash
+# Is this estate coherent? Reports and instructs; never fixes.
+#
+# Takes no subcommand — there is one question to ask.
+# Every check corresponds to something wrong in the estate today.
+
+set -euo pipefail
+cd "$(dirname "$0")"
+
+source ./lib/config.sh
+source ./lib/estate.sh
+load_config
+
+WORST=0
+note() { echo "  $*"; }
+warn() { echo "  WARN  $*"; [ "$WORST" -lt 1 ] && WORST=1; return 0; }
+bad()  { echo "  FAIL  $*"; WORST=2; return 0; }
+
+echo "estate: $ESTATE   target: $TARGET   domain: $DOMAIN"
+echo
+
+# 1. cert coverage: a wildcard matches exactly one label.
+echo "certs — does the cert cover every name the services serve?"
+issued=$(estate_get "certs.issued" | python3 -c 'import json,sys
+try: print("\n".join(json.load(sys.stdin)))
+except Exception: pass')
+if [ -z "$issued" ]; then
+    warn "no certs.issued in the estate file — cannot check coverage."
+else
+    while IFS=$'\x1f' read -r name host up kind raw placement peer port lhost; do
+        [ -z "$host" ] && continue
+        fqdn="${host}.${DOMAIN}"
+        # A '*' host stands for "any single label here" — check the deepest
+        # name it can produce, which is the one that fails.
+        probe="$fqdn"
+        case "$host" in \*.*) probe="anyroom.${host#\*.}.${DOMAIN}" ;; esac
+        covered=""
+        while IFS= read -r san; do
+            [ -z "$san" ] && continue
+            if san_covers "$probe" "$san"; then covered=1; break; fi
+        done <<< "$issued"
+        if [ -z "$covered" ]; then
+            bad "$name: '$probe' is covered by NO issued SAN"
+            note "      issued: $(echo "$issued" | tr '\n' ' ')"
+            note "      a wildcard matches exactly ONE label — reissue with"
+            note "      -d '*.${host#\*.}.${DOMAIN}' or move the name one level up"
+        fi
+    done < <(estate_services "$TARGET")
+    [ "$WORST" -lt 2 ] && note "every service name is covered."
+fi
+echo
+
+# 2. unmatched names: DNS and the cert are wildcard, nginx is exact, so
+#    without a :443 default_server the fallback is whichever vhost loads first.
+echo "gateway — is there a deliberate answer for unmatched names?"
+DEFAULT_CONF="${PPL_DIR:-$HOME/wdir/semester/ppl}/gateway/nginx/conf.d/default.conf"
+if [ ! -f "$DEFAULT_CONF" ]; then
+    note "ppl not on this machine at $DEFAULT_CONF — skipped."
+elif grep -qE '^\s*listen\s+443.*default_server' "$DEFAULT_CONF"; then
+    note "default.conf has a :443 default_server."
+else
+    bad "default.conf has NO :443 default_server."
+    note "      Unmatched names fall through to the first-loaded vhost."
+    note "      This is a PREREQUISITE for generating any config: adding a"
+    note "      generated include changes load order, and load order is what"
+    note "      currently decides the fallback."
+fi
+echo
+
+# 3. a rule allowing a port nothing listens on is dead config; a service no
+#    compose file declares is undocumented state. Needs both halves to see.
+echo "firewall — rules against listeners"
+estate_get "firewall" | python3 -c '
+import json,sys
+try: fw = json.load(sys.stdin)
+except Exception: fw = []
+for r in fw:
+    n = r.get("note")
+    print("  %-6s %-5s %s" % (r["port"], r.get("proto","tcp"), r.get("desc","")))
+    if n: print("        UNRESOLVED: " + n)
+'
+note "listener side: unknown until captured (ss -ltnp over ssh $HOST)."
+# Ask the structure, not the prose: the condition is "does a peer still lack a
+# public key", not "is there a _status string". _status is ALWAYS non-empty —
+# capture rewrites it to "CAPTURED ..." — so testing it for emptiness pinned
+# this warning on permanently, including after the capture it asks for.
+uncaptured=$(estate_get "vpn.overlays.estate.peers" 2>/dev/null | python3 -c '
+import json, sys
+try:
+    peers = json.load(sys.stdin)
+except Exception:
+    sys.exit(0)
+print(" ".join(n for n, p in peers.items() if not p.get("public_key")))
+' 2>/dev/null)
+if [ -n "$uncaptured" ]; then
+    warn "overlay: public keys not captured for:$uncaptured — see 'make vpn check'"
+    note "      10.8.0.1 carries the registry and woodpecker gRPC;"
+    note "      10.8.0.2 backs langfuse. Nothing in the tree creates the"
+    note "      interface — a fresh box cannot start the gateway compose."
+    note "      capture with: sudo wg show | make vpn capture --write"
+else
+    note "overlay: $(estate_get 'vpn._status')"
+fi
+note "overlay detail: make vpn show estate"
+echo
+
+# 4. ssh aliases, never hostnames: a bare hostname offers every agent key and
+#    trips MaxAuthTries before reaching the right one.
+echo "ssh — aliases, never hostnames"
+for var in HOST HOST_ADMIN; do
+    val="${!var:-}"
+    if [ -z "$val" ]; then
+        warn "$var is unset."
+    elif [[ "$val" == *.* ]]; then
+        bad "$var='$val' looks like a hostname, not a ~/.ssh/config alias."
+        note "      A bare hostname trips MaxAuthTries before reaching the key."
+    elif [ -f "$HOME/.ssh/config" ] && grep -qiE "^\s*Host\s+.*\b${val}\b" "$HOME/.ssh/config"; then
+        note "$var=$val — Host block present."
+    else
+        warn "$var='$val' has no matching Host block in ~/.ssh/config."
+    fi
+done
+for f in "$HOME/wdir/semester/ppl/ctrl/.env"; do
+    [ -f "$f" ] || continue
+    if grep -qE '^SERVER=.*\.' "$f"; then
+        warn "$f sets SERVER to a hostname, not an alias — every ppl script inherits it."
+    fi
+done
+echo
+
+# ── 5. toolchain ───────────────────────────────────────────────────────────
+echo "toolchain"
+for t in python3 "$TOFU_BIN" aws gcloud ssh rsync wg; do
+    if command -v "$t" >/dev/null 2>&1; then
+        note "$(printf '%-8s' "$t") present"
+    else
+        note "$(printf '%-8s' "$t") MISSING  — $(case $t in
+            tofu)   echo 'blocks: estate plan/apply; terraform works identically' ;;
+            wg)     echo 'blocks: vpn keygen and the overlay checks' ;;
+            aws)    echo 'blocks: the control-plane inventory, dns on route53' ;;
+            gcloud) echo 'blocks: the gcp estate' ;;
+            *)      echo 'blocks: most things' ;;
+        esac)"
+    fi
+done
+echo
+
+case "$WORST" in
+    0) echo "OK" ;;
+    1) echo "OK, with warnings" ;;
+    2) echo "PROBLEMS FOUND — see FAIL lines above" ;;
+esac
+exit 0

berth/ctrl/dns.sh

@@ -0,0 +1,81 @@
+#!/usr/bin/env bash
+# DNS records, over whichever provider the target names.
+#
+# Usage:
+#   ./dns.sh list
+#   ./dns.sh add <subdomain>       # <sub>.<domain> -> <domain>
+#   ./dns.sh add-wildcard <sub>
+#   ./dns.sh remove <subdomain>
+#
+# Read-only verbs announce and wait; mutating ones refuse.
+
+set -euo pipefail
+cd "$(dirname "$0")"
+
+source ./lib/config.sh
+source ./lib/estate.sh
+load_config
+
+announce() {
+    echo "would run:"
+    printf '  %s\n' "$*"
+}
+
+list() {
+    case "$DNS_PROVIDER" in
+        route53)
+            announce "aws route53 list-resource-record-sets --hosted-zone-id $AWS_HOSTED_ZONE_ID" \
+                     "--query 'ResourceRecordSets[].[Name,Type,TTL,ResourceRecords[0].Value]' --output table"
+            ;;
+        google)
+            announce "gcloud dns record-sets list --zone=${ESTATE}-zone --project=${GCP_PROJECT:-<unset>}"
+            ;;
+        *) echo "unknown DNS_PROVIDER: $DNS_PROVIDER" >&2; exit 1 ;;
+    esac
+    echo
+    echo "read-only, but NOT run: each batch is announced and"
+    echo "waited on. Approve it and it runs."
+}
+
+mutate() {
+    local verb="$1" sub="${2:-}"
+    [ -z "$sub" ] && { echo "usage: $0 $verb <subdomain>" >&2; exit 1; }
+    local name
+    case "$verb" in
+        add)          name="${sub}.${DOMAIN}" ;;
+        add-wildcard) name="*.${sub}.${DOMAIN}" ;;
+        remove)       name="${sub}.${DOMAIN}" ;;
+    esac
+    echo "$verb: $name  ->  $DOMAIN   (provider: $DNS_PROVIDER)"
+    echo
+
+    # The wildcard-depth rule again, applied BEFORE the record is created rather
+    # than discovered in a browser afterwards.
+    if [ "$verb" = "add" ]; then
+        local issued; issued=$(estate_get "certs.issued" | python3 -c 'import json,sys
+try: print("\n".join(json.load(sys.stdin)))
+except Exception: pass')
+        local covered=""
+        while IFS= read -r san; do
+            [ -z "$san" ] && continue
+            san_covers "$name" "$san" && { covered=1; break; }
+        done <<< "$issued"
+        [ -z "$covered" ] && {
+            echo "WARNING: '$name' is covered by no issued SAN." >&2
+            echo "         A wildcard matches ONE label. The record would" >&2
+            echo "         resolve and then fail TLS. Reissue the cert first." >&2
+            echo >&2
+        }
+    fi
+
+    echo "REFUSING: this changes live DNS." >&2
+    echo "  Nothing is created, modified or deleted on any account without" >&2
+    echo "  explicit approval for that specific action." >&2
+    exit 1
+}
+
+case "${1:-list}" in
+    list)                     list ;;
+    add|add-wildcard|remove)  mutate "$@" ;;
+    *) echo "usage: $0 [list|add <sub>|add-wildcard <sub>|remove <sub>]" >&2; exit 1 ;;
+esac

berth/ctrl/docs.sh

@@ -0,0 +1,27 @@
+#!/usr/bin/env bash
+# Documentation.
+#
+# Usage:
+#   ./docs.sh serve|graphs
+
+set -euo pipefail
+cd "$(dirname "$0")"
+
+source ./lib/config.sh
+source ./lib/estate.sh
+load_config
+
+case "${1:-serve}" in
+    serve)
+        echo "berth has no doc server yet — the README is the documentation."
+        echo "  $(cd .. && pwd)/README.md"
+        echo
+        echo "the estate, resolved:  make estate show"
+        ;;
+    graphs)
+        echo "not implemented. The estate file is the graph's source; a"
+        echo "renderer belongs with docgen/graphgen, which is another thread's"
+        echo "another thread's — so this is a handoff, not a stub to fill in here."
+        ;;
+    *) echo "usage: $0 [serve|graphs]" >&2; exit 1 ;;
+esac

berth/ctrl/env.d/aws.env

@@ -0,0 +1,14 @@
+# Target: AWS — the estate that actually runs today (mcrn.ar).
+TARGET_NAME=aws
+CLOUD=aws
+REGION=us-east-1
+INSTANCE_TYPE=t3.small
+
+# The Route53 hosted zone. It ALREADY EXISTS and is reused, never created —
+# see estate.sh's refusal to create a zone on this target.
+AWS_HOSTED_ZONE_ID=Z02279903503ZMIB5FC1N
+SSH_KEY_NAME=mcrn
+
+# certbot's DNS-01 plugin for this provider.
+CERTBOT_IMAGE_VAR=CERTBOT_AWS_IMAGE
+DNS_PROVIDER=route53

berth/ctrl/env.d/gcp.env

@@ -0,0 +1,17 @@
+# Target: GCP — the replica, on its own domain.
+#
+# The domain differs from AWS's on purpose: this target CREATES a DNS zone
+# where aws reuses one, so a shared domain would create a second authoritative
+# zone and break live DNS. The domain lives in estate/nrft.json, not here.
+TARGET_NAME=gcp
+CLOUD=gcp
+REGION=us-central1
+INSTANCE_TYPE=e2-small
+
+GCP_PROJECT=
+GCP_ZONE=us-central1-a
+
+# Delegation order: create the zone and let it answer first, then set the
+# nameservers at the registrar — it validates that they respond.
+CERTBOT_IMAGE_VAR=CERTBOT_GCP_IMAGE
+DNS_PROVIDER=google

berth/ctrl/estate.sh

@@ -0,0 +1,101 @@
+#!/usr/bin/env bash
+# The estate: what it is, what would change, and — behind a gate — changing it.
+#
+# Usage:
+#   ./estate.sh              # show
+#   ./estate.sh list         # every estate/*.json
+#   ./estate.sh plan         # tofu plan, read-only
+#   ./estate.sh apply --yes  # refuses without --yes
+#   ./estate.sh destroy --yes
+#
+# Why the default is read-only: ../README.md
+
+set -euo pipefail
+cd "$(dirname "$0")"
+
+source ./lib/config.sh
+source ./lib/estate.sh
+load_config
+
+show() {
+    echo "estate:   $ESTATE   ($ESTATE_FILE)"
+    echo "target:   $TARGET   (cloud=$CLOUD region=$REGION)"
+    echo "domain:   $DOMAIN"
+    echo "host:     $HOST   (sudo: $HOST_ADMIN)"
+    echo "workspace: $TOFU_WORKSPACE"
+    echo
+    local status; status=$(estate_get "_meta.status")
+    [ -n "$status" ] && echo "  !! $status" && echo
+
+    echo "services in scope for '$TARGET':"
+    local name host up kind raw
+    while IFS=$'\x1f' read -r name host up kind raw placement peer port lhost; do
+        [ -z "$name" ] && continue
+        printf '  %-12s %-14s %-24s %s%s\n' \
+            "$name" "${host:--}" "${up:--}" "$kind" \
+            "$([ -n "$raw" ] && echo '  [hand-written]')"
+    done < <(estate_services "$TARGET")
+
+    echo
+    echo "cert SANs (derived, not listed):"
+    estate_sans | sed 's/^/  /'
+}
+
+list() {
+    local f n
+    printf '%-12s %-16s %s\n' ESTATE DOMAIN STATUS
+    for f in ../estate/*.json; do
+        [ -f "$f" ] || continue
+        n=$(basename "$f" .json)
+        printf '%-12s %-16s %s%s\n' "$n" \
+            "$(python3 -c 'import json,sys;print(json.load(open(sys.argv[1])).get("domain",""))' "$f")" \
+            "$(python3 -c 'import json,sys;print(json.load(open(sys.argv[1])).get("_meta",{}).get("status",""))' "$f")" \
+            "$([ "$n" = "$ESTATE" ] && echo '   <- this one')"
+    done
+}
+
+# Read-only. Meaningful only once state is imported: against empty state,
+# plan reports "create N resources", which is not drift.
+plan() {
+    echo "== $TOFU_BIN plan =="
+    if ! command -v "$TOFU_BIN" >/dev/null; then
+        echo "   $TOFU_BIN not installed — skipped." >&2
+        echo "   OpenTofu is the MPL-2.0 fork; 'terraform' works identically." >&2
+    else
+        echo "   would run: $TOFU_BIN plan -var-file=<(estate)"
+    fi
+    echo
+    echo "NOTE: not wired up yet, and that is the point. tofu plan against empty"
+    echo "      state reports \"create N resources\" — which is not drift, it is an"
+    echo "      empty state. It becomes the check that proves the description"
+    echo "      matches reality only once state is IMPORTED from the inventory."
+    echo "      ppl/infra/ describes an aspiration: it was never applied."
+}
+
+# The gate. Two things have to be true: --yes present, AND the plan shown first.
+require_yes() {
+    local verb="$1"; shift
+    local yes=""
+    for a in "$@"; do [ "$a" = "--yes" ] && yes=1; done
+    if [ -z "$yes" ]; then
+        echo "refusing to $verb without --yes." >&2
+        echo >&2
+        echo "  $verb changes a live, billable estate and can take DNS with it." >&2
+        echo "  Read the plan first:  make estate plan" >&2
+        echo "  Then:                 ./ctrl/estate.sh $verb --yes" >&2
+        exit 1
+    fi
+    echo "refusing to $verb: not implemented, and deliberately so." >&2
+    echo "  The executor is not wired up, and nothing is imported yet, so" >&2
+    echo "  there is nothing truthful to apply." >&2
+    exit 1
+}
+
+case "${1:-show}" in
+    show)    show ;;
+    list)    list ;;
+    plan)    plan ;;
+    apply)   shift; require_yes apply "$@" ;;
+    destroy) shift; require_yes destroy "$@" ;;
+    *)       echo "usage: $0 [show|list|plan|apply --yes|destroy --yes]" >&2; exit 1 ;;
+esac

berth/ctrl/host.sh

@@ -0,0 +1,63 @@
+#!/usr/bin/env bash
+# The remote box — the other half of the inventory.
+#
+# Usage:
+#   ./host.sh status|ports|services
+#
+# Announces what it would run over `ssh $HOST` and does not run it. Always the
+# ssh alias, never a hostname: there is no `Host mcrn.ar` block, so a bare
+# hostname offers every agent key and trips MaxAuthTries.
+
+set -euo pipefail
+cd "$(dirname "$0")"
+
+source ./lib/config.sh
+source ./lib/estate.sh
+load_config
+
+guard_alias() {
+    if [[ "$HOST" == *.* ]]; then
+        echo "HOST='$HOST' is a hostname, not a ~/.ssh/config alias. Refusing." >&2
+        exit 1
+    fi
+}
+
+announce_batch() {
+    echo "would run, over 'ssh $HOST':"
+    printf '  %s\n' "$@"
+    echo
+    echo "read-only, and NOT run. Announce-first applies to EACH batch, not"
+    echo "once per session — so the commands can be read and"
+    echo "learned rather than scrolled past."
+}
+
+guard_alias
+case "${1:-status}" in
+    status)
+        announce_batch \
+            "uname -a; uptime; df -h /" \
+            "docker ps --format '{{.Names}}\t{{.Image}}\t{{.Ports}}'" \
+            "docker network inspect gateway --format '{{range .Containers}}{{.Name}} {{end}}'" \
+            "systemctl list-units --type=service --state=running --no-pager" \
+            "systemctl list-timers --no-pager"
+        echo
+        echo "sudo-only, over 'ssh $HOST_ADMIN' and only where genuinely needed:"
+        echo "  wg show           # the WireGuard peers that exist nowhere in the tree"
+        ;;
+    ports)
+        announce_batch "ss -ltnp"
+        echo "the estate declares these firewall rules:"
+        estate_get "firewall" | python3 -c '
+import json,sys
+for r in json.load(sys.stdin):
+    print("  %-6s %-5s %s" % (r["port"], r.get("proto","tcp"), r.get("desc","")))'
+        ;;
+    services)
+        announce_batch "docker compose -f ~/ppl/gateway/docker-compose.yml ps"
+        echo "the estate declares $(estate_services "$TARGET" | grep -c . ) service(s) for target '$TARGET'."
+        echo "the gateway compose declares 8. The difference is sibling repos'"
+        echo "stacks joining the shared 'gateway' network — intended design, but"
+        echo "nothing in the tree lists it. The inventory produces that list."
+        ;;
+    *) echo "usage: $0 [status|ports|services]" >&2; exit 1 ;;
+esac

berth/ctrl/lib/config.sh

@@ -0,0 +1,94 @@
+# Shared config loading. Sourced, never executed. Run from ctrl/.
+#
+# Precedence, weakest first:
+#   ctrl/versions.env        pinned toolchain          (committed)
+#   ctrl/env.d/<target>.env  provider shape: aws|gcp   (committed)
+#   ctrl/.env                machine-local + secrets   (gitignored)
+#   the caller's env         `make estate plan TARGET=gcp`  (always wins)
+
+CONFIG_OVERRIDABLE="TARGET ESTATE CLOUD REGION INSTANCE_TYPE
+                    HOST HOST_ADMIN AWS_PROFILE AWS_HOSTED_ZONE_ID
+                    GCP_PROJECT GCP_ZONE TOFU_WORKSPACE"
+
+# rig's port formula, reproduced rather than imported. cksum because it is
+# POSIX and gives the same value on every machine. Used to verify, not allocate.
+derive_port_base() {
+    local h; h=$(printf '%s' "$1" | cksum | awk '{print $1}')
+    echo $((20000 + (h % 200) * 10))
+}
+
+_config_restore() {
+    local line
+    while IFS= read -r line; do
+        if [ -n "$line" ]; then
+            eval "export $line"
+        fi
+    done <<< "$1"
+    # A while loop returns its last body command's status; the trailing empty
+    # line would otherwise make this return 1 and trip `set -e` in the caller.
+    return 0
+}
+
+load_config() {
+    local k saved=""
+    for k in $CONFIG_OVERRIDABLE; do
+        # ${!k+x} distinguishes "set but empty" from "unset" — an explicit
+        # FOO= on the command line is a real choice and must survive.
+        if [ -n "${!k+x}" ]; then
+            saved+="$k=$(printf '%q' "${!k}")"$'\n'
+        fi
+    done
+
+    set -a
+    source ./versions.env
+    [ -f ./.env ] && source ./.env
+    set +a
+
+    # Re-apply overrides now so TARGET is the caller's before we pick the file.
+    _config_restore "$saved"
+
+    local target="${TARGET:-aws}"
+    if [ ! -f "./env.d/${target}.env" ]; then
+        echo "no such target: env.d/${target}.env" >&2
+        echo "available: $(ls env.d/*.env 2>/dev/null | xargs -n1 basename | sed 's/\.env$//' | tr '\n' ' ')" >&2
+        exit 1
+    fi
+
+    set -a
+    source "./env.d/${target}.env"
+    [ -f ./.env ] && source ./.env
+    set +a
+
+    _config_restore "$saved"
+
+    TARGET="$target"
+
+    # Identity is explicit: berth never guesses which estate it is acting on.
+    # The one convenience: a single estate/*.json is used without being asked.
+    if [ -z "${ESTATE:-}" ]; then
+        local n; n=$(ls ../estate/*.json 2>/dev/null | wc -l)
+        if [ "$n" = "1" ]; then
+            ESTATE=$(basename "$(ls ../estate/*.json)" .json)
+        else
+            echo "ESTATE is not set and estate/ holds $n candidates — refusing to guess." >&2
+            echo "available: $(ls ../estate/*.json 2>/dev/null | xargs -n1 basename | sed 's/\.json$//' | tr '\n' ' ')" >&2
+            echo "set it: make estate show ESTATE=<name>, or ESTATE= in ctrl/.env" >&2
+            exit 1
+        fi
+    fi
+
+    ESTATE_FILE="../estate/${ESTATE}.json"
+    if [ ! -f "$ESTATE_FILE" ]; then
+        echo "no such estate: estate/${ESTATE}.json" >&2
+        echo "available: $(ls ../estate/*.json 2>/dev/null | xargs -n1 basename | sed 's/\.json$//' | tr '\n' ' ')" >&2
+        exit 1
+    fi
+
+    # Facts come from the estate file, never restated in a target env.
+    DOMAIN=$(estate_get "domain")
+    HOST="${HOST:-$(estate_get "host")}"
+    HOST_ADMIN="${HOST_ADMIN:-$(estate_get "host_admin")}"
+
+    # Workspace == target, so the two can never mean different things.
+    TOFU_WORKSPACE="${TOFU_WORKSPACE:-$TARGET}"
+}

berth/ctrl/lib/estate.sh

@@ -0,0 +1,147 @@
+# Reading and projecting estate/<name>.json. Sourced, never executed.
+#
+# python3 rather than jq: berth's floor already includes python3, so it is a
+# dependency berth has rather than one it adds.
+
+estate_get() {
+    python3 -c '
+import json, sys
+d = json.load(open(sys.argv[1]))
+for k in sys.argv[2].split("."):
+    if isinstance(d, list):
+        try: k = int(k)
+        except ValueError: sys.exit(0)
+    try: d = d[k]
+    except Exception: sys.exit(0)
+print("" if d is None else d if isinstance(d, str) else json.dumps(d))
+' "$ESTATE_FILE" "$1"
+}
+
+# Services in scope for one target. A service names its targets; absent = all.
+# Fields are US-separated (0x1f), not tab: tab is IFS whitespace, so bash
+# collapses a run of them and an empty field would shift every later column.
+estate_services() {
+    local target="${1:-$TARGET}"
+    python3 -c '
+import json, sys
+d = json.load(open(sys.argv[1]))
+target = sys.argv[2]
+for s in d.get("services", []):
+    tg = s.get("targets")
+    if tg is not None and target not in tg:
+        continue
+    print("\x1f".join([
+        s.get("name", ""),
+        s.get("host", ""),
+        str(s.get(target + "_upstream", s.get("upstream", "")) or ""),
+        s.get("kind", "proxy"),
+        "raw" if s.get("raw") else "",
+        s.get("placement", "box"),
+        s.get("peer", ""),
+        str(s.get("port", "") or ""),
+        s.get("local_host", s.get("host", "")),
+    ]))
+' "$ESTATE_FILE" "$target"
+}
+
+# The SAN list the services need, derived — never a literal list.
+estate_sans() {
+    python3 -c '
+import json, sys
+d = json.load(open(sys.argv[1]))
+domain = d["domain"]
+sans = [domain]
+depths = set()
+for s in d.get("services", []):
+    h = s.get("host", "")
+    if not h:
+        continue
+    # A wildcard matches exactly ONE label. "git" needs *.domain; "dlt.spr"
+    # needs *.spr.domain. The parent of the leaf is what has to be covered.
+    parent = h.split(".", 1)[1] if "." in h else ""
+    depths.add(parent)
+for p in sorted(depths):
+    sans.append("*." + (p + "." if p else "") + domain)
+for s in sans:
+    print(s)
+' "$ESTATE_FILE"
+}
+
+# Is <fqdn> covered by <san>? A wildcard matches exactly one label.
+san_covers() {
+    local fqdn="$1" san="$2"
+    [ "$fqdn" = "$san" ] && return 0
+    case "$san" in
+        \*.*)
+            local suffix="${san#\*.}"
+            # Must end in .suffix AND have exactly one extra label.
+            case "$fqdn" in
+                *".$suffix") [ "${fqdn%".$suffix"}" = "${fqdn%%.*}" ] && return 0 ;;
+            esac
+            ;;
+    esac
+    return 1
+}
+
+# ── the overlay ────────────────────────────────────────────────────────────
+
+# Every overlay name, one per line.
+overlay_names() {
+    python3 -c '
+import json, sys
+d = json.load(open(sys.argv[1]))
+for n in d.get("vpn", {}).get("overlays", {}):
+    print(n)
+' "$ESTATE_FILE"
+}
+
+# Peers of one overlay, US-separated:
+#   name, address, role, endpoint, public_key, allowed_ips, keepalive
+overlay_peers() {
+    python3 -c '
+import json, sys
+d = json.load(open(sys.argv[1]))
+ov = d.get("vpn", {}).get("overlays", {}).get(sys.argv[2], {})
+for name, p in ov.get("peers", {}).items():
+    print("\x1f".join(str(x) if x is not None else "" for x in [
+        name, p.get("address"), p.get("role"), p.get("endpoint"),
+        p.get("public_key"), p.get("allowed_ips"), p.get("keepalive"),
+    ]))
+' "$ESTATE_FILE" "$1"
+}
+
+overlay_get() { estate_get "vpn.overlays.$1.$2"; }
+
+# Is an address inside a CIDR? Pure python so there is no ipcalc dependency —
+# berth's floor already includes python3 because the IaC side needs it.
+addr_in_subnet() {
+    python3 -c '
+import ipaddress, sys
+try:
+    sys.exit(0 if ipaddress.ip_address(sys.argv[1]) in ipaddress.ip_network(sys.argv[2], strict=False) else 1)
+except ValueError:
+    sys.exit(2)
+' "$1" "$2"
+}
+
+# The upstream a service actually resolves to, as "host:port".
+#
+# A PLACED service has no literal `upstream` field: ✖ B9 replaced langfuse's
+# hand-written `10.8.0.2:3000` with placement+peer+port, because being reached
+# by address on the overlay is ONE decision, not three properties. Everything
+# that asks "what does this service point at" must therefore resolve it the
+# same way, or it silently sees an empty string and skips the service — which
+# is exactly how vpn.sh's bindings invariant went quiet after B9 landed.
+#
+# usage: service_upstream <up> <placement> <peer> <port>
+service_upstream() {
+    local up="$1" placement="$2" peer="$3" port="$4"
+    case "$placement" in
+        local|instance)
+            local addr; addr="$(overlay_get estate "peers.${peer}.address")"
+            [ -z "$addr" ] && return 1
+            printf '%s:%s' "$addr" "$port"
+            ;;
+        *) printf '%s' "$up" ;;
+    esac
+}

berth/ctrl/ports.sh

@@ -0,0 +1,78 @@
+#!/usr/bin/env bash
+# The local port map, and whether it still agrees with rig.
+#
+# Usage:
+#   ./ports.sh show     # DERIVED / ACTIVE / SOURCE
+#   ./ports.sh verify   # recompute rig's formula, report drift
+#
+# berth recomputes rig's port formula rather than importing it, so neither
+# depends on the other. See ../README.md.
+
+set -euo pipefail
+cd "$(dirname "$0")"
+
+source ./lib/config.sh
+source ./lib/estate.sh
+load_config
+
+# name<US>host<US>local_port for everything that has one.
+_local_ports() {
+    python3 -c '
+import json, sys
+d = json.load(open(sys.argv[1]))
+for s in d.get("services", []):
+    p = s.get("local_port")
+    if p:
+        print("\x1f".join([s.get("name",""), s.get("host",""), str(p)]))
+' "$ESTATE_FILE"
+}
+
+show() {
+    local ld; ld=$(estate_get "local_domain"); : "${ld:=local.ar}"
+    printf '%-12s %-22s %-8s %-8s %s\n' NAME ADDRESS ACTIVE DERIVED SOURCE
+    local name host port base
+    while IFS=$'\x1f' read -r name host port; do
+        [ -z "$name" ] && continue
+        base=$(derive_port_base "$host")
+        if [ "$port" = "$base" ]; then
+            printf '%-12s %-22s %-8s %-8s %s\n' "$name" "${host}.${ld}" "$port" "$base" "derived"
+        else
+            printf '%-12s %-22s %-8s %-8s %s\n' "$name" "${host}.${ld}" "$port" "$base" "override"
+        fi
+    done < <(_local_ports)
+    echo
+    echo "DERIVED is what rig's formula gives for that name. ACTIVE is what the"
+    echo "estate records. 'override' is not an error — most of these were never"
+    echo "rigs. 'make ports verify' says which ones should have matched."
+}
+
+verify() {
+    local name host port base rc=0 checked=0
+    while IFS=$'\x1f' read -r name host port; do
+        [ -z "$name" ] && continue
+        # Only 20000-21999 is rig's to predict; anything else was never derived.
+        if [ "$port" -lt 20000 ] || [ "$port" -gt 21999 ]; then
+            continue
+        fi
+        checked=$((checked + 1))
+        base=$(derive_port_base "$host")
+        if [ "$port" != "$base" ]; then
+            echo "DRIFT  $name (${host}): estate says $port, rig's formula gives $base"
+            echo "       either the rig pinned HTTP_PORT in its ctrl/.env, or the"
+            echo "       folder was renamed. The Caddy map is stale either way."
+            rc=1
+        else
+            echo "ok     $name (${host}): $port"
+        fi
+    done < <(_local_ports)
+    echo
+    echo "checked $checked rig-shaped port(s) in 20000-21999."
+    [ "$rc" = 0 ] && echo "no drift." || echo "drift found — regenerate with 'make services render local'."
+    return 0
+}
+
+case "${1:-show}" in
+    show)   show ;;
+    verify) verify ;;
+    *)      echo "usage: $0 [show|verify]" >&2; exit 1 ;;
+esac

berth/ctrl/registry.sh

@@ -0,0 +1,33 @@
+#!/usr/bin/env bash
+# The image registry — remote, and reachable only over the overlay.
+#
+# Usage:
+#   ./registry.sh status
+#
+# One verb: berth reports on a registry running on someone else's box. Starting
+# and stopping it is that box's business.
+
+set -euo pipefail
+cd "$(dirname "$0")"
+
+source ./lib/config.sh
+source ./lib/estate.sh
+load_config
+
+status() {
+    local wg_server; wg_server=$(overlay_get estate "peers.box.address")
+    echo "registry:  registry.${DOMAIN}   (public pull via /v2/)"
+    echo "push:      ${wg_server}:5000    (WireGuard-only, not in the firewall)"
+    echo
+    echo "would run:"
+    echo "  curl -s https://registry.${DOMAIN}/v2/_catalog"
+    echo
+    echo "NOTE: the push endpoint binds ${wg_server}, and $(estate_get 'vpn._status')."
+    echo "      A freshly-provisioned box cannot start the gateway compose file"
+    echo "      at all, because that bind fails."
+}
+
+case "${1:-status}" in
+    status) status ;;
+    *) echo "usage: $0 [status]" >&2; exit 1 ;;
+esac

berth/ctrl/render/nginx-proxy.tmpl

@@ -0,0 +1,32 @@
+# ${NAME} — GENERATED by berth from estate/${ESTATE}.json. Do not edit.
+# Edit the estate file and re-run: make services render aws
+
+server {
+    listen 80;
+    server_name ${FQDN};
+    return 301 https://$host$request_uri;
+}
+
+server {
+    listen 443 ssl;
+    server_name ${FQDN};
+
+    ssl_certificate     /etc/nginx/certs/live/${DOMAIN}/fullchain.pem;
+    ssl_certificate_key /etc/nginx/certs/live/${DOMAIN}/privkey.pem;
+
+    # Docker's embedded DNS. Naming the upstream in a VARIABLE forces runtime
+    # resolution, so nginx STARTS even when the upstream container is absent.
+    # With a literal proxy_pass, one stopped container takes the whole gateway
+    # down at reload — which is what makes one nginx able to front a dozen
+    # independent compose stacks.
+    resolver 127.0.0.11 valid=30s;
+
+    location / {
+        set $upstream_${NAME} ${UPSTREAM_HOST};
+        proxy_pass http://$upstream_${NAME}:${UPSTREAM_PORT};
+        proxy_set_header Host              $host;
+        proxy_set_header X-Real-IP         $remote_addr;
+        proxy_set_header X-Forwarded-For   $proxy_add_x_forwarded_for;
+        proxy_set_header X-Forwarded-Proto $scheme;
+    }
+}

berth/ctrl/render/nginx-static.tmpl

@@ -0,0 +1,23 @@
+# ${NAME} — GENERATED by berth from estate/${ESTATE}.json. Do not edit.
+# Edit the estate file and re-run: make services render aws
+
+server {
+    listen 80;
+    server_name ${FQDN};
+    return 301 https://$host$request_uri;
+}
+
+server {
+    listen 443 ssl;
+    server_name ${FQDN};
+
+    ssl_certificate     /etc/nginx/certs/live/${DOMAIN}/fullchain.pem;
+    ssl_certificate_key /etc/nginx/certs/live/${DOMAIN}/privkey.pem;
+
+    root /usr/share/nginx/html/${NAME};
+    index index.html;
+
+    location / {
+        try_files $uri $uri/ =404;
+    }
+}

berth/ctrl/render/nginx-upstream.tmpl

@@ -0,0 +1,32 @@
+# ${NAME} — GENERATED by berth from estate/${ESTATE}.json. Do not edit.
+# Placement: ${PLACEMENT} (${PEER}) — reached over the overlay, not the docker network.
+
+upstream ${NAME}_backend {
+    server ${UPSTREAM_HOST}:${UPSTREAM_PORT};
+}
+
+server {
+    listen 80;
+    server_name ${FQDN};
+    return 301 https://$host$request_uri;
+}
+
+server {
+    listen 443 ssl;
+    server_name ${FQDN};
+
+    ssl_certificate /etc/nginx/certs/live/${DOMAIN}/fullchain.pem;
+    ssl_certificate_key /etc/nginx/certs/live/${DOMAIN}/privkey.pem;
+
+    # No `resolver`, and no `set $var` indirection — deliberately. Those exist so
+    # nginx starts when a CONTAINER is absent; this upstream is a literal address
+    # on the overlay, which needs no DNS at all. The three properties are one
+    # decision, and placement is what decides them.
+    location / {
+        proxy_pass http://${NAME}_backend;
+        proxy_set_header Host              $host;
+        proxy_set_header X-Real-IP         $remote_addr;
+        proxy_set_header X-Forwarded-For   $proxy_add_x_forwarded_for;
+        proxy_set_header X-Forwarded-Proto $scheme;
+    }
+}

berth/ctrl/selftest.sh

@@ -0,0 +1,210 @@
+#!/usr/bin/env bash
+# What berth has settled, and what it has withdrawn, written down as assertions.
+#
+# Two halves:
+#   - decisions that hold. Failing one means "you are about to undo this".
+#   - every entry in ../STALE.md. Failing one means a withdrawn assumption came
+#     back. That is the half that makes STALE.md an audit surface and not an
+#     archive — a retraction nobody re-reads is a retraction that decays.
+#
+# Scope: no cloud, no ssh, no sudo, no network. Cheap enough to actually run.
+# `make check` reports on the world and never fails; this exits 1, like rig's.
+#
+# Usage: make selftest   (or: bash ctrl/selftest.sh)
+set -uo pipefail          # NOT -e: one failing check must not abort the rest
+cd "$(dirname "$0")"
+
+source ./lib/config.sh
+
+rc=0
+passed=0
+check() {   # name, expected, actual
+    if [ "$2" = "$3" ]; then
+        printf '  ok    %s\n' "$1"
+        passed=$((passed + 1))
+    else
+        printf '  FAIL  %s\n        expected: %s\n        got:      %s\n' "$1" "$2" "$3"
+        rc=1
+    fi
+}
+note() { printf '\n%s\n' "$1"; }
+skip() { printf '  skip  %s (%s)\n' "$1" "$2"; }
+
+# An absence check must not match the files that RECORD the absence. STALE.md
+# names every withdrawn thing by definition, and this file names them again to
+# assert them — so both are excluded, or every check fails on itself. rig hits
+# the same wall and assembles its pattern from fragments for the same reason.
+NOSELF="--exclude=selftest.sh --exclude=STALE.md"
+absent() { grep -rIl $NOSELF "$@" 2>/dev/null | wc -l; }
+
+# A throwaway estate, for the checks that have to run berth rather than read it.
+TMP_ESTATE=_selftest
+cleanup() { rm -f "../estate/${TMP_ESTATE}.json"; }
+trap cleanup EXIT
+
+note "the withdrawn assumptions — ../STALE.md, one check each"
+
+# B1 — Pulumi. The two surviving mentions are historical fact about ppl/infra
+# and live in README.md and the estate, not in anything that runs.
+check "B1  no pulumi in the code" "0" "$(absent -i pulumi . ../Makefile)"
+
+# B2 — the ctlptl precedent. Withdrawn; the argument stands on its own now.
+check "B2  the withdrawn precedent is cited nowhere" "0" "$(absent -i ctlptl ..)"
+
+# B3 — `wg show <if> dump` leaks the private key in field 1. Stating only
+# show-vs-showconf makes the dump form read as safe.
+check "B3  all three wg forms are named" "yes" \
+    "$(grep -q 'dump' vpn.sh && grep -q 'showconf' vpn.sh && echo yes || echo no)"
+check "B3  capture refuses showconf-shaped input" "1" \
+    "$(printf '[Interface]\nPrivateKey = x\n' | bash vpn.sh capture >/dev/null 2>&1; echo $?)"
+check "B3  capture refuses dump-shaped input" "1" \
+    "$(printf 'priv\tpub\t51820\toff\n' | bash vpn.sh capture >/dev/null 2>&1; echo $?)"
+
+# B4 — keepalive belongs to the peer that DIALS, not the one that roams. The
+# first version warned on a correctly configured overlay, so the check is run
+# against one: hub carries the keepalive, nrft roams.
+python3 - <<'PY'
+import json, collections
+d = json.load(open("../estate/mcrn.json"), object_pairs_hook=collections.OrderedDict)
+d["vpn"]["overlays"]["estate"]["peers"]["box"]["keepalive"] = 25
+json.dump(d, open("../estate/_selftest.json", "w"), indent=2, ensure_ascii=False)
+PY
+check "B4  a correct overlay raises no keepalive warning" "0" \
+    "$(ESTATE=$TMP_ESTATE bash vpn.sh check 2>/dev/null | grep -ci 'no peer entry carries')"
+
+# B6 — public keys are 44-char base64 too, so shape alone would flag correct
+# data. Same fixture, with a real-shaped public key on a peer.
+python3 - <<'PY'
+import base64, collections, json, os
+d = json.load(open("../estate/_selftest.json"), object_pairs_hook=collections.OrderedDict)
+d["vpn"]["overlays"]["estate"]["peers"]["box"]["public_key"] = base64.b64encode(os.urandom(32)).decode()
+json.dump(d, open("../estate/_selftest.json", "w"), indent=2, ensure_ascii=False)
+PY
+check "B6  a public key does not trip the secret check" "0" \
+    "$(ESTATE=$TMP_ESTATE bash vpn.sh check 2>/dev/null | grep -c 'FAIL.*key')"
+
+cleanup   # the fixture is done with; two estate files would make load_config
+          # refuse to guess below, which is right but reads as a config failure
+
+# B5 — the pass-through block must be LAST, or a subcommand that names a real
+# target runs that target too. Checked through make, not by reading the file.
+note "B5  a subcommand that names a target dispatches once"
+for combo in "host ports" "host services" "vpn check" "vpn show estate" "estate show"; do
+    check "    make $combo" "1" \
+        "$(cd .. && make -n $combo 2>/dev/null | grep -c 'bash ctrl/')"
+done
+
+# B7 — the overlay moved out of network.wireguard into a top-level vpn block.
+check "B7  nothing reads network.wireguard" "0" "$(absent 'network\.wireguard' .)"
+
+# B8 — peers, not relatives. berth sources nothing from rig.
+check "B8  berth sources nothing from rig" "0" "$(absent -E 'rig/ctrl|\.\./rig' .)"
+
+# B9 — langfuse was filed as an exception a template could not express. It was
+# the general case. The proof is a live route: render it and diff against the
+# hand-written file, normalised for comments and whitespace.
+LIVE=/home/mariano/wdir/semester/ppl/gateway/nginx/conf.d/langfuse.conf
+if [ -f "$LIVE" ]; then
+    norm() { sed -e 's/#.*//' -e 's/[[:space:]]\+/ /g' -e 's/^ //' -e 's/ $//' -e '/^$/d' "$1"; }
+    bash services.sh render aws >/dev/null 2>&1
+    check "B9  the generated vhost reproduces the live one" "same" \
+        "$(diff -q <(norm ./render/out/aws/langfuse.conf) <(norm "$LIVE") >/dev/null 2>&1 \
+            && echo same || echo different)"
+else
+    skip "B9  generated vhost matches the live one" "ppl not on this machine"
+fi
+
+note "the safety contract — berth's verbs are not all safe"
+
+check "estate defaults to show"   "show"   "$(cd .. && make -n estate 2>/dev/null | grep -oE 'estate\.sh [a-z]+' | awk '{print $2}')"
+check "certs defaults to status"  "status" "$(cd .. && make -n certs  2>/dev/null | grep -oE 'certs\.sh [a-z]+'  | awk '{print $2}')"
+check "dns defaults to list"      "list"   "$(cd .. && make -n dns    2>/dev/null | grep -oE 'dns\.sh [a-z]+'    | awk '{print $2}')"
+check "vpn defaults to list"      "list"   "$(cd .. && make -n vpn    2>/dev/null | grep -oE 'vpn\.sh [a-z]+'    | awk '{print $2}')"
+
+for verb in apply destroy; do
+    check "estate $verb refuses without --yes" "1" \
+        "$(bash estate.sh "$verb" >/dev/null 2>&1; echo $?)"
+done
+for verb in renew push; do
+    check "certs $verb refuses"                "1" \
+        "$(bash certs.sh "$verb" >/dev/null 2>&1; echo $?)"
+done
+check "dns add refuses to change live DNS" "1" \
+    "$(bash dns.sh add selftest >/dev/null 2>&1; echo $?)"
+check "vpn up refuses without --yes" "1" \
+    "$(bash vpn.sh up >/dev/null 2>&1; echo $?)"
+
+note "config — the caller's env beats the files"
+
+# Generated from CONFIG_OVERRIDABLE, so a new key enrols itself.
+test_value() {
+    case "$1" in
+        TARGET)  echo "gcp" ;;
+        ESTATE)  echo "mcrn" ;;
+        *)       echo "selftest-sentinel" ;;
+    esac
+}
+for key in $CONFIG_OVERRIDABLE; do
+    want="$(test_value "$key")"
+    got="$(export "$key=$want"; load_config >/dev/null 2>&1; echo "${!key}")"
+    check "  caller's $key wins" "$want" "$got"
+done
+
+note "rig agreement — recomputed, never imported"
+
+# rig pins these same constants in its own selftest. Both arrive at them from
+# the same formula with no shared code, which is the coupling rule made testable.
+check "derive_port_base rig"     "20310" "$(derive_port_base rig)"
+check "derive_port_base foo"     "21690" "$(derive_port_base foo)"
+check "derive_port_base my-proj" "21030" "$(derive_port_base my-proj)"
+
+note "containment — berth writes nothing outside berth/"
+
+check "no tracked change outside berth/" "0" \
+    "$(cd ../.. && git status --porcelain 2>/dev/null | grep -vc '^.. berth/')"
+check "generated output is ignored" "yes" \
+    "$(cd .. && git check-ignore -q ctrl/render/out && echo yes || echo no)"
+# A trailing-slash pattern matches directories only, so ask about a path
+# inside it rather than the (not-yet-existing) directory itself.
+check "key material is ignored" "yes" \
+    "$(cd .. && git check-ignore -q ctrl/.secrets/vpn/any.key && echo yes || echo no)"
+
+note "capture and the checks that read it — three bugs found by running, 2026-09-14"
+
+# 1. A placed service's upstream is DERIVED (✖ B9). Anything reading the raw
+#    `upstream` field sees "" and skips it — which is how vpn.sh's bindings
+#    invariant, the "my configurations broke" detector, went quiet the day
+#    placement landed while still printing OK. Vacuous passes are the failure
+#    mode this whole file exists to catch.
+check "a placed service resolves to a real upstream" "10.8.0.2:3000" \
+    "$(bash -c 'source ./lib/config.sh; source ./lib/estate.sh; load_config >/dev/null;
+       service_upstream "" local nrft 3000')"
+check "bindings actually inspects a service" "1" \
+    "$(bash ./vpn.sh check 2>/dev/null | grep -c 'no service currently has an overlay address' \
+       | awk '{print 1-$1}')"
+
+# 2. A listen port belongs to a PEER. The roaming peer's is an ephemeral source
+#    port; writing it to the overlay renames the port the firewall rule is
+#    checked against — silently, since both are plausible integers.
+check "a roaming peer's port is not the overlay's port" "51820" \
+    "$(python3 -c 'import json;print(json.load(open("../estate/mcrn.json"))["vpn"]["overlays"]["estate"]["listen_port"])')"
+
+# 3. _status is always non-empty — capture rewrites it rather than clearing it —
+#    so a warning gated on "is it set" can never turn off, including after the
+#    capture it asks for. Gate on the structure instead.
+check "the capture warning clears once keys are in" "0" \
+    "$(bash ./check.sh 2>/dev/null | grep -c 'public keys not captured')"
+
+note "every STALE entry has a check here"
+
+# Not "$0": line 15 cd's into this script's directory, so a relative $0 no
+# longer resolves. After the cd the file is simply selftest.sh.
+# Ids are counted wherever they appear — B5's sits in a note(), not a check name.
+entries="$(grep -c '^\*\*✖ B' ../STALE.md)"
+checked="$(grep -oE '\bB[1-9][0-9]?\b' selftest.sh | sort -u | wc -l)"
+check "STALE.md entries are all covered" "$entries" "$checked"
+
+printf '\n%d passed' "$passed"
+[ "$rc" -ne 0 ] && printf ', SOME FAILED'
+printf '\n'
+exit "$rc"

berth/ctrl/services.sh

@@ -0,0 +1,183 @@
+#!/usr/bin/env bash
+# Gateway routes, projected from the estate onto one target.
+#
+# Usage:
+#   ./services.sh                 # list
+#   ./services.sh render aws      # -> render/out/aws/*.conf   (nginx vhosts)
+#   ./services.sh render local    # -> render/out/local/Caddyfile
+#   ./services.sh deploy          # refuses; ppl/ctrl/deploy.sh ships config
+#
+# Each target is a projection with its own rules, not a format conversion.
+# The nine axes they disagree on, and the install order: ../README.md
+
+set -euo pipefail
+cd "$(dirname "$0")"
+
+source ./lib/config.sh
+source ./lib/estate.sh
+load_config
+
+OUT_ROOT="./render/out"
+
+list() {
+    printf '%-12s %-16s %-22s %-9s %s\n' NAME FQDN UPSTREAM PLACEMENT SOURCE
+    local name host up kind raw
+    while IFS=$'\x1f' read -r name host up kind raw placement peer port lhost; do
+        [ -z "$name" ] && continue
+        # A placed service has no literal `upstream` — it is derived from the
+        # peer's overlay address, so show what it actually resolves to.
+        local shown; shown="$(service_upstream "$up" "$placement" "$peer" "$port")" || shown=""
+        printf '%-12s %-16s %-22s %-9s %s\n' \
+            "$name" "${host}.${DOMAIN}" "${shown:--}" "$placement" \
+            "$([ -n "$raw" ] && echo 'hand-written' || echo 'generated')"
+    done < <(estate_services "$TARGET")
+    echo
+    echo "hand-written entries are NOT generated and NOT overwritten."
+    echo "run 'make estate show' to see why each one is an exception."
+}
+
+render_cloud() {
+    # Two statements: `local a="$1" b="$a"` expands all arguments before any
+    # assignment, so $a would still be unset.
+    local target="$1"
+    local out="$OUT_ROOT/$target"
+    rm -rf "$out"; mkdir -p "$out"
+    local name host up kind raw uhost uport n=0 skipped=0
+
+    while IFS=$'\x1f' read -r name host up kind raw placement peer port lhost; do
+        [ -z "$name" ] && continue
+        if [ -n "$raw" ]; then
+            skipped=$((skipped + 1))
+            continue
+        fi
+        # Placement picks the rendering. A container on the estate's own network
+        # is reached by NAME through docker's resolver; anything on the overlay
+        # is reached by ADDRESS and needs no DNS. That is one decision, not the
+        # three properties (upstream{}, no resolver, no set $var) it produces.
+        local tmpl
+        case "$placement" in
+            local|instance)
+                tmpl=./render/nginx-upstream.tmpl
+                uhost="$(overlay_get estate "peers.${peer}.address")"
+                uport="$port"   # resolved via service_upstream's same rule
+                if [ -z "$uhost" ]; then
+                    echo "  ! $name: placement '$placement' names peer '$peer', which has no address" >&2
+                    continue
+                fi
+                ;;
+            hosted)
+                echo "  ! $name: placement 'hosted' is declared but not rendered yet" >&2
+                continue
+                ;;
+            *)
+                if [ "$kind" = "static" ]; then
+                    tmpl=./render/nginx-static.tmpl
+                    uhost=""; uport=""
+                else
+                    tmpl=./render/nginx-proxy.tmpl
+                    uhost="${up%%:*}"; uport="${up##*:}"
+                fi
+                ;;
+        esac
+        sed -e "s|\${NAME}|${name}|g" \
+            -e "s|\${ESTATE}|${ESTATE}|g" \
+            -e "s|\${FQDN}|${host}.${DOMAIN}|g" \
+            -e "s|\${DOMAIN}|${DOMAIN}|g" \
+            -e "s|\${PLACEMENT}|${placement}|g" \
+            -e "s|\${PEER}|${peer}|g" \
+            -e "s|\${UPSTREAM_HOST}|${uhost}|g" \
+            -e "s|\${UPSTREAM_PORT}|${uport}|g" \
+            "$tmpl" > "$out/${name}.conf"
+        n=$((n + 1))
+    done < <(estate_services "$target")
+
+    echo "wrote $n vhost(s) to $out/  ($skipped hand-written, left alone)"
+    cat <<EONOTE
+
+TO INSTALL THESE, THREE THINGS MUST HAPPEN IN THIS ORDER — and the order is the
+whole reason this is not a one-liner:
+
+ 1. These land in conf.d/generated/, NOT conf.d/. ppl/ctrl/deploy.sh rsyncs the
+    gateway with --delete; generated and hand-written config sharing one
+    directory means one of them gets erased.
+
+ 2. nginx.conf needs a THIRD include line. Its conf.d/*.conf glob does not
+    recurse — which is exactly why conf.d/soleprint/*.conf already needs its
+    own line at nginx.conf:28-30.
+
+ 3. That new include changes LOAD ORDER, and load order decides which :443
+    block catches unmatched names. So default.conf's commented-out
+    ':443 default_server' must be restored FIRST. 'make check' fails on this
+    today, deliberately — it is a gate, not a warning.
+EONOTE
+}
+
+render_local() {
+    local out="$OUT_ROOT/local"; mkdir -p "$out"
+    local ld; ld=$(estate_get "local_domain")
+    : "${ld:=local.ar}"
+    local name host up kind raw port drift=0
+
+    {
+        cat <<EOH
+# GENERATED by berth from estate/${ESTATE}.json. Do not edit.
+# Regenerate: make services render local
+#
+# Install: sudo ln -sf \$PWD/Caddyfile /etc/caddy/Caddyfile && sudo systemctl reload caddy
+# All *.${ld} resolve to 127.0.0.1 via dnsmasq.
+#
+# Every site address carries an explicit :80. Without it Caddy 2 defaults to
+# :443 with auto-HTTPS, which on *.${ld} means cert provisioning attempts that
+# fail and break the listener. Plain HTTP only on this host.
+#
+# Caddy matches the MOST SPECIFIC site address, not the first — the opposite of
+# nginx, which matches exactly and otherwise falls to default_server. A name set
+# that is unambiguous here can be ambiguous on the box.
+EOH
+        while IFS=$'\x1f' read -r name host up kind raw placement peer port lhost; do
+            [ -z "$name" ] && continue
+            port=$(python3 -c '
+import json,sys
+d=json.load(open(sys.argv[1]))
+for s in d.get("services",[]):
+    if s.get("name")==sys.argv[2]:
+        print(s.get("local_port") or ""); break
+' "$ESTATE_FILE" "$name")
+            [ -z "$port" ] && continue
+            echo
+            echo "${lhost}.${ld}:80, *.${lhost}.${ld}:80 {"
+            echo "	reverse_proxy localhost:${port}"
+            echo "}"
+        done < <(estate_services local)
+    } > "$out/Caddyfile"
+
+    echo "wrote $out/Caddyfile"
+    echo
+    echo "rig is not consulted and does not know this exists — its handover"
+    echo "scrub refuses the string '${ld}'. Where a port belongs to a rig,"
+    echo "'make ports verify' RECOMPUTES rig's formula to check it rather than"
+    echo "importing rig's code. Convention, verified; not a dependency."
+}
+
+case "${1:-list}" in
+    list)   list ;;
+    render)
+        shift
+        # `case "${1:-X}"` defaults the match but leaves $1 empty.
+        t="${1:-$TARGET}"
+        case "$t" in
+            local)     render_local ;;
+            aws|gcp)   render_cloud "$t" ;;
+            *) echo "usage: $0 render [aws|gcp|local]" >&2; exit 1 ;;
+        esac
+        ;;
+    deploy)
+        echo "berth does not ship config; ppl/ctrl/deploy.sh does." >&2
+        echo "  berth's half is the DESCRIPTION and the render. Shipping is" >&2
+        echo "  rsync + compose against a live box, and it belongs where the" >&2
+        echo "  credentials are: berth is the tool, ppl is the estate that" >&2
+        echo "  holds the secrets." >&2
+        exit 1
+        ;;
+    *) echo "usage: $0 [list|render [aws|gcp|local]|deploy]" >&2; exit 1 ;;
+esac

berth/ctrl/versions.env

@@ -0,0 +1,11 @@
+# Pinned toolchain. Committed. The weakest config layer.
+
+# The infra executor. One, not a pair — see README.md.
+# This pin is a placeholder; set it from `tofu version` once installed.
+TOFU_VERSION=1.9.0
+TOFU_BIN=tofu
+
+# certbot runs as a throwaway container so the DNS plugin's credentials never
+# have to be installed on this machine.
+CERTBOT_AWS_IMAGE=certbot/dns-route53:latest
+CERTBOT_GCP_IMAGE=certbot/dns-google:latest

berth/ctrl/vpn.sh

@@ -0,0 +1,478 @@
+#!/usr/bin/env bash
+# Overlays — WireGuard as berth's network layer.
+#
+# Usage:
+#   ./vpn.sh                        # list
+#   ./vpn.sh show <overlay>         # topology
+#   ./vpn.sh check                  # invariants
+#   ./vpn.sh render <peer>          # peer's wg0.conf -> render/out/vpn/
+#   ./vpn.sh keygen <peer>          # keypair -> .secrets/; prints only the public key
+#   ./vpn.sh up|down --yes          # refuses; the host operates its own tunnel
+#
+#   sudo wg show | ./vpn.sh capture [--write]
+#
+# `wg show` is the only safe form: `wg show <if> dump` puts the private key in
+# field 1, and `wg showconf` prints it outright. capture refuses both.
+#
+# Rationale, topology and key handling: ../README.md
+
+set -euo pipefail
+cd "$(dirname "$0")"
+
+source ./lib/config.sh
+source ./lib/estate.sh
+load_config
+
+SECRETS_DIR="./.secrets/vpn"
+OUT_DIR="./render/out/vpn"
+
+WORST=0
+note() { echo "  $*"; }
+warn() { echo "  WARN  $*"; [ "$WORST" -lt 1 ] && WORST=1; return 0; }
+bad()  { echo "  FAIL  $*"; WORST=2; return 0; }
+
+# Which addresses belong to THIS machine, so checks can distinguish what they
+# can actually see from what needs capturing elsewhere.
+my_overlay_addrs() { ip -4 -o addr show 2>/dev/null | awk '{split($4,a,"/"); print a[1]}'; }
+is_me() { my_overlay_addrs | grep -qxF "$1"; }
+
+list() {
+    local n sub port peers
+    for n in $(overlay_names); do
+        sub=$(overlay_get "$n" subnet)
+        port=$(overlay_get "$n" listen_port)
+        peers=$(overlay_peers "$n" | grep -c . || true)
+        printf '%-10s %-16s port %-7s %s peer(s)\n' "$n" "$sub" "$port" "$peers"
+        note "$(overlay_get "$n" purpose)"
+    done
+    local st; st=$(estate_get "vpn._status")
+    [ -n "$st" ] && { echo; echo "  !! $st"; }
+}
+
+show() {
+    local ov="${1:-}"
+    [ -z "$ov" ] && { echo "usage: $0 show <overlay>" >&2; exit 1; }
+    overlay_names | grep -qxF "$ov" || {
+        echo "no such overlay: $ov" >&2
+        echo "available: $(overlay_names | tr '\n' ' ')" >&2; exit 1; }
+
+    echo "overlay: $ov   subnet $(overlay_get "$ov" subnet)   udp/$(overlay_get "$ov" listen_port)"
+    echo
+    printf '%-8s %-12s %-9s %-22s %s\n' PEER ADDRESS ROLE ENDPOINT PUBKEY
+    local name addr role ep pk aips ka
+    while IFS=$'\x1f' read -r name addr role ep pk aips ka; do
+        [ -z "$name" ] && continue
+        printf '%-8s %-12s %-9s %-22s %s%s\n' \
+            "$name" "$addr" "$role" "${ep:-—}" "${pk:-—}" \
+            "$(is_me "$addr" && echo '   <- this machine')"
+    done < <(overlay_peers "$ov")
+}
+
+check() {
+    local ov name addr role ep pk aips ka
+    for ov in $(overlay_names); do
+        local sub port
+        sub=$(overlay_get "$ov" subnet); port=$(overlay_get "$ov" listen_port)
+        echo "overlay '$ov' — $sub udp/$port"
+
+        # 1. addresses: unique, and inside the subnet. Two peers sharing an
+        #    address is a silent misroute, never an error message.
+        local addrs; addrs=$(overlay_peers "$ov" | cut -d$'\x1f' -f2 | grep -v '^$' || true)
+        local dupes; dupes=$(echo "$addrs" | sort | uniq -d)
+        [ -n "$dupes" ] && bad "duplicate peer addresses: $(echo "$dupes" | tr '\n' ' ')"
+        while IFS= read -r a; do
+            [ -z "$a" ] && continue
+            addr_in_subnet "$a" "$sub" || bad "$a is outside $sub"
+        done <<< "$addrs"
+
+        while IFS=$'\x1f' read -r name addr role ep pk aips ka; do
+            [ -z "$name" ] && continue
+
+            # AllowedIPs is cryptokey routing — route table and ACL at once.
+            case "$aips" in
+                *0.0.0.0/0*) warn "$name: AllowedIPs includes 0.0.0.0/0 — full-tunnel. Deliberate?" ;;
+            esac
+
+            # A peer with no endpoint cannot be dialed; it must initiate.
+            if [ -z "$ep" ] && [ "$role" != "roaming" ]; then
+                warn "$name: role '$role' but no endpoint — nothing can dial it."
+            fi
+            # Keepalive is NOT on the roaming peer's own entry — it is set on
+            # the entry for the peer it dials. Checked per-overlay below.
+            [ -z "$pk" ] && note "$name: public_key not captured yet"
+        done < <(overlay_peers "$ov")
+
+        # If anything roams, some peer entry must carry a keepalive.
+        if overlay_peers "$ov" | cut -d$'\x1f' -f3 | grep -qx roaming; then
+            if ! overlay_peers "$ov" | cut -d$'\x1f' -f7 | grep -qE '^[0-9]+$'; then
+                warn "a peer roams but no peer entry carries PersistentKeepalive"
+                note "      the roaming side sets it on the entry for the peer it dials;"
+                note "      without it the NAT mapping expires and the tunnel works only"
+                note "      while traffic flows outward — 'works sometimes'"
+            else
+                note "keepalive present on the dialed peer."
+            fi
+        fi
+
+        # 4. the listen port must be open wherever a peer is dialable.
+        local fwports; fwports=$(estate_get "firewall" | python3 -c '
+import json,sys
+try: print(" ".join(str(r.get("port")) for r in json.load(sys.stdin)))
+except Exception: pass')
+        case " $fwports " in
+            *" $port "*) note "udp/$port present in the firewall description." ;;
+            *) bad "udp/$port is in no firewall rule — no peer could be dialed." ;;
+        esac
+        echo
+    done
+
+    # Public keys are also 44-char base64, so shape alone proves nothing. The
+    # assertions are: no field named private, no key outside a public_key field.
+    echo "secrets — the description must never carry a private key"
+    local leaked
+    leaked=$(python3 - estate/../../estate/*.json <<'PY' 2>/dev/null || true
+import json, re, sys, glob
+KEY = re.compile(r'^[A-Za-z0-9+/]{43}=$')
+bad = []
+for f in glob.glob("../estate/*.json"):
+    def walk(node, path):
+        if isinstance(node, dict):
+            for k, v in node.items():
+                if re.search(r'priv', k, re.I):
+                    bad.append(f"{f}: field '{'.'.join(path+[k])}' is named private")
+                walk(v, path + [k])
+        elif isinstance(node, list):
+            for i, v in enumerate(node): walk(v, path + [str(i)])
+        elif isinstance(node, str) and KEY.match(node):
+            if not path or 'public' not in path[-1]:
+                bad.append(f"{f}: base64 key at '{'.'.join(path)}' is not a public_key field")
+    walk(json.load(open(f)), [])
+print("\n".join(bad))
+PY
+)
+    if [ -n "$leaked" ]; then
+        echo "$leaked" | while IFS= read -r l; do [ -n "$l" ] && bad "$l"; done
+    else
+        note "clean — no private-named field, no stray key material."
+    fi
+    echo
+
+    # A service reached over the overlay must bind an address the tunnel can
+    # reach. Loopback cannot be reached through a tunnel.
+    echo "bindings — services reached over the overlay must bind a reachable address"
+    local checked=0
+    while IFS=$'\x1f' read -r name host up kind raw placement peer port lhost; do
+        # Resolve placement first: a placed service's upstream is derived, not
+        # literal, so reading `up` alone skips it and this invariant goes quiet.
+        up="$(service_upstream "$up" "$placement" "$peer" "$port")" || true
+        [ -z "$up" ] && continue
+        local uhost="${up%%:*}" uport="${up##*:}"
+        addr_in_subnet "$uhost" "$(overlay_get estate subnet)" 2>/dev/null || continue
+        checked=$((checked + 1))
+        if is_me "$uhost"; then
+            local binds; binds=$(ss -ltn 2>/dev/null | awk -v p=":$uport\$" '$4 ~ p {print $4}')
+            if [ -z "$binds" ]; then
+                bad "$name: nothing listens on :$uport here, but $uhost:$uport is its upstream"
+            elif echo "$binds" | grep -q '^127\.0\.0\.1:'; then
+                bad "$name: :$uport binds 127.0.0.1 — unreachable over the overlay"
+                note "      the tunnel cannot reach loopback; bind 0.0.0.0 or $uhost"
+            else
+                note "$name: :$uport binds $(echo "$binds" | tr '\n' ' ')— reachable"
+                echo "$binds" | grep -q '^0\.0\.0\.0:' && \
+                    note "      (0.0.0.0 also exposes it to the LAN; $uhost alone would be tighter)"
+            fi
+        else
+            note "$name: upstream $uhost is another peer — needs capture there"
+        fi
+    done < <(estate_services "$TARGET")
+    [ "$checked" = 0 ] && note "no service currently has an overlay address as its upstream."
+    echo
+
+    case "$WORST" in
+        0) echo "OK" ;;
+        1) echo "OK, with warnings" ;;
+        2) echo "PROBLEMS FOUND — see FAIL lines above" ;;
+    esac
+    return 0
+}
+
+# A .gitignore pattern containing a slash anchors to its own directory, so the
+# only way to know a path is ignored is to ask git.
+assert_ignored() {
+    local path="$1"
+    if ! git check-ignore -q "$path" 2>/dev/null; then
+        echo "REFUSING: '$path' is not gitignored." >&2
+        echo "  Writing key material there would stage it on the next 'git add'." >&2
+        echo "  Verify with: git check-ignore -v $path" >&2
+        exit 1
+    fi
+}
+
+keygen() {
+    local peer="${1:-}"
+    [ -z "$peer" ] && { echo "usage: $0 keygen <peer>" >&2; exit 1; }
+    command -v wg >/dev/null || { echo "wg not installed." >&2; exit 1; }
+
+    mkdir -p "$SECRETS_DIR"
+    assert_ignored "$SECRETS_DIR"
+    local kf="$SECRETS_DIR/${peer}.key"
+    [ -e "$kf" ] && { echo "REFUSING: $kf exists. Delete it deliberately to rotate." >&2; exit 1; }
+
+    ( umask 077; wg genkey > "$kf" )
+    echo "private key -> $kf  (0600, gitignored, never leaves this machine)"
+    echo
+    echo "public key for the estate description:"
+    echo "  $(wg pubkey < "$kf")"
+    echo
+    echo "Paste that into estate/*.json under vpn.overlays.<ov>.peers.${peer}.public_key."
+    echo "The private key stays here and is injected only at render time."
+}
+
+render() {
+    local peer="${1:-}"
+    [ -z "$peer" ] && { echo "usage: $0 render <peer>" >&2; exit 1; }
+    mkdir -p "$OUT_DIR"
+    assert_ignored "$OUT_DIR"
+
+    local ov=estate
+    local found=""
+    local name addr role ep pk aips ka
+    while IFS=$'\x1f' read -r name addr role ep pk aips ka; do
+        [ "$name" = "$peer" ] && found=1 && break
+    done < <(overlay_peers "$ov")
+    [ -z "$found" ] && { echo "no such peer '$peer' in overlay '$ov'" >&2; exit 1; }
+
+    local missing=""
+    while IFS=$'\x1f' read -r name addr role ep pk aips ka; do
+        [ -z "$pk" ] && missing="$missing $name"
+    done < <(overlay_peers "$ov")
+    if [ -n "$missing" ]; then
+        echo "REFUSING to render: public keys not captured for:$missing" >&2
+        echo "  A config without every peer's public key is a config that silently" >&2
+        echo "  drops those peers. Capture first: sudo wg show | $0 capture --write" >&2
+        exit 1
+    fi
+    echo "would write $OUT_DIR/${peer}.conf (all keys present)"
+}
+
+# Reads `wg show` on stdin. Peers match by allowed-ips address, not public key,
+# because the keys are what is missing. A roaming peer's endpoint is a home
+# address and has no stable value — dropped in the parser, not just unused.
+capture() {
+    local write="" as_peer=""
+    while [ $# -gt 0 ]; do
+        case "$1" in
+            --write) write=1 ;;
+            --as)    shift; as_peer="${1:-}"
+                     [ -z "$as_peer" ] && { echo "--as needs a peer name" >&2; exit 1; } ;;
+            *)       echo "capture: unknown argument '$1'" >&2; exit 1 ;;
+        esac
+        shift
+    done
+
+    local input; input=$(cat)
+    if [ -z "$input" ]; then
+        echo "nothing on stdin." >&2
+        echo "  run: sudo wg show | $0 capture" >&2
+        exit 1
+    fi
+    # Refuse the unsafe forms outright rather than parsing around them.
+    if printf '%s' "$input" | grep -qiE '^\s*PrivateKey\s*=|^\[Interface\]'; then
+        echo "REFUSING: this looks like 'wg showconf' output — it contains a PRIVATE KEY." >&2
+        echo "  Use 'sudo wg show' (plain). It prints 'private key: (hidden)'." >&2
+        exit 1
+    fi
+    if ! printf '%s' "$input" | grep -q 'interface:'; then
+        echo "REFUSING: this does not look like 'wg show' output." >&2
+        echo "  If it was 'wg show <if> dump': that form's first field IS the" >&2
+        echo "  private key. Use 'sudo wg show' with no subcommand." >&2
+        exit 1
+    fi
+
+    WRITE="$write" AS_PEER="$as_peer" INPUT="$input" python3 - "$ESTATE_FILE" <<'PYCAP'
+import collections, ipaddress, json, os, re, sys
+
+text  = os.environ["INPUT"]
+write = os.environ.get("WRITE") == "1"
+path  = sys.argv[1]
+
+iface, peers, cur = {}, [], None
+for line in text.splitlines():
+    st = line.strip()
+    if st.startswith("interface:"):
+        cur = iface; cur["name"] = st.split(":", 1)[1].strip(); continue
+    if st.startswith("peer:"):
+        cur = {"public_key": st.split(":", 1)[1].strip()}; peers.append(cur); continue
+    if cur is None or ":" not in st:
+        continue
+    k, v = st.split(":", 1)
+    k, v = k.strip().lower(), v.strip()
+    if k == "private key":
+        continue                     # never recorded, whatever it says
+    if k == "public key":   cur["public_key"] = v
+    elif k == "listening port": cur["listen_port"] = v
+    elif k == "allowed ips":    cur["allowed_ips"] = v
+    elif k == "endpoint":       cur["endpoint"] = v
+    elif k == "persistent keepalive":
+        m = re.search(r"(\d+)", v)
+        if m: cur["keepalive"] = int(m.group(1))
+
+d  = json.load(open(path), object_pairs_hook=collections.OrderedDict)
+ov = d["vpn"]["overlays"]["estate"]
+
+# address -> peer name, from what the estate already declares
+by_addr = {p["address"]: n for n, p in ov["peers"].items() if p.get("address")}
+by_key  = {p["public_key"]: n for n, p in ov["peers"].items() if p.get("public_key")}
+subnet  = ipaddress.ip_network(ov["subnet"]) if ov.get("subnet") else None
+hubs    = [n for n, p in ov["peers"].items() if p.get("role") == "hub"]
+
+# Whose interface block is this? `--as` names it explicitly, and that is the only
+# thing that works for output captured over ssh: the addresses on THIS machine
+# say nothing about the machine the output came from.
+as_peer = os.environ.get("AS_PEER") or ""
+if as_peer:
+    if as_peer not in ov["peers"]:
+        print("no peer named %r in this overlay. known: %s"
+              % (as_peer, ", ".join(ov["peers"])))
+        raise SystemExit(1)
+    me = as_peer
+else:
+    me = None
+    local = os.popen(
+        "ip -4 -o addr show 2>/dev/null | awk '{split($4,a,\"/\"); print a[1]}'"
+    ).read().split()
+    for n, p in ov["peers"].items():
+        if p.get("address") and p["address"] in local:
+            me = n
+
+changes = []
+conflicts = []
+staged = {}
+def setf(peer, field, val, why=""):
+    p = ov["peers"][peer]
+    if val is None or p.get(field) == val:
+        return
+    # Two values for one field means the input is from another machine.
+    prev = staged.get((peer, field))
+    if prev is not None and prev != val:
+        conflicts.append((peer, field, prev, val))
+        return
+    staged[(peer, field)] = val
+    changes.append((peer, field, p.get(field), val, why))
+    if write:
+        p[field] = val
+
+if me and iface.get("public_key"):
+    setf(me, "public_key", iface["public_key"], "(this machine's interface)")
+
+def match(pr):
+    # 1. The public key IS the identity. Use it whenever the estate knows it.
+    n = by_key.get(pr["public_key"])
+    if n:
+        return n
+    nets = [a.strip() for a in pr.get("allowed_ips", "").split(",") if a.strip()]
+    # 2. An allowed-ip that is a declared peer address — the ordinary spoke case.
+    for a in nets:
+        if a.split("/")[0] in by_addr:
+            return by_addr[a.split("/")[0]]
+    # 3. A peer routing the WHOLE overlay is the hub seen from a spoke. Its
+    #    allowed_ips is the subnet itself, so no single address ever matches it.
+    if subnet and len(hubs) == 1:
+        for a in nets:
+            try:
+                if ipaddress.ip_network(a, strict=False).supernet_of(subnet):
+                    return hubs[0]
+            except ValueError:
+                continue
+    return None
+
+for pr in peers:
+    name = match(pr)
+    if not name:
+        changes.append(("?", "UNMATCHED", None,
+                        "allowed_ips=%s key=%s" % (pr.get("allowed_ips"), pr["public_key"][:12] + "..."),
+                        "no estate peer has this key, this address, or this route"))
+        continue
+    setf(name, "public_key", pr.get("public_key"))
+    setf(name, "allowed_ips", pr.get("allowed_ips"))
+    setf(name, "keepalive", pr.get("keepalive"))
+    # endpoint: recorded ONLY for a non-roaming peer. For a roaming one the
+    # value is a home ISP address and is deliberately dropped here.
+    if pr.get("endpoint"):
+        if ov["peers"][name].get("role") == "roaming":
+            changes.append((name, "endpoint", None, "(dropped: roaming peer)",
+                            "a home address is the one sensitive field; roaming peers have no stable endpoint"))
+        else:
+            setf(name, "endpoint", pr["endpoint"])
+
+if me and iface.get("listen_port"):
+    try:
+        lp = int(iface["listen_port"])
+    except ValueError:
+        lp = None
+    if lp is not None:
+        # A listen port belongs to the PEER, not to the overlay. A roaming peer's
+        # is an ephemeral source port chosen by the kernel; writing it to the
+        # overlay would rename the port the firewall rule is checked against.
+        setf(me, "listen_port", lp)
+        if ov["peers"][me].get("role") == "hub" and ov.get("listen_port") != lp:
+            changes.append(("(overlay)", "listen_port", ov.get("listen_port"), lp,
+                            "the hub's port is the overlay's port"))
+            if write:
+                ov["listen_port"] = lp
+
+if conflicts:
+    print("REFUSING: the same field was reported twice with different values.\n")
+    for peer, field, a, b in conflicts:
+        print("  %s.%s: %s  vs  %s" % (peer, field, a, b))
+    print("\nThis usually means `wg show` output from one machine was piped into")
+    print("capture on another. Run capture on the machine the output came from.")
+    raise SystemExit(1)
+
+if not changes:
+    print("nothing to record — the estate already matches what wg reports.")
+else:
+    print("%-9s %-12s %-22s %s" % ("PEER", "FIELD", "WAS", "WOULD BE"))
+    for peer, field, was, val, why in changes:
+        print("%-9s %-12s %-22s %s" % (peer, field, was if was is not None else "—", val))
+        if why: print("          %s" % why)
+
+if write:
+    still = [n for n, p in ov["peers"].items() if not p.get("public_key")]
+    if not still:
+        d["vpn"]["_status"] = ("CAPTURED %s — public keys, allowed-ips and keepalive read from "
+                               "`wg show`. Roaming endpoints deliberately not recorded."
+                               % __import__("datetime").date.today())
+    json.dump(d, open(path, "w"), indent=2, ensure_ascii=False)
+    open(path, "a").write("\n")
+    print("\nwritten to %s" % path)
+else:
+    print("\nnothing written. Add --write to record it.")
+PYCAP
+}
+
+refuse() {
+    local verb="$1"; shift
+    local yes=""
+    for a in "$@"; do [ "$a" = "--yes" ] && yes=1; done
+    [ -z "$yes" ] && {
+        echo "refusing to $verb without --yes." >&2
+        echo "  $verb changes live networking — it can cut the path this session" >&2
+        echo "  is reaching the estate through. Read 'make vpn check' first." >&2
+        exit 1; }
+    echo "refusing to $verb: not implemented. Bringing a tunnel up or down is" >&2
+    echo "  the host's business, and the live one is systemd-managed" >&2
+    echo "  (wg-quick@wg0). berth describes and renders; it does not operate." >&2
+    exit 1
+}
+
+case "${1:-list}" in
+    list)     list ;;
+    show)     shift; show "${1:-}" ;;
+    check)    check ;;
+    render)   shift; render "${1:-}" ;;
+    keygen)   shift; keygen "${1:-}" ;;
+    capture)  shift; capture "$@" ;;
+    up|down)  v="$1"; shift; refuse "$v" "$@" ;;
+    *) echo "usage: $0 [list|show <ov>|check|render <peer>|keygen <peer>|capture [--as <peer>] [--write]|up --yes|down --yes]" >&2; exit 1 ;;
+esac

berth/estate/mcrn.json

@@ -0,0 +1,326 @@
+{
+  "_meta": {
+    "status": "UNVERIFIED — derived from the repos, not from the estate",
+    "why": "ppl/infra/ was written and never applied: no ~/.pulumi, no infra/venv, no stack state, files dated 'mar 6'. The estate was built in the console and the IaC is aspirational. B1's inventory is what replaces these values with observed ones; until it runs, every field here is a CLAIM.",
+    "never_record": "credential values. Resource ids and settings only. nova's gateway secret is deliberately absent from this file even though it is committed in plaintext in ppl/gateway/nginx/conf.d/nova.conf — see services[].raw.",
+    "sources": [
+      "ppl/infra/__main__.py",
+      "ppl/ctrl/dns.sh",
+      "ppl/ctrl/certs.sh",
+      "ppl/gateway/docker-compose.yml",
+      "ppl/gateway/nginx/conf.d/",
+      "ppl/local/Caddyfile"
+    ],
+    "placement": {
+      "box": "a container on the estate's own docker network — upstream is the container name",
+      "local": "a rig cluster on a peer, reached over the overlay — upstream is that peer's address",
+      "instance": "a dedicated cloud instance on the overlay — same rendering as `local`",
+      "hosted": "a managed endpoint. Declared so moving to one is a one-line change; unused.",
+      "_why": "A service says WHERE it runs. How it is reached follows from that, and the three properties of a static-upstream vhost — upstream{}, no resolver, no set $var — are one decision rather than three."
+    }
+  },
+  "domain": "mcrn.ar",
+  "local_domain": "local.ar",
+  "host": "mcrn",
+  "host_admin": "mcrn-admin",
+  "instance": {
+    "type": "t3.small",
+    "disk_gb": 30,
+    "disk_type": "gp3",
+    "image": "debian-12",
+    "user": "mariano"
+  },
+  "firewall": [
+    {
+      "port": 22,
+      "proto": "tcp",
+      "desc": "SSH"
+    },
+    {
+      "port": 80,
+      "proto": "tcp",
+      "desc": "HTTP"
+    },
+    {
+      "port": 443,
+      "proto": "tcp",
+      "desc": "HTTPS"
+    },
+    {
+      "port": 3022,
+      "proto": "tcp",
+      "desc": "Gitea SSH",
+      "note": "compose maps 3022:22 but GITEA__server__SSH_PORT=22, so gitea advertises :22 in clone URLs while listening on :3022. B1 confirms which is real."
+    },
+    {
+      "port": 51820,
+      "proto": "udp",
+      "desc": "WireGuard",
+      "note": "ABSENT from ppl/infra/__main__.py's four rules — but the tunnel is live (ping 10.8.0.1 succeeds), so the real security group must already allow it. The code therefore does not describe the estate. Confirm in V1."
+    }
+  ],
+  "network": {
+    "docker_network": "gateway",
+    "docker_network_note": "A fixed, externally-joinable bridge name. Every unrelated app stack on the box joins it so nginx can resolve them by container name. This is why the gateway compose declares 8 services while nginx routes 20+ hostnames.",
+    "wireguard_moved": "superseded by the top-level `vpn` block"
+  },
+  "vpn": {
+    "_status": "CAPTURED 2026-09-14 — public keys, allowed-ips and keepalive read from `wg show`. Roaming endpoints deliberately not recorded.",
+    "_never_record": "private keys. `wg show` prints 'private key: (hidden)' and is the safe capture command. `wg showconf` dumps PrivateKey= in clear — never use it.",
+    "overlays": {
+      "estate": {
+        "purpose": "Connects the estate's machines across clouds without a shared VPC, and carries everything that does not need to be publicly reachable.",
+        "subnet": "10.8.0.0/24",
+        "listen_port": 51820,
+        "peers": {
+          "box": {
+            "address": "10.8.0.1",
+            "role": "hub",
+            "note": "mcrn.ar. Has a public IP, so it is the peer others dial. Carries the registry (:5000) and woodpecker's gRPC (:9000), both bound to this address and therefore overlay-only.",
+            "endpoint": "3.23.204.197:51820",
+            "public_key": "zVYCmi3xucuX7k/aDhrOUPyN4GRk96ffSDD6dUFQjh4=",
+            "allowed_ips": "10.8.0.0/24",
+            "keepalive": 25,
+            "listen_port": 51820
+          },
+          "nrft": {
+            "address": "10.8.0.2",
+            "role": "roaming",
+            "note": "The dev box. Behind NAT, so it must initiate and needs PersistentKeepalive. Verified: wg0 UP at 10.8.0.2/24, ping 10.8.0.1 0% loss at 153ms.",
+            "endpoint": null,
+            "public_key": "zlIBGs4y5rt6uVdmFBasHpafht6ErxG+R3ySCg5rh3s=",
+            "allowed_ips": "10.8.0.2/32, 192.168.1.0/24",
+            "keepalive": null,
+            "listen_port": 36145
+          },
+          "work": {
+            "address": "10.8.0.3",
+            "role": "roaming",
+            "note": "A work computer, granted access when it was needed. Identified by the user at capture time, 2026-09-14 — it was NOT in the description before, and the wire is where it was found. No handshake and no transfer have ever been recorded for it, so it is a standing grant rather than a live peer: it can connect, and never has. Whether to keep or revoke it is the host's call.",
+            "endpoint": null,
+            "public_key": "ruSZwKt/p60GVsTLSAhcKBIXKkSZsf0gWSmSH1+UgE0=",
+            "allowed_ips": "10.8.0.3/32",
+            "keepalive": null
+          }
+        }
+      }
+    }
+  },
+  "databases": [
+    "gitea",
+    "woodpecker",
+    "umami"
+  ],
+  "certs": {
+    "issued": [
+      "mcrn.ar",
+      "*.mcrn.ar",
+      "*.spr.mcrn.ar"
+    ],
+    "issued_source": "ppl/ctrl/certs.sh:92 — the -d flags passed to certbot",
+    "note": "What the cert ACTUALLY covers. estate_sans() derives what the services NEED. check.sh compares the two; the difference is the finding, not a restatement."
+  },
+  "services": [
+    {
+      "name": "gitea",
+      "host": "git",
+      "upstream": "gitea:3000",
+      "targets": [
+        "aws"
+      ]
+    },
+    {
+      "name": "woodpecker",
+      "host": "ci",
+      "upstream": "woodpecker-server:8000",
+      "targets": [
+        "aws"
+      ]
+    },
+    {
+      "name": "registry",
+      "host": "registry",
+      "upstream": "registry:5000",
+      "targets": [
+        "aws"
+      ]
+    },
+    {
+      "name": "umami",
+      "host": "analytics",
+      "upstream": "umami:3000",
+      "targets": [
+        "aws"
+      ]
+    },
+    {
+      "name": "docserve",
+      "host": "docs",
+      "upstream": "docserve:8020",
+      "targets": [
+        "aws"
+      ]
+    },
+    {
+      "name": "ghost",
+      "host": "notes",
+      "upstream": "ghost:2368",
+      "targets": [
+        "aws"
+      ]
+    },
+    {
+      "name": "deskmeter",
+      "host": "deskmeter",
+      "upstream": "dmweb:10000",
+      "targets": [
+        "aws"
+      ],
+      "local_port": 10000
+    },
+    {
+      "name": "sysmonstm",
+      "host": "sysmonstm",
+      "upstream": "sysmonstm-edge:8080",
+      "targets": [
+        "aws"
+      ],
+      "local_port": 8020
+    },
+    {
+      "name": "malvalava",
+      "host": "malvalava",
+      "upstream": "mlvclean-frontend:80",
+      "targets": [
+        "aws"
+      ],
+      "local_port": 30090
+    },
+    {
+      "name": "soleprint",
+      "host": "soleprint",
+      "upstream": "soleprint:8000",
+      "targets": [
+        "aws"
+      ],
+      "local_port": 12000
+    },
+    {
+      "name": "dlt",
+      "host": "dlt.spr",
+      "upstream": "dlt_spr:8000",
+      "targets": [
+        "aws"
+      ]
+    },
+    {
+      "name": "sample",
+      "host": "sample.spr",
+      "upstream": "sample_spr:8000",
+      "targets": [
+        "aws"
+      ]
+    },
+    {
+      "name": "mariano",
+      "host": "mariano",
+      "kind": "static",
+      "targets": [
+        "aws"
+      ]
+    },
+    {
+      "name": "rigui",
+      "host": "rig",
+      "kind": "static",
+      "targets": [
+        "aws"
+      ],
+      "local_port": 20310
+    },
+    {
+      "name": "unt",
+      "host": "unt",
+      "targets": [
+        "local"
+      ],
+      "local_port": 8040
+    },
+    {
+      "name": "mpr",
+      "host": "mpr",
+      "targets": [
+        "local"
+      ],
+      "local_port": 30080
+    },
+    {
+      "name": "nvi",
+      "host": "nvi",
+      "targets": [
+        "local"
+      ],
+      "local_port": 8060
+    },
+    {
+      "name": "eth",
+      "host": "eth",
+      "targets": [
+        "local"
+      ],
+      "local_port": 8050
+    },
+    {
+      "name": "amar",
+      "host": "amar",
+      "targets": [
+        "local"
+      ],
+      "local_port": 8030
+    },
+    {
+      "name": "nova",
+      "host": "nova",
+      "upstream": "nova-ui:80",
+      "targets": [
+        "aws"
+      ],
+      "raw": true,
+      "raw_why": "Gated on an X-Gateway-Secret header whose value is committed in plaintext. The value is NOT recorded here. Worse: stellarair.conf proxies to the SAME nova-ui:80 upstream WITHOUT the check, so the gate is bypassable by hostname. Stays hand-written until that is decided."
+    },
+    {
+      "name": "stellarair",
+      "host": "stellarair",
+      "upstream": "nova-ui:80",
+      "targets": [
+        "aws"
+      ],
+      "raw": true,
+      "raw_why": "See nova. Same upstream, no header gate."
+    },
+    {
+      "name": "langfuse",
+      "host": "langfuse",
+      "local_host": "lng",
+      "placement": "local",
+      "peer": "nrft",
+      "port": 3000,
+      "targets": [
+        "aws",
+        "local"
+      ],
+      "local_port": 3000,
+      "note": "One service, one socket, two names. It was two entries with one flagged `raw`; placement is what made the exception expressible, so it is generated now."
+    },
+    {
+      "name": "legacy",
+      "host": "*.soleprint",
+      "upstream": "soleprint:8000",
+      "targets": [
+        "aws"
+      ],
+      "raw": true,
+      "raw_why": "A regex server_name with a named capture plus sub_filter injection — not expressible as a template. ALSO BROKEN: its /api/, /admin/, /static/ and / blocks proxy to 127.0.0.1, i.e. inside the nginx container where nothing listens, so every legacy room 502s. Only /wrapper/ uses the correct container-name form."
+    }
+  ]
+}

ctrl/cluster.sh

@@ -6,16 +6,49 @@
 #   ./ctrl/cluster.sh down     # delete it (drops every room's namespace)
 #   ./ctrl/cluster.sh status   # what's running on it
 #
-# One target, one script — the variants live here. The kind-*.sh files stay
-# exactly as they are and remain runnable on their own; this only dispatches.
+# spr depends on rig, never the other way round. Building and deleting a cluster
+# is rig's job, so up and down hand straight to rig/ctrl/cluster.sh, carrying the
+# four things that make this cluster spr's rather than rig's defaults:
+#
+#   CLUSTER=spr          rooms deploy into the kind-spr context
+#   KIND_CONFIG          spr's own shape, which maps the rooms' gateway NodePorts
+#   REGISTRY_MODE=none   rooms load images straight into the node
+#   PROFILE=minimal      pinned here, so a change to rig's own ctrl/.env can never
+#                        quietly add addons to spr's cluster
+#
+# status stays here: it answers a question about rooms, not about the cluster.
 set -e
 
 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+RIG_CTRL="$SCRIPT_DIR/../rig/ctrl"
+
+rig() {
+    CLUSTER=spr \
+    KIND_CONFIG="$SCRIPT_DIR/k8s/kind-config.yaml" \
+    REGISTRY_MODE=none \
+    PROFILE=minimal \
+        bash "$RIG_CTRL/cluster.sh" "$@"
+}
 
 case "${1:-status}" in
-    up)     exec "$SCRIPT_DIR/kind-up.sh" ;;
-    down)   exec "$SCRIPT_DIR/kind-down.sh" ;;
-    status) exec "$SCRIPT_DIR/kind-status.sh" ;;
+    up)
+        rig up
+        echo
+        echo "Per-room deploy:"
+        echo "  cd gen/<room> && ./ctrl/k8s-up.sh"
+        ;;
+    down)
+        rig down
+        ;;
+    status)
+        if ! kind get clusters 2>/dev/null | grep -qx spr; then
+            echo "No 'spr' kind cluster — run: make cluster up"
+            exit 0
+        fi
+        kubectl --context kind-spr get namespaces -l soleprint-room
+        echo
+        kubectl --context kind-spr get pods -A -l soleprint-room
+        ;;
     *)
         echo "Unknown subcommand: $1" >&2
         echo "Usage: cluster.sh [up|down|status]" >&2

ctrl/k8s/kind-config.yaml

@@ -3,9 +3,15 @@ apiVersion: kind.x-k8s.io/v1alpha4
 # Single shared cluster for all soleprint rooms.
 # Each room deploys into its own namespace; gateway Services pick a
 # NodePort from the 30080-30099 range mapped here.
-name: spr
+#
+# Built by rig, not by spr: ctrl/cluster.sh hands this file to
+# rig/ctrl/cluster.sh, which substitutes CLUSTER and NODE_IMAGE (named without
+# braces here so this comment survives the substitution). The shape is spr's —
+# what its cluster needs is spr's business. Building it is rig's.
+name: ${CLUSTER}
 nodes:
   - role: control-plane
+    image: ${NODE_IMAGE}
     extraPortMappings:
       # Room gateway NodePorts (one per active room).
       - {containerPort: 30080, hostPort: 30080, protocol: TCP}

ctrl/kind-down.sh

@@ -1,12 +0,0 @@
-#!/bin/bash
-# Delete the shared `spr` kind cluster (drops every room's namespace too).
-# Use `gen/<room>/ctrl/k8s-down.sh` instead if you only want to remove
-# a single room's namespace.
-set -e
-
-if kind get clusters 2>/dev/null | grep -q '^spr$'; then
-    echo "Deleting kind cluster 'spr'..."
-    kind delete cluster --name spr
-else
-    echo "No kind cluster 'spr' to delete."
-fi

ctrl/kind-status.sh

@@ -1,12 +0,0 @@
-#!/bin/bash
-# Show what's running on the shared `spr` cluster.
-set -e
-
-if ! kind get clusters 2>/dev/null | grep -q '^spr$'; then
-    echo "No 'spr' kind cluster — run ctrl/kind-up.sh"
-    exit 0
-fi
-
-kubectl --context kind-spr get namespaces -l soleprint-room
-echo
-kubectl --context kind-spr get pods -A -l soleprint-room

ctrl/kind-up.sh

@@ -1,21 +0,0 @@
-#!/bin/bash
-# Create (or no-op) the single shared `spr` kind cluster used by every
-# soleprint room. Per-room work happens inside namespaces — see
-# `gen/<room>/ctrl/k8s-up.sh`.
-set -e
-
-SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-KIND_CONFIG="$SCRIPT_DIR/k8s/kind-config.yaml"
-
-if kind get clusters 2>/dev/null | grep -q '^spr$'; then
-    echo "Kind cluster 'spr' already exists."
-else
-    echo "Creating kind cluster 'spr'..."
-    kind create cluster --config "$KIND_CONFIG"
-fi
-
-kubectl config use-context kind-spr >/dev/null
-
-echo
-echo "Cluster ready. Per-room deploy:"
-echo "  cd gen/<room> && ./ctrl/k8s-up.sh"

ctrl/theme.sh

@@ -0,0 +1,58 @@
+#!/usr/bin/env bash
+# Bake the theme and its parts into the pages that use them.
+#
+# Usage:
+#   ./ctrl/theme.sh             # bake — rewrite every generated block
+#   ./ctrl/theme.sh check       # fail if any page is stale; changes nothing
+#   ./ctrl/theme.sh new [title] # a scaffold page to start from
+#   ./ctrl/theme.sh parts       # what can be added, and the markup that adds it
+#   ./ctrl/theme.sh export [name...]   # the contract for a subset, as one doc
+#
+# `export` is for handing a vetted LLM what it needs to write an ad-hoc page —
+# the chosen parts, their markup, and the tokens resolved to literal values, so
+# the document stands alone. Naming parts is the point: hand over everything and
+# you get back a page built from Vue components that cannot run standalone.
+#
+#   ./ctrl/theme.sh export panel split > /tmp/contract.md
+#
+# Call the script directly when piping; `make` echoes its recipe to stdout.
+# For whole-repo context this is the wrong tool — station/tools/distill already
+# flattens a tree to one budgeted document.
+#
+# A page that says `background: var(--bg)` and never gets `--bg` is UNSTYLED,
+# not merely unbranded — the declaration is invalid at computed-value time. That
+# is why every page carries a baked default, and why `check` is worth running.
+#
+# This exists because bake.py was reachable by no command at all: not from the
+# Makefile, not from ctrl/, not from build.py. A drift check nobody runs is a
+# drift check that reports nothing, and the evidence was already on disk —
+# histgen's page linked /theme.css for months, was missing from the old
+# hardcoded page list, and so was never baked once.
+set -e
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+ROOT_DIR="$(dirname "$SCRIPT_DIR")"
+cd "$ROOT_DIR/soleprint"
+
+PYTHON="${PYTHON:-python3}"
+
+case "${1:-bake}" in
+    bake)  exec "$PYTHON" common/theme/bake.py ;;
+    check) exec "$PYTHON" common/theme/bake.py --check ;;
+    parts)
+        exec "$PYTHON" common/theme/bake.py --parts
+        ;;
+    new)
+        shift
+        exec "$PYTHON" common/theme/bake.py --new "$@"
+        ;;
+    export)
+        shift
+        exec "$PYTHON" common/theme/bake.py --export "$@"
+        ;;
+    *)
+        echo "Unknown: $1" >&2
+        echo "Usage: ./ctrl/theme.sh [new [title]|parts|bake|check|export [name...]]" >&2
+        exit 1
+        ;;
+esac

rig/.gitignore

@@ -9,12 +9,13 @@ ctrl/.env
 # generated: the .dot is a build artifact rendered from arch/*.json, never hand-edited.
 # The .svg IS committed — onboarding material should render in a repo browser.
 arch/*.dot
-ctrl/Tiltfile.gen
+# ctrl/Tiltfile.gen was here for a generator that no longer exists. ctrl/Tiltfile
+# is now a real, committed file that derives its values when Tilt parses it, so
+# there is nothing generated to ignore.
 
-# binaries pulled by `make deps-bundle` for the air-gapped wizard image
+# binaries pulled by `make deps-bundle` for the air-gapped installer image
 vendor
 
 # Client rigs are NOT ignored here. A copy is a SIBLING of this directory
-# (spr/acme-rig), so a rule in this file cannot see it — the rules live in
-# spr/.gitignore, anchored at spr's root, where `*-rig/` matches the siblings and
-# `!rig/sample-rig/` keeps the committed stand-in.
+# (../acme-rig), so a rule in this file cannot see it — the rules live in the
+# parent repo's .gitignore, anchored at its root, where `*-rig/` matches them.

rig/BOOTSTRAP.md

@@ -4,9 +4,8 @@ The README says the prerequisite is Docker and nothing else. This is what that
 actually looks like end to end: a bare Linux box, and a new project running under
 Tilt at the end of it.
 
-rig lives inside soleprint, at `spr/rig` — it is soleprint's cluster half, and
-a client copy is a sibling (`spr/acme-rig`). Paths below are relative to
-soleprint's checkout.
+A copy of this directory is a sibling of it, named after the environment it
+models (`acme-rig`). Paths below are relative to the parent checkout.
 
 It spans three repos because the work does. **rig** prepares the machine — the
 pinned toolchain, the cluster, the port arithmetic. **all** owns the shape a
@@ -50,7 +49,7 @@ isn't.
 ## Read the docs before installing anything
 
 ```bash
-cd spr/rig
+cd rig
 make docs
 ```
 
@@ -67,11 +66,11 @@ persists; ctrl-c ends it.
 ## Ask what is wrong with this machine
 
 ```bash
-make station
+make check
 cp ctrl/.env.example ctrl/.env
 ```
 
-`station.sh` reports and instructs, and fixes nothing. It runs bare rather than
+`check.sh` reports and instructs, and fixes nothing. It runs bare rather than
 in a container because host detection only ever reads `/proc` and `/etc` — no
 dependency beyond coreutils.
 
@@ -80,7 +79,7 @@ port rig binds derives from this directory's name, so the answer is specific to
 this copy, and a clash here surfaces as an opaque `failed to bind host port` in
 the middle of cluster creation if you skip it.
 
-Copy the `.env` even though station only warns about it. It is gitignored, it is
+Copy the `.env` even though the check only warns about it. It is gitignored, it is
 where a machine-local override goes, and `ports.sh persist` expects it to exist.
 
 
@@ -88,33 +87,33 @@ where a machine-local override goes, and `ports.sh persist` expects it to exist.
 
 This is the step where "nothing installed" stops being rhetorical.
 
-`make deps` runs `ctrl/wizard.sh install` directly on the host, and the wizard
+`make deps` runs `ctrl/deps.sh install` directly on the host, and the installer
 fetches with `curl`. A stock `debian:trixie-slim` has no curl — detection runs
 fine, then the first download dies with `curl: command not found` and an exit
-code of 127. That is the bootstrap paradox `ctrl/Dockerfile.wizard` exists
-to kill — the wizard carries its own toolchain so the host needs only Docker —
+code of 127. That is the bootstrap paradox `ctrl/Dockerfile.deps` exists
+to kill — the installer carries its own toolchain so the host needs only Docker —
 but building the image and running it are two different things, and only the
 build has a Makefile target today. **On a genuinely bare machine, run it by
 hand:**
 
 ```bash
-make wizard                          # builds rig-wizard:wizard
+make deps-image                          # builds rig-deps:deps
 mkdir -p ~/.local/bin
 docker run --rm \
   -v /:/host:ro \
   -v /var/run/docker.sock:/var/run/docker.sock \
   -v "$HOME/.local/bin:/out/bin" \
   -e HOST_UID="$(id -u)" -e HOST_GID="$(id -g)" \
-  rig-wizard:wizard install dev
+  rig-deps:deps install dev
 ```
 
-The image name follows the directory, like everything else here: in `spr/rig`
-it is `rig-wizard`, in a copy called `spr/acme-rig` it is `acme-rig-wizard`. The
-tag is `wizard` (or `full`, below), not `latest`.
+The image name follows the directory, like everything else here: in `rig`
+it is `rig-deps`, in a copy called `acme-rig` it is `acme-rig-deps`. The
+tag is `deps` (or `full`, below), not `latest`.
 
 None of the four arguments are guessable, so:
 
-- **`/:/host:ro`** — the wizard reads the *host's* `/etc/os-release` and
+- **`/:/host:ro`** — the installer reads the *host's* `/etc/os-release` and
   `/etc/wsl.conf`, not the container's. `HOST_ROOT=/host` is already baked into
   the image; this is what it points at. Read-only, and it is the only reason
   detection inside a container tells you anything about the machine.
@@ -122,7 +121,7 @@ None of the four arguments are guessable, so:
   and how it counts kind clusters already running.
 - **`/out/bin`** — the image's `OUT_BIN`. Whatever you mount here is where the
   four binaries land.
-- **`HOST_UID` / `HOST_GID`** — the wizard runs as root so it can reach that
+- **`HOST_UID` / `HOST_GID`** — the installer runs as root so it can reach that
   socket, which means everything it writes into a mounted volume is root-owned
   and useless to you. These drive the `chown` back. Omit them and the install
   looks like it worked.
@@ -131,18 +130,18 @@ None of the four arguments are guessable, so:
 tooling — which is the right answer on a managed or corporate-issued machine and
 is why the split exists.
 
-Then put them on PATH, which the wizard will remind you about because it cannot
+Then put them on PATH, which the installer will remind you about because it cannot
 edit your shell for you:
 
 ```bash
 export PATH="$HOME/.local/bin:$PATH"     # and add the same line to ~/.bashrc
 ```
 
-If something else on this machine already provides `kubectl`, the wizard says so
+If something else on this machine already provides `kubectl`, the installer says so
 by name rather than shadowing it quietly. `OUT_BIN=$PWD/def/bin` installs
 somewhere private instead.
 
-**Two variants worth knowing before you need them.** `make wizard full` bakes
+**Two variants worth knowing before you need them.** `make deps-image full` bakes
 every pinned binary into the image at build time (`DEPS_SOURCE=baked`), so
 `docker save` gives you the entire installer as one file to carry into an
 air-gapped network. And `DEPS_SOURCE=artifactory` with `DEPS_ARTIFACTORY_URL`
@@ -195,8 +194,8 @@ follows is only the mechanical part.
 
 ```bash
 SLUG=<slug>                                    # short, lowercase, no separators
-cp -r ~/wdir/all/projects/templates/broad ~/wdir/"$SLUG"
-cd ~/wdir/"$SLUG"
+cp -r ~/wdir/semester/all/projects/templates/broad ~/wdir/semester/"$SLUG"
+cd ~/wdir/semester/"$SLUG"
 grep -rl '<slug>' ctrl | xargs sed -i "s/<slug>/$SLUG/g"
 cp ctrl/k8s/.env.example ctrl/k8s/.env
 git init && git add -A && git commit -m "scaffold $SLUG from broad"
@@ -214,7 +213,7 @@ scaffold's Makefile from the directory, so there is nothing to edit for either.
 is already in use, so copying it unchanged puts two projects on one port:
 
 ```bash
-grep -h '^TILT_PORT=' ~/wdir/*/ctrl/k8s/.env 2>/dev/null | sort
+grep -h '^TILT_PORT=' ~/wdir/semester/*/ctrl/k8s/.env 2>/dev/null | sort
 ```
 
 Choose a free one in `10300–10399` — the range ALL reserves in
@@ -244,12 +243,22 @@ The workload is an nginx placeholder so a fresh copy reaches something that
 answers; replace it. Keep `30080` in step between the overlay patch and
 `kind-config.yaml`'s `containerPort` — the hostPort is this project's to pick.
 Reachability is a plain kind port mapping: no ingress controller and no MetalLB.
-Caddy maps `<slug>.local.ar` onto the host port (`~/wdir/ppl/local/Caddyfile`),
+Caddy maps `<slug>.local.ar` onto the host port (`~/wdir/semester/ppl/local/Caddyfile`),
 with `*.local.ar` resolving to 127.0.0.1 through dnsmasq. That is the whole chain.
 
-**The one file the scaffold still does not ship is `ctrl/Tiltfile`** — `make
-tilt-up` runs `cd ctrl && tilt up`, and there is nothing to run until you write
-one. Copy it from a live project; `unt` and `nvi` are closest to the plain shape.
+**For `ctrl/Tiltfile`, copy rig's** rather than a live project's. rig ships one
+that derives its cluster, context, ports and manifest directory from
+`ctrl/ports.sh active` instead of hardcoding a slug, and carries a catalogue of
+the blocks every project here ends up needing. Copying from `unt` or `nvi` is
+what the estate did until now, and it is why the same Tiltfile preamble exists
+in six places with the slug typed in by hand five times each.
+
+> **Two things in this document disagree with rig and are not settled.** It
+> mandates Tilt ports in `10300–10399`, while rig derives a block from the
+> directory name at `20000+` so copies cannot collide — a rig-managed project
+> takes rig's. And it names `ctrl/k8s/.env.example`, which is the `broad`
+> scaffold's layout; rig's is `ctrl/.env.example`. Both are this document
+> describing the house scaffold from inside rig's tree.
 
 
 ## Run it
@@ -270,9 +279,9 @@ delete-and-recreate for when a cluster wedges.
 ## Register it
 
 The project exists; now it is findable. Add an entry to
-`~/wdir/all/projects/index.json` and write its `projects/<slug>.md` beside the
+`~/wdir/semester/all/projects/index.json` and write its `projects/<slug>.md` beside the
 others. Structured fields in the index, prose in the markdown.
 
 Putting it on the CI server and deploying it is `ppl`'s half, and it starts at
-`~/wdir/ppl/ctrl/init-repo.sh` — gitea remote, then Woodpecker. That is a
+`~/wdir/semester/ppl/ctrl/init-repo.sh` — gitea remote, then Woodpecker. That is a
 different document.

rig/Makefile

@@ -16,11 +16,27 @@
 # Identity follows the FOLDER NAME, so this directory can be copied elsewhere,
 # renamed, and run as a separate environment with no edits. ctrl/.env overrides
 # it when you want a name that differs from the directory.
+#
+# Asked once, of ctrl/ports.sh, which resolves it through lib/config.sh:
+#
+#   CLUSTER KUBECONTEXT HTTP HTTPS TILT REGISTRY MANIFESTS_DIR
+#
+# Read positionally, so the order is a contract — ctrl/selftest.sh pins it.
+#
+# This used to be sed over ctrl/.env plus a slug computed here, which is a
+# SECOND derivation of values lib/config.sh already owns — and the two could
+# disagree about the port after `make ports persist`, or about the name for any
+# directory whose sanitised form differs from its raw one. One source now; the
+# Tiltfile reads the same line.
+FACTS     := $(shell bash ctrl/ports.sh active 2>/dev/null)
 SLUG      := $(shell echo '$(notdir $(CURDIR))' | tr '[:upper:]' '[:lower:]' | tr -c 'a-z0-9-' '-' | sed 's/^-*//; s/-*$$//')
-CLUSTER   := $(or $(shell sed -n 's/^CLUSTER=//p' ctrl/.env 2>/dev/null),$(SLUG))
-KCTX      := --context kind-$(CLUSTER)
-TILT_PORT := $(shell sed -n 's/^TILT_PORT=//p' ctrl/.env 2>/dev/null)
-WIZARD    := $(SLUG)-wizard
+# The fallback matters: ports.sh sources config.sh, and if a profile or .env is
+# broken it exits non-zero. Losing the cluster name would send --context to the
+# wrong place, so fall back to the folder rather than to empty.
+CLUSTER   := $(or $(word 1,$(FACTS)),$(SLUG))
+KCTX      := --context $(or $(word 2,$(FACTS)),kind-$(SLUG))
+TILT_PORT := $(word 5,$(FACTS))
+DEPSIMG   := $(SLUG)-deps
 
 # Words after the target become the script's subcommand. Make would otherwise
 # treat them as goals of their own, so each gets a no-op rule.
@@ -35,7 +51,7 @@ $(eval $(ARGS):;@:)
 .PHONY: $(ARGS)
 endif
 
-.PHONY: help setup station deps wizard cluster registry addons ports \
+.PHONY: help setup check selftest mem deps deps-image pins cluster registry addons ports \
         newbox dockerhost docs tilt \
         kind-up kind-down kind-reset tilt-up tilt-down
 
@@ -47,16 +63,28 @@ help:                          ## list targets
 setup:                         ## prepare this machine  [core] [--share-docker] [--cluster]
 	bash ctrl/setup.sh $(ARGS)
 
-station:                       ## is this workstation ready? reports, never fixes
-	bash ctrl/station.sh
+check:                         ## is this machine ready? reports, never fixes
+	bash ctrl/check.sh
+
+# The counterpart to check: that one asks about the MACHINE and never fails,
+# this one asks about RIG and exits 1, the way pins does. The checks are written
+# as the decisions they defend, so a failure names what is being undone.
+selftest:                      ## does rig still do what it says? exits 1 if not
+	bash ctrl/selftest.sh
+
+mem:                           ## memory, and any cap holding it  [status|backup|restore]
+	bash ctrl/mem.sh $(or $(ARGS),status)
 
 deps:                          ## install the toolchain  [core|dev]  (default dev)
-	bash ctrl/wizard.sh install $(or $(ARGS),dev)
+	bash ctrl/deps.sh install $(or $(ARGS),dev)
+
+pins:                          ## standalone/rigdeps.sh still installs what rig pins?
+	bash ctrl/pins.sh
 
-wizard:                        ## build the installer image  [full]
-	docker build -f ctrl/Dockerfile.wizard \
-	  --target $(if $(filter full,$(ARGS)),wizard-full,wizard) \
-	  -t $(WIZARD):$(if $(filter full,$(ARGS)),full,wizard) .
+deps-image:                    ## build the installer image  [full]
+	docker build -f ctrl/Dockerfile.deps \
+	  --target $(if $(filter full,$(ARGS)),deps-full,deps) \
+	  -t $(DEPSIMG):$(if $(filter full,$(ARGS)),full,deps) .
 
 # ── cluster ────────────────────────────────────────────────────────────────
 
@@ -85,11 +113,16 @@ dockerhost:                    ## share Docker between distros  [status|share|un
 docs:                          ## documentation  [serve|graphs]        (default serve)
 	bash ctrl/docs.sh $(or $(ARGS),serve)
 
-# --port is only passed when TILT_PORT is actually set. It comes from ctrl/.env,
-# which does NOT carry it by default — ports are derived at runtime in
-# lib/config.sh unless `make ports persist` has written them. Without the guard
-# tilt receives a bare `--port` with no value and fails on the flag rather than
-# on anything real. `make ports show` prints the derived block.
+# --port is only passed when TILT_PORT resolved. It normally does, since FACTS
+# above asks ports.sh — but ports.sh can fail on a broken profile, and without
+# the guard tilt receives a bare `--port` with no value and fails on the flag
+# rather than on anything real. Tilt's own default is 10350, which is the number
+# every project on this machine is trying not to collide on, so falling back to
+# it silently is worse than not passing the flag.
+#
+# The Tiltfile asks ports.sh for the rest itself — cluster, registry and where
+# the manifests are — so nothing needs passing here beyond what tilt's own flags
+# require.
 tilt:                          ## dev loop  [up|down]                  (default up)
 	cd ctrl && tilt $(or $(ARGS),up) $(KCTX) $(if $(filter down,$(ARGS)),,$(if $(TILT_PORT),--port $(TILT_PORT)))
 
@@ -104,6 +137,11 @@ tilt:                          ## dev loop  [up|down]                  (default
 #
 # `cluster list` and `cluster free` have no hyphenated twin on purpose — they
 # are rig's own, with nothing to be consistent with.
+#
+# Nothing outside this file reads these names: the script is `ctrl/cluster.sh`
+# and it takes the verb. So rename them, delete the ones you never type, or add
+# the spelling your own projects use — an alias is two lines, and adding one
+# costs nothing but a line in .PHONY above.
 
 kind-up:                       ## alias for `cluster up`
 	bash ctrl/cluster.sh up
@@ -114,10 +152,10 @@ kind-down:                     ## alias for `cluster down`
 kind-reset:                    ## alias for `cluster reset`
 	bash ctrl/cluster.sh reset
 
-# These two match the other projects' spelling, but rig has no Tiltfile — there
-# is nothing to run yet, and they fail the same way `make tilt` does.
-tilt-up:                       ## alias for `tilt up`   (rig has no Tiltfile yet)
+# These two match the other projects' spelling. rig ships ctrl/Tiltfile, so they
+# run — it deploys the examples in k8s/base until you replace them.
+tilt-up:                       ## alias for `tilt up`
 	cd ctrl && tilt up $(KCTX) $(if $(TILT_PORT),--port $(TILT_PORT))
 
-tilt-down:                     ## alias for `tilt down` (rig has no Tiltfile yet)
+tilt-down:                     ## alias for `tilt down`
 	cd ctrl && tilt down $(KCTX)

rig/README.md

@@ -9,6 +9,46 @@ topology, not the workloads.
 
 **Docker.** Nothing else — no curl, no jq, no python, no apt repositories.
 
+### Starting from plain Windows
+
+Everything here is bash and runs *inside* a Linux shell, so on a Windows machine
+that means WSL. Nothing in rig installs WSL, and nothing will: `wsl --install`
+enables Windows features and requires a reboot, which is not something a script
+should do to a machine on your behalf — and there is no tested undo for it.
+
+From an elevated PowerShell or Command Prompt, once:
+
+```powershell
+wsl --install
+```
+
+Then reboot and open the Linux shell it installed.
+
+**If you cloned this on the Windows side, copy it into WSL before carrying on.**
+WSL can reach the Windows drives at `/mnt/c`, and working from there mostly
+functions — slowly — but file watching does not: that filesystem raises no
+inotify events, so anything watching for edits silently stops seeing them.
+
+```bash
+cp -r /mnt/c/Users/<you>/rig ~/rig
+cd ~/rig
+```
+
+`make deps` reports it if you are running from `/mnt/...`. Then carry on below.
+
+If it fails, the usual causes give unhelpful messages:
+
+| symptom | cause |
+| --- | --- |
+| "the virtual machine could not be started" | virtualization disabled in BIOS/UEFI |
+| the command is not recognised | Windows build too old — needs 2004 or later |
+| the install starts, then nothing works | a reboot is still pending |
+
+Running the scripts from **Git Bash, MSYS or Cygwin does not work** — those look
+close enough to a Linux shell to get started and then fail without `/proc` or a
+docker socket. `ctrl/deps.sh` detects that and says so rather than letting you
+find out the slow way.
+
 ## Read the docs first
 
 ```bash
@@ -21,16 +61,65 @@ instructions for everything else. No cluster and no toolchain required.
 ## Then
 
 ```bash
-make station        # report host and config problems; changes nothing
+make check         # report host and config problems; changes nothing
 make deps          # install the toolchain (add `core` on a managed machine)
 make cluster up    # build the cluster for the active profile
 ```
 
+`make cluster up` also starts this environment's local registry and wires it
+into the node, so an image built locally is pullable by the cluster without
+going near docker.io:
+
+```bash
+make registry status        # prints: endpoint  localhost:<port>
+docker build -t localhost:<port>/app:1 .
+docker push localhost:<port>/app:1
+kubectl --context kind-$(basename $PWD) run app --image=localhost:<port>/app:1
+```
+
+The port block is derived from the directory name, so two copies of rig never
+collide:
+
+```bash
+make ports show                            # HTTP / HTTPS / TILT / REGISTRY
+make cluster list                          # every cluster on this machine, with memory
+make cluster free                          # stop the others if memory is tight
+make cluster down                          # remove this cluster and its registry
+```
+
+**The verbs are yours to change.** `cluster` is the script — `ctrl/cluster.sh` —
+and every spelling above is a `Makefile` target that calls it. `make kind-up` is
+an alias for `make cluster up`, kept because the other projects on this machine
+answer to that spelling and muscle memory spans repos rather than stopping at
+one. Nothing outside the `Makefile` reads these names, so rename them, drop the
+ones you never type, or add whatever your own projects already say: each alias
+is two lines at the bottom of the file, calling the same script the canonical
+target does.
+
+**`make tilt` works on a fresh copy, unedited.** rig ships `ctrl/Tiltfile`, and
+`k8s/base` already boots, so the dev loop comes up with the two examples running
+and nothing to configure first.
+
+It hardcodes nothing. It asks `ctrl/ports.sh active` for this environment's
+cluster name, kube context, ports and manifest directory — the same values every
+other rig script resolves through `ctrl/lib/config.sh` — so a copied and renamed
+rig deploys into its own cluster with no edits. Every other project here writes
+its slug into the Tiltfile five or six times by hand, which is exactly the
+collision `kind-config.yaml.tpl` exists to avoid.
+
+What it deploys is `MANIFESTS_DIR`, defaulting to rig's own `ctrl/k8s/overlays/dev`.
+Point that at an overlay versioned elsewhere and rig stops owning the manifests.
+
+Replace the examples, then add your images and resources in the two marked
+sections. The catalogue below them holds the blocks that recur across every
+project here — `docker_build`, resource ordering, gateway reload, port-forwards —
+with the parts that are easy to get wrong already commented.
+
 `make help` lists every target.
 
 On a machine where Docker really is the only thing installed, `make deps` has
 nothing to download with — see [BOOTSTRAP.md](BOOTSTRAP.md), which runs the
-toolchain through the wizard container and carries on to scaffolding and running
+toolchain through the installer container and carries on to scaffolding and running
 a new project.
 
 ## One directory is one environment
@@ -39,10 +128,10 @@ Copy this directory, rename it, run it. Cluster name, kubectl context, image
 tags and the host port block all derive from the directory name, so copies never
 collide and neither one's teardown can touch the other.
 
-rig lives inside soleprint, at `spr/rig` — it is soleprint's cluster half, and a
-copy is a **sibling**: `spr/acme-rig`. That is why the ignore rules for client
-rigs sit in `spr/.gitignore` rather than here; a rule in this directory cannot
-see a directory beside it.
+A copy of this directory is a **sibling** of it, named after the environment it
+models (`acme-rig`). That is why the ignore rules for copies sit in the *parent*
+repo's `.gitignore` rather than here: a rule in this directory cannot see a
+directory beside it.
 
 ## Profiles
 
@@ -54,7 +143,7 @@ apiserver audits. They live in `ctrl/env.d/`, and the active one is `PROFILE`.
 | `minimal` | the default. One node, no addons, boots fast. |
 | `client` | the regulated-estate shape — multi-node, audit on, registry mirror. |
 | `offline` | air-gapped: everything from a preloaded local registry. |
-| `data` | the dependency containers a soleprint room asks for. |
+| `data` | the cabinets an environment asks for. |
 
 ```bash
 PROFILE=data make cluster up
@@ -98,10 +187,10 @@ cluster does.
 | `redis` | cache and broker |
 | `airflow` | scheduled pipelines; needs postgres and redis |
 
-The last three are the cluster half of **soleprint's cabinets**. A room declares
-what it needs once, in `cfg/<room>/data/cabinets.json`; soleprint's `build.py`
-composes those services into `docker-compose.yml` for a laptop, and these
-install the same ones here. The names match on purpose — each cabinet carries a
+The last three are **cabinets**: a public service dropped in as-is, the upstream
+image unmodified, reachable at a known address. A cabinet is declared once and
+installs on either target — a `service.yml` composes it for a laptop, and these
+install the same one here. The names match on purpose: each cabinet carries a
 `rig_addon` field pointing at `ctrl/addons/<name>.sh`.
 
 Plain manifests rather than helm charts, like every other addon: a chart repo is

rig/ctrl/.env.example

@@ -26,10 +26,10 @@ PROFILE=minimal
 #   MANIFESTS_DIR=../platform-manifests/overlays/dev
 MANIFESTS_DIR=ctrl/k8s/overlays/dev
 
-# Where the wizard fetches the pinned binaries from.
+# Where the installer fetches the pinned binaries from.
 #   upstream    GitHub releases / dl.k8s.io  (needs internet)
 #   artifactory a generic repo — what a locked-down client usually allows
-#   baked       already inside the wizard image; no network at all
+#   baked       already inside the installer image; no network at all
 DEPS_SOURCE=upstream
 DEPS_ARTIFACTORY_URL=
 
@@ -43,7 +43,7 @@ REGISTRY_PASSWORD=
 # Corporate root CA, if Artifactory is fronted by an internal CA (it usually is).
 # Trust has to reach THREE places and nothing does it for you: the host docker
 # daemon, every kind node's containerd, and any in-cluster client. registry.sh
-# handles the first two; station.sh reports when it's configured but not trusted.
+# handles the first two; check.sh reports when it's configured but not trusted.
 # Symptom when missing: x509: certificate signed by unknown authority
 REGISTRY_CA_FILE=
 

rig/ctrl/Dockerfile.deps

@@ -1,18 +1,18 @@
-# The installation wizard. It does NOT run the cluster — it installs a toolchain
+# The toolchain installer image. It does NOT run the cluster — it installs a toolchain
 # onto the host and gets out of the way.
 #
 # This exists to kill a bootstrap paradox: a plain bash installer needs curl, jq
 # and sha256sum to already be present, and a minimal Debian has none of them.
-# The wizard carries its own toolchain, so the only host prerequisite is Docker.
+# It carries its own toolchain, so the only host prerequisite is Docker.
 #
 # Two variants from one file:
-#   docker build -f ctrl/Dockerfile.wizard --target wizard      -t <slug>-wizard        .
-#   docker build -f ctrl/Dockerfile.wizard --target wizard-full -t <slug>-wizard:full   .
+#   docker build -f ctrl/Dockerfile.deps --target deps        -t <slug>-deps        .
+#   docker build -f ctrl/Dockerfile.deps --target deps-full -t <slug>-deps:full   .
 #
-# wizard-full bakes every pinned binary in at build time. `docker save` it and
+# deps-full bakes every pinned binary in at build time. `docker save` it and
 # you have the whole installer as one file to carry into an air-gapped network.
 
-FROM debian:trixie-slim AS wizard
+FROM debian:trixie-slim AS deps
 
 # ca-certificates + curl: fetch and verify. graphviz + python3: render diagrams
 # and validate the arch model, so the host never needs an apt package.
@@ -26,21 +26,21 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
 
 WORKDIR /work
 COPY ctrl/versions.env /work/ctrl/versions.env
-COPY ctrl/wizard.sh    /work/ctrl/wizard.sh
-RUN chmod +x /work/ctrl/wizard.sh
+COPY ctrl/deps.sh    /work/ctrl/deps.sh
+RUN chmod +x /work/ctrl/deps.sh
 
 # Defaults; every one is overridable with -e at run time.
 ENV DEPS_SOURCE=upstream \
     OUT_BIN=/out/bin \
     HOST_ROOT=/host
 
-ENTRYPOINT ["/work/ctrl/wizard.sh"]
+ENTRYPOINT ["/work/ctrl/deps.sh"]
 CMD ["install"]
 
 
 # ---------------------------------------------------------------------------
-# wizard-full — same wizard, binaries baked in, works with no network at all.
-FROM wizard AS wizard-full
-RUN /work/ctrl/wizard.sh fetch --to /opt/rig/bin
+# deps-full — same image, binaries baked in, works with no network at all.
+FROM deps AS deps-full
+RUN /work/ctrl/deps.sh fetch --to /opt/rig/bin
 ENV DEPS_SOURCE=baked \
     BAKED_BIN=/opt/rig/bin

rig/ctrl/Dockerfile.example

@@ -0,0 +1,50 @@
+# EXAMPLE — a component image. Copy, rename, replace.
+#
+# Named like the manifest it feeds and the resource it becomes:
+#
+#   ctrl/Dockerfile.api   ->  image <cluster>-api  ->  image: in k8s/base/api.yaml
+#
+# That image string is the ONLY thing connecting the three. Nothing checks it;
+# a typo shows up as a pod stuck in ImagePullBackOff pulling from the public
+# index, which reads like a network problem and is not one.
+#
+# ── the one that catches everyone ──────────────────────────────────────────
+# The Tiltfile passes two paths with DIFFERENT bases, in adjacent arguments:
+#
+#   context='..'                 the REPO ROOT   (the Tiltfile is in ctrl/)
+#   dockerfile='Dockerfile.api'  relative to the TILTFILE, so ctrl/Dockerfile.api
+#
+# So every COPY below is resolved against the repo root, NOT against this file's
+# directory. A file sitting right beside this one is still reached as `ctrl/`:
+#
+#   COPY ctrl/nginx.conf /etc/nginx/conf.d/default.conf     # correct
+#   COPY nginx.conf      /etc/nginx/conf.d/default.conf     # fails — no such file
+#
+# Nothing warns you. The build just cannot find a file that is visibly there.
+
+FROM python:3.12-slim
+
+WORKDIR /app
+
+# Dependencies first, in their own layer: they change far less often than the
+# code, so a source edit does not reinstall them on every rebuild.
+COPY api/requirements.txt ./
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Repo-root relative — see above.
+COPY api/ ./api/
+
+# Match this with the containerPort in the manifest and the target of the
+# Service in front of it.
+EXPOSE 8000
+
+CMD ["python", "-m", "api"]
+
+# ── live_update ────────────────────────────────────────────────────────────
+# The sync in the Tiltfile's docker_build must land where this image expects it:
+#
+#   live_update=[sync('../api', '/app/api')]
+#
+# matches `COPY api/ ./api/` with WORKDIR /app. If the two disagree, Tilt syncs
+# into a path nothing reads and the container keeps serving the built copy —
+# edits appear to do nothing, with no error anywhere.

rig/ctrl/Tiltfile

@@ -0,0 +1,139 @@
+# The dev loop. `make tilt` from the project root, or `cd ctrl && tilt up`.
+#
+# This file ships with rig and works unedited: rig's own k8s/base already boots,
+# so `make tilt` comes up with a running cluster and no editing at all. What it
+# deploys is two EXAMPLES — replace them, and add your own images and resources
+# in the two marked sections near the bottom. The catalogue after them has the
+# blocks to paste, with the parts that are easy to get wrong already commented.
+#
+# rig supplies this file; it does not own it. Nothing in rig reads it back, and
+# nothing here is regenerated — edit it freely, the way you would edit
+# k8s/base/example-mock.yaml. rig owns the machine, you own the workload.
+#
+# Nothing below is hardcoded to this directory, deliberately. Every other
+# project here writes its slug into the Tiltfile five or six times by hand, so a
+# copy of the project deploys into the original's cluster until someone
+# remembers to edit all of them. A rig is meant to be copied and renamed, so it
+# asks instead.
+
+# ── who we are, and on which ports ─────────────────────────────────────────
+# One question to rig, answered by ctrl/ports.sh, which resolves it through
+# lib/config.sh — the same path every other rig script takes. That is the point:
+# the cluster name is NOT the bare directory name (it is lowercased and reduced
+# to a DNS label), and the ports honour anything pinned in ctrl/.env. Recomputing
+# either of those here in Starlark is how two copies end up disagreeing about
+# which cluster they are talking to.
+_facts   = str(local('bash ports.sh active', quiet=True)).split()
+CLUSTER  = _facts[0]
+CTX      = _facts[1]
+HTTP     = _facts[2]
+HTTPS    = _facts[3]
+TILT     = _facts[4]
+REGISTRY = _facts[5]
+
+# Where the manifests live. rig's own are the default; point MANIFESTS_DIR in
+# ctrl/.env at an overlay versioned somewhere else and rig stops owning them —
+# see k8s/README.md. Real manifests usually change on a different cadence, by
+# different people, under different review.
+#
+# The value is REPO-ROOT relative, because that is the root everything else in
+# rig is expressed against. This file runs in ctrl/, so prefix rather than
+# assume: '../' + 'ctrl/k8s/overlays/dev' and '../' + '../platform/overlays/dev'
+# are both right, where stripping a leading 'ctrl/' would only fix the first.
+MANIFESTS = '../' + _facts[6]
+
+# ── refuse to deploy into the wrong cluster ────────────────────────────────
+# Tilt snapshots the kubectl context at startup, BEFORE parsing this file, so it
+# cannot be switched from here — only refused. `make tilt` passes --context for
+# you; this catches a bare `tilt up` after some other project moved the global
+# context.
+allow_k8s_contexts(CTX)
+if k8s_context() != CTX:
+    fail("Wrong kubectl context: '%s'. This is %s — run: make tilt, or tilt up --context %s"
+         % (k8s_context(), CLUSTER, CTX))
+
+# The namespace has to exist before anything lands in it, and kustomize does not
+# guarantee ordering across resources. Creating it here is idempotent.
+local('kubectl --context %s create namespace %s --dry-run=client -o yaml | kubectl --context %s apply -f -'
+      % (CTX, CLUSTER, CTX), quiet=True)
+
+# ── images go to this environment's own registry ───────────────────────────
+# Fail closed. Tilt can usually infer the kind registry on its own, but "usually"
+# is an inference, and when it misses, an unqualified name like 'app' quietly
+# means docker.io/library/app — a push to the public index instead of the
+# registry two lines away. rig runs that registry; name it.
+default_registry('localhost:' + REGISTRY)
+
+k8s_yaml(kustomize(MANIFESTS))
+
+# ── Images ─────────────────────────────────────────────────────────────────
+# (nothing yet — rig's examples run upstream images. Add docker_build calls here.)
+
+
+# ── Resources ──────────────────────────────────────────────────────────────
+# (nothing yet — add k8s_resource calls here to name and order what you deploy.)
+
+
+# Everything with no dev loop of its own, gathered so it does not clutter the UI.
+k8s_resource(
+    objects=[CLUSTER + ':namespace'],
+    new_name='infra',
+)
+
+# ═══════════════════════════════════════════════════════════════════════════
+# Catalogue — paste what you need, delete the rest.
+#
+# These are the shapes that recur across every project here, with the reasoning
+# kept next to them. They are comments so this file runs as-is.
+# ═══════════════════════════════════════════════════════════════════════════
+#
+# ── build an image ─────────────────────────────────────────────────────────
+# The one genuinely non-obvious thing in the whole corpus: `context` and
+# `dockerfile` are relative to DIFFERENT directories, in adjacent arguments,
+# and nothing warns you.
+#
+#   context=    the REPO ROOT — this file is in ctrl/, so '..'
+#   dockerfile= relative to THIS file — so 'Dockerfile.api' is ctrl/Dockerfile.api
+#
+# Every COPY inside those Dockerfiles is therefore repo-root relative: a file
+# sitting BESIDE the Dockerfile is still reached as `COPY ctrl/nginx.conf`.
+#
+# docker_build(
+#     CLUSTER + '-api',                  # must match `image:` in the manifest —
+#     context='..',                      #   that string is the only thing
+#     dockerfile='Dockerfile.api',       #   connecting the two
+#     ignore=['.git', 'def', '.venv', 'node_modules', '__pycache__'],
+#     live_update=[sync('../api', '/app/api')],
+# )
+#
+# ── name and order a resource ──────────────────────────────────────────────
+# k8s_resource('api', resource_deps=['postgres'], labels=['app'])
+# k8s_resource('gateway', resource_deps=['api', 'ui'], labels=['app'])
+#
+# ── reload the gateway when its config changes ─────────────────────────────
+# A Caddyfile arriving via configMapGenerator with disableNameSuffixHash does
+# NOT roll the pod — the ConfigMap name never changes, so nothing tells the
+# Deployment anything happened. Without this you edit the routes and watch
+# nothing take effect.
+#
+# local_resource(
+#     'gateway-reload',
+#     cmd='kubectl --context %s -n %s rollout restart deployment/gateway' % (CTX, CLUSTER),
+#     deps=['k8s/base/Caddyfile'],
+#     resource_deps=['gateway'],
+#     auto_init=False,
+# )
+#
+# ── an overlay whose secretGenerator reads outside its own directory ───────
+# kustomize refuses to read above the kustomization root unless told to. Only
+# add this if you actually have such a generator; it loosens a safety check.
+#
+# k8s_yaml(kustomize(MANIFESTS, flags=['--load-restrictor=LoadRestrictionsNone']))
+#
+# ── reach a service directly, bypassing the gateway ────────────────────────
+# For a DB client or an admin UI. Prefer routing through the gateway: host ports
+# are a single shared namespace across every project on this machine, which is
+# why rig derives a block per environment in the first place. If you do need
+# one, take it from this environment's own block rather than picking a number.
+#
+# k8s_resource('postgres', port_forwards=[str(int(HTTP) + 5) + ':5432'])

rig/ctrl/addons/airflow.sh

@@ -1,5 +1,5 @@
 #!/usr/bin/env bash
-# Apache Airflow — the cluster half of soleprint's airflow cabinet.
+# Apache Airflow — the cluster half of the airflow cabinet.
 #
 # Airflow needs a metadata database before it will start at all, so this refuses
 # rather than rolls a pod that will CrashLoopBackOff while the real problem
@@ -7,7 +7,7 @@
 #
 # One pod on `standalone`, matching the compose cabinet: migration, admin user,
 # scheduler and webserver in a single container. The official chart's five
-# deployments model an installation; a room switching this on wants pipelines.
+# deployments model an installation; switching this on means wanting pipelines.
 set -euo pipefail
 cd "$(dirname "$0")/.."
 

rig/ctrl/addons/postgres.sh

@@ -1,9 +1,10 @@
 #!/usr/bin/env bash
-# PostgreSQL — the cluster half of soleprint's postgres cabinet.
+# PostgreSQL — the cluster half of the postgres cabinet.
 #
-# A room declares the dependency once, in cfg/<room>/data/cabinets.json. On a
-# laptop `build.py` composes it into docker-compose.yml; here it becomes a pod,
-# so the same declaration works either way and nothing has to be remembered
+# A cabinet is a public service dropped into the environment as-is — the
+# upstream image, unmodified, reachable at a known address. This is the cluster
+# half of it; the compose half is a `service.yml` beside a `cabinet.json`. The
+# declaration is made once and both paths read it, so nothing is remembered
 # twice.
 #
 # Plain manifests rather than a helm chart, matching the other addons: a chart
@@ -32,8 +33,8 @@ if $K get secret -n "$NS" postgres >/dev/null 2>&1; then
 else
     password=$(head -c 18 /dev/urandom | base64 | tr -d '/+=' | head -c 24)
     $K create secret generic postgres -n "$NS" \
-        --from-literal=POSTGRES_DB="${POSTGRES_DB:-soleprint}" \
-        --from-literal=POSTGRES_USER="${POSTGRES_USER:-soleprint}" \
+        --from-literal=POSTGRES_DB="${POSTGRES_DB:-postgres}" \
+        --from-literal=POSTGRES_USER="${POSTGRES_USER:-postgres}" \
         --from-literal=POSTGRES_PASSWORD="$password" >/dev/null
     echo "  generated a password (read it back with the command printed below)"
 fi

rig/ctrl/addons/redis.sh

@@ -1,5 +1,5 @@
 #!/usr/bin/env bash
-# Redis — the cluster half of soleprint's redis cabinet.
+# Redis — the cluster half of the redis cabinet.
 #
 # Cache, and the broker anything queue-shaped runs on. No persistence: a broker
 # that loses its queue on restart is the honest local model, and a PVC here buys

rig/ctrl/check.sh

@@ -0,0 +1,221 @@
+#!/usr/bin/env bash
+# Readiness check: is this machine ready to run rig?
+#
+# Reports and instructs; never silently fixes anything. Everything it finds is
+# either already fine, or something a human has to decide on.
+#
+# Runs ctrl/deps.sh host detection in a container when Docker is the only thing
+# installed, or directly when the toolchain is already present. Then adds the
+# checks that need this repo's config: profile sanity, CA trust, port clashes.
+set -euo pipefail
+cd "$(dirname "$0")"
+
+DEPS_IMAGE="${DEPS_IMAGE:-$(basename "$(cd .. && pwd)")-deps}"
+
+# Host detection. Prefer running it bare — it needs no dependencies beyond
+# coreutils — and fall back to the container only if this shell can't.
+bash ./deps.sh detect
+
+# ── repo-level checks ──────────────────────────────────────────────────────
+
+source ./lib/config.sh
+load_config
+
+echo
+echo "config"
+echo "  profile     ${PROFILE_NAME}  (nodes=${NODES} audit=${AUDIT})"
+echo "  cluster     ${CLUSTER}  (context ${KUBECONTEXT})"
+echo "  registry    ${REGISTRY_MODE}"
+echo "  ingress     ${INGRESS_MODE}"
+
+if [ ! -f ./.env ]; then
+    echo "  ! ctrl/.env missing — copy it:  cp ctrl/.env.example ctrl/.env"
+fi
+
+# ── memory ─────────────────────────────────────────────────────────────────
+#
+# A profile on a box that is already full is the most common first failure, and
+# it presents as pods stuck Pending rather than anything that says "memory".
+# Warns; never blocks. Whether to try anyway is the user's call.
+
+# A /proc/meminfo field in MB, 0 if absent. MEMINFO and OVERCOMMIT_FILE exist
+# only so the tight and does-not-fit branches can be exercised against another
+# machine's real numbers; in normal use they are the kernel's own files.
+mb_of() {
+    awk -v k="$1:" '$1 == k { printf "%d", $2 / 1024; found = 1 }
+                    END      { if (!found) printf "0" }' "${MEMINFO:-/proc/meminfo}"
+}
+
+# What one node costs, measured rather than guessed. On 2026-09-11 a minimal
+# control-plane node ran at 620 MiB idle and ~728 MiB with a small mock, plus
+# 16 MiB for the local registry — ~745 MiB of working set. 800 rounds that up,
+# and agrees with the 800 MB observed independently on a larger rig. Worker
+# nodes carry no etcd or apiserver and are lighter, so for a multi-node shape
+# this errs high. It is the cluster alone: whatever you deploy comes on top.
+NODE_MB=800
+
+# Every running container's working set in MB, tagged with the kind cluster it
+# belongs to ('-' when it is not kind). docker stats reports usage minus page
+# cache, which is what actually competes — cache is handed back under pressure.
+# Counting only kind would hide the usual culprit on a managed workspace, where
+# the memory is held by other containers entirely.
+container_mb() {
+    docker info >/dev/null 2>&1 || return 0
+    awk -F'\t' '
+        FILENAME == ARGV[1] { cl[$1] = ($2 == "" ? "-" : $2); if ($2 != "") isc[$2] = 1; next }
+        {
+            grp = ($1 in cl ? cl[$1] : "-")
+            # A kind cluster'"'"'s local registry is a plain container with no kind
+            # label, named <cluster>-registry, so on its own it would read as a
+            # stranger. It belongs to its cluster — but only if that cluster exists:
+            # a registry whose cluster is gone is a genuine stray, and says so.
+            if (grp == "-" && $1 ~ /-registry$/) {
+                base = $1; sub(/-registry$/, "", base)
+                if (base in isc) grp = base
+            }
+            split($2, u, " "); v = u[1]; mb = 0
+            if      (v ~ /GiB$/) { sub(/GiB$/, "", v); mb = v * 1024 }
+            else if (v ~ /MiB$/) { sub(/MiB$/, "", v); mb = v }
+            else if (v ~ /KiB$/) { sub(/KiB$/, "", v); mb = v / 1024 }
+            else if (v ~ /B$/)   { sub(/B$/,   "", v); mb = v / 1048576 }
+            printf "%d\t%s\t%s\n", mb, grp, $1
+        }
+    ' <(docker ps --format '{{.Names}}\t{{.Label "io.x-k8s.kind.cluster"}}' 2>/dev/null) \
+      <(docker stats --no-stream --format '{{.Name}}\t{{.MemUsage}}' 2>/dev/null)
+}
+
+total_mb=$(mb_of MemTotal)
+avail_mb=$(mb_of MemAvailable)
+swap_used_mb=$(( $(mb_of SwapTotal) - $(mb_of SwapFree) ))
+overcommit=$(cat "${OVERCOMMIT_FILE:-/proc/sys/vm/overcommit_memory}" 2>/dev/null || echo '?')
+need_mb=$(( NODES * NODE_MB ))
+
+rows=$(container_mb)
+# Once this environment's own cluster is running, its real footprint is already
+# out of MemAvailable and the per-node estimate stops being relevant. Subtracting
+# the measurement from the estimate would count the same memory twice, and a
+# running cluster that happens to sit under 800 MB would still "need" the gap.
+ours_mb=$(awk -F'\t' -v c="$CLUSTER" '$2 == c { s += $1 } END { print s + 0 }' <<< "$rows")
+still_mb=$(( ours_mb > 0 ? 0 : need_mb ))
+
+echo
+echo "memory"
+printf "  this profile ~%d MB   %s node(s) x %d MB — the cluster alone, your workload on top\n" \
+    "$need_mb" "$NODES" "$NODE_MB"
+if [ "$ours_mb" -gt 0 ]; then
+    printf "  already held %d MB    by '%s', which is up\n" "$ours_mb" "$CLUSTER"
+fi
+printf "  available    %d MB   of %d MB\n" "$avail_mb" "$total_mb"
+
+# The biggest things holding memory right now, other than this cluster: kind
+# clusters summed per cluster, everything else by container name.
+others=$(awk -F'\t' -v c="$CLUSTER" '
+            $2 != c && $2 != "-" && $2 != "" { k["kind cluster \x27" $2 "\x27"] += $1 }
+            $2 == "-"                         { k["container \x27" $3 "\x27"]    += $1 }
+            END { for (n in k) printf "%d\t%s\n", k[n], n }' <<< "$rows" | sort -rn)
+if [ -n "$others" ]; then
+    echo "  held elsewhere:"
+    head -6 <<< "$others" | awk -F'\t' '{ printf "    %6d MB  %s\n", $1, $2 }'
+    n_others=$(wc -l <<< "$others")
+    if [ "$n_others" -gt 6 ]; then
+        echo "    ... and $((n_others - 6)) more"
+    fi
+fi
+
+headroom=$(( avail_mb - still_mb ))
+if [ "$still_mb" -eq 0 ]; then
+    if [ "$headroom" -ge 512 ]; then
+        printf "  fits — already up; %d MB headroom for what you deploy\n" "$headroom"
+    else
+        printf "  ! already up, but only %d MB headroom for anything you deploy\n" "$headroom"
+    fi
+elif [ "$headroom" -ge 512 ]; then
+    printf "  fits — %d MB headroom for what you deploy\n" "$headroom"
+elif [ "$headroom" -ge 0 ]; then
+    printf "  ! fits, but only %d MB headroom for anything you deploy\n" "$headroom"
+else
+    printf "  ! does not fit right now: ~%d MB needed, %d MB available\n" "$still_mb" "$avail_mb"
+    # Two failures with opposite fixes, and telling them apart is the point.
+    if [ "$still_mb" -le "$total_mb" ]; then
+        echo "    The machine is big enough; something else is holding memory (above)."
+        echo "    Stopping that is what helps — a bigger VM would not."
+        if grep -q 'kind cluster' <<< "$others"; then
+            echo "    'make cluster free' stops the other kind clusters. It stops, never deletes."
+        fi
+    else
+        echo "    The machine itself is too small: ~${still_mb} MB needed, ${total_mb} MB total."
+    fi
+fi
+
+if [ "$swap_used_mb" -gt 0 ]; then
+    printf "  ! %d MB already in swap — available memory does not count it, so expect a\n" "$swap_used_mb"
+    echo "    cluster here to be slow well before it fails"
+fi
+if [ "$overcommit" = "1" ]; then
+    echo "  ! overcommit=1: allocations never fail here, so read 'fits' as a ceiling."
+    echo "    A cluster that starts cleanly can still lose processes to the OOM killer."
+fi
+
+# The CA reaches three places and only one of them is ours. Report the other two.
+if [ -n "${REGISTRY_CA_FILE:-}" ]; then
+    echo
+    echo "registry CA"
+    if [ ! -r "$REGISTRY_CA_FILE" ]; then
+        echo "  ! REGISTRY_CA_FILE not readable: $REGISTRY_CA_FILE"
+    else
+        echo "  file        $REGISTRY_CA_FILE"
+        host="${REGISTRY_REMOTE_URL#*://}"; host="${host%%/*}"
+        if [ -n "$host" ] && [ ! -f "/etc/docker/certs.d/${host}/ca.crt" ]; then
+            echo "  ! the HOST docker daemon does not trust it yet:"
+            echo "      sudo mkdir -p /etc/docker/certs.d/${host}"
+            echo "      sudo cp ${REGISTRY_CA_FILE} /etc/docker/certs.d/${host}/ca.crt"
+            echo "    (kind nodes are handled by registry.sh; in-cluster clients are the workload's job)"
+        fi
+    fi
+fi
+
+# Host ports this environment will try to bind. Checked before cluster creation
+# because docker reports a clash halfway through, as an opaque
+# "failed to bind host port ...: address already in use".
+echo
+echo "ports  (block derived from the directory name — see 'make ports')"
+
+port_busy() {
+    if command -v ss >/dev/null 2>&1; then
+        ss -ltn "sport = :$1" 2>/dev/null | grep -q LISTEN && return 0 || return 1
+    fi
+    # iproute2 is absent from a minimal Debian, so fall back to procfs rather
+    # than silently reporting everything as free.
+    local hex; hex=$(printf ':%04X' "$1")
+    grep -qi "^ *[0-9]*: [0-9A-F]*$hex " /proc/net/tcp /proc/net/tcp6 2>/dev/null
+}
+
+# A port held by THIS environment's own cluster is not a clash — it is the thing
+# working. Reporting it as a problem every time the cluster is up would train
+# people to ignore this section, which is the opposite of the point.
+# Extract with a second grep rather than `tr -d ':->'`: in tr, ':->' is the
+# character RANGE ':' to '>', which does not contain '-', so the trailing dash
+# survives and nothing ever matches.
+ours=$(docker ps --filter "label=io.x-k8s.kind.cluster=${CLUSTER}" \
+        --format '{{.Ports}}' 2>/dev/null | tr ',' '\n' \
+        | grep -oE ':[0-9]+->' | grep -oE '[0-9]+' || true)
+
+clash=0
+for entry in "HTTP:${HTTP_PORT}" "HTTPS:${HTTPS_PORT}" \
+             "TILT:${TILT_PORT}" "REGISTRY:${REGISTRY_PORT}"; do
+    name="${entry%%:*}"; p="${entry#*:}"
+    [ -n "$p" ] || continue
+    if ! port_busy "$p"; then
+        printf "    %-9s %-6s free\n" "$name" "$p"
+    elif echo "$ours" | grep -qx "$p"; then
+        printf "    %-9s %-6s in use by this environment's cluster\n" "$name" "$p"
+    else
+        printf "  ! %-9s %-6s IN USE by something else\n" "$name" "$p"
+        clash=1
+    fi
+done
+
+if [ "$clash" -eq 1 ]; then
+    echo "    override the clashing one in ctrl/.env, e.g.  HTTP_PORT=21080"
+    echo "    (or rename this directory — the whole block follows the name)"
+fi

rig/ctrl/cluster.sh

@@ -25,7 +25,7 @@ up() {
         # Say what this profile locks in BEFORE spending minutes building it:
         # the audit policy is an apiserver flag and cannot be changed later.
         echo "creating cluster '$CLUSTER' from profile '$PROFILE_NAME'"
-        echo "  shape       ctrl/k8s/$KIND_CONFIG"
+        echo "  shape       ${KIND_CONFIG_SHOWN}"
         echo "  nodes       $NODES"
         echo "  image       $NODE_IMAGE"
         echo "  audit       $AUDIT"

rig/ctrl/deps.sh

@@ -0,0 +1,621 @@
+#!/usr/bin/env bash
+# Toolchain installer: detect the host, install a pinned toolchain onto it, then
+# report what it could not do.
+#
+# It never runs the cluster, never uses sudo or apt, and writes only into
+# $OUT_BIN (default ~/.local/bin). Everything that would touch the host proper —
+# systemd, inotify limits, .wslconfig, docker group — is REPORTED for a human to
+# decide on, never performed. That is what makes it safe to run on a machine that
+# already has a working setup.
+#
+# Usage (normally via `make deps`, or directly):
+#   deps.sh detect            # report host facts only, change nothing
+#   deps.sh fetch [core|dev] [--to DIR]   # download + verify into DIR
+#   deps.sh install [core|dev]            # detect, fetch, install, report
+#
+# Tiers: 'core' is kubectl + jq (talk to a cluster); 'dev' adds kind and tilt
+# Default is dev.
+#
+# Runs both inside the installer container and bare on a host. Inside the
+# container, host files are read through $HOST_ROOT (mount / as :ro); bare, it
+# falls back to /.
+
+set -euo pipefail
+
+# Keep the caller's cwd so a relative --to resolves where the user expects,
+# not against ctrl/ once we've moved.
+INVOKED_FROM="$PWD"
+cd "$(dirname "$0")"
+
+source ./versions.env
+
+# Resolve a possibly-relative path against the caller's original directory.
+abspath() {
+    case "$1" in
+        /*) echo "$1" ;;
+        *)  echo "$INVOKED_FROM/$1" ;;
+    esac
+}
+
+OUT_BIN="${OUT_BIN:-$HOME/.local/bin}"
+HOST_ROOT="${HOST_ROOT:-/}"
+DEPS_SOURCE="${DEPS_SOURCE:-upstream}"
+DEPS_ARTIFACTORY_URL="${DEPS_ARTIFACTORY_URL:-}"
+BAKED_BIN="${BAKED_BIN:-/opt/rig/bin}"
+
+# Collected by detect(), printed by report_manual() at the very end.
+MANUAL=()
+
+# Host FILES (/etc/..., /mnt/c/...) must be read through the mount. Kernel-level
+# facts (kernel version, meminfo, inotify) are shared with the container, so the
+# container's own view is already the host's.
+# A /proc/meminfo field in MB, 0 if the field is absent. MEMINFO exists so the
+# tight and does-not-fit branches can be exercised against a real machine's
+# numbers from somewhere else; in normal use it is always /proc/meminfo.
+mb_of() {
+    awk -v k="$1:" '$1 == k { printf "%d", $2 / 1024; found = 1 }
+                    END      { if (!found) printf "0" }' "${MEMINFO:-/proc/meminfo}"
+}
+
+host_file() {
+    local p="${1#/}"
+    if [ "$HOST_ROOT" != "/" ] && [ -e "$HOST_ROOT/$p" ]; then
+        echo "$HOST_ROOT/$p"
+    else
+        echo "/$p"
+    fi
+}
+
+# ── detect ─────────────────────────────────────────────────────────────────
+
+# Windows outside WSL — Git Bash, MSYS, Cygwin — looks close enough to work and
+# then fails in a pile of confusing ways: no /proc, no docker socket, none of
+# the tooling. Detectable, so name it instead.
+require_linux() {
+    case "$(uname -s)" in
+        MINGW*|MSYS*|CYGWIN*)
+            cat >&2 <<'EOF'
+This has to run inside WSL, not Git Bash / MSYS / Cygwin.
+
+If WSL is not installed yet, from an elevated PowerShell or Command Prompt:
+
+    wsl --install
+
+That enables Windows features and needs a reboot, so it is not something this
+script will do for you. Afterwards, open the Linux shell it installs and run
+this from there.
+
+See "Starting from plain Windows" in README.md.
+EOF
+            exit 1 ;;
+    esac
+}
+
+is_wsl() { grep -qi microsoft /proc/version 2>/dev/null; }
+
+detect() {
+    echo "host"
+    echo "  kernel      $(uname -r)"
+
+    local osr; osr=$(host_file /etc/os-release)
+    [ -r "$osr" ] && echo "  distro      $(sed -n 's/^PRETTY_NAME="\(.*\)"/\1/p' "$osr")"
+
+    # In MB. Whole gigabytes lose nearly half a GB on exactly the machines where
+    # it matters: 1874 MB available used to print as "1 GB". Facts only — whether
+    # that is enough depends on the profile, which check.sh knows and this does not.
+    local total_mb avail_mb swap_total_mb swap_used_mb om
+    total_mb=$(mb_of MemTotal)
+    avail_mb=$(mb_of MemAvailable)
+    swap_total_mb=$(mb_of SwapTotal)
+    swap_used_mb=$(( swap_total_mb - $(mb_of SwapFree) ))
+    printf "  memory      %d MB total, %d MB available\n" "$total_mb" "$avail_mb"
+    if [ "$swap_total_mb" -gt 0 ]; then
+        printf "  swap        %d MB used of %d MB\n" "$swap_used_mb" "$swap_total_mb"
+    fi
+
+    # How the kernel answers an allocation it cannot really satisfy. With 1 it
+    # always says yes and settles up later with the OOM killer, so a cluster that
+    # starts cleanly can still lose processes afterwards.
+    om=$(cat "${OVERCOMMIT_FILE:-/proc/sys/vm/overcommit_memory}" 2>/dev/null || echo '?')
+    case "$om" in
+        0) echo "  overcommit  0 heuristic — allocations are granted on a guess" ;;
+        1) echo "  overcommit  1 always — every allocation succeeds; the OOM killer is the only limit" ;;
+        2) echo "  overcommit  2 strict — an allocation fails honestly instead of killing later" ;;
+    esac
+
+    detect_wsl
+    detect_filesystem
+    detect_docker
+    detect_inotify
+    detect_toolchain
+}
+
+detect_wsl() {
+    if ! is_wsl; then
+        echo "  platform    native linux"
+        return
+    fi
+
+    echo "  platform    WSL"
+
+    # systemd is off by default in WSL, and the ingress/DNS paths that use a
+    # host service need it. Enabling it requires a Windows-side restart, which
+    # cannot be issued from inside the distro.
+    local wc; wc=$(host_file /etc/wsl.conf)
+    if [ -r "$wc" ] && grep -qE '^\s*systemd\s*=\s*true' "$wc"; then
+        echo "  systemd     enabled in wsl.conf"
+    else
+        echo "  ! systemd not enabled in /etc/wsl.conf"
+        MANUAL+=("Enable systemd — add to /etc/wsl.conf:
+      [boot]
+      systemd=true
+  then from a WINDOWS terminal (not this shell):  wsl --shutdown")
+    fi
+
+    # WSL regenerates /etc/resolv.conf on every boot, which silently reverts any
+    # local DNS setup.
+    if [ -r "$wc" ] && grep -qE '^\s*generateResolvConf\s*=\s*false' "$wc"; then
+        echo "  resolv.conf pinned (generateResolvConf=false)"
+    else
+        echo "  - resolv.conf is WSL-generated; DNS_MODE=dnsmasq would be reverted on reboot"
+    fi
+
+    local wcfg
+    wcfg=$(ls "$HOST_ROOT"/mnt/c/Users/*/.wslconfig 2>/dev/null | head -1 || true)
+    if [ -n "$wcfg" ] && grep -qE '^\s*memory\s*=' "$wcfg"; then
+        echo "  wslconfig   memory set: $(grep -E '^\s*memory\s*=' "$wcfg" | tr -d ' ')"
+    else
+        MANUAL+=("Cap/raise the WSL VM memory — see what is set versus what booted:
+      make mem status
+  It prints the edit to make and the command to apply it.")
+    fi
+}
+
+# Not a path check: /mnt is an ordinary mount point and an ext4 disk mounted
+# there is perfectly fine. What matters is the filesystem. The Windows drives
+# arrive as 9p (WSL2) or drvfs (WSL1); network and fuse mounts behave the same
+# way. None of them deliver inotify events, so anything watching files goes
+# quiet without saying why.
+watch_hostile_fs() {
+    local dir="$1" fstype
+    fstype=$(findmnt -no FSTYPE --target "$dir" 2>/dev/null || true)
+    [ -n "$fstype" ] || fstype=$(stat -f -c %T "$dir" 2>/dev/null || true)
+    case "$fstype" in
+        9p|v9fs|drvfs|cifs|smb3|nfs|nfs4|fuse.sshfs|fuseblk) echo "$fstype" ;;
+        *) echo "" ;;
+    esac
+}
+
+detect_filesystem() {
+    local root fstype
+    root=$(cd .. && pwd -P)
+    fstype=$(watch_hostile_fs "$root")
+    if [ -n "$fstype" ]; then
+        echo "  ! this directory is on $fstype — file watching will not work"
+        MANUAL+=("Move this onto the local disk. Nothing watching files sees changes
+  on a $fstype mount, and everything else is slower:
+      cp -r \"$root\" ~/ && cd ~/$(basename "$root")")
+    else
+        echo "  filesystem  $root  ($(findmnt -no FSTYPE --target "$root" 2>/dev/null || echo local))"
+    fi
+}
+
+detect_docker() {
+    # Reachability of the daemon is the real question, and the CLI is only how
+    # we ask it. Note that when this runs inside the installer container, Docker
+    # necessarily exists on the host — otherwise nothing would be executing —
+    # so a missing CLI in here is an installer packaging bug, not a host problem.
+    if ! command -v docker >/dev/null 2>&1; then
+        if [ -S /var/run/docker.sock ]; then
+            echo "  docker      socket present (no cli in this context)"
+        else
+            echo "  ! docker not found and no socket at /var/run/docker.sock"
+            MANUAL+=("Install Docker — the one true prerequisite:
+      sudo apt-get install -y docker.io && sudo usermod -aG docker \"\$USER\"
+  then log out and back in.")
+        fi
+        return
+    fi
+    if docker info >/dev/null 2>&1; then
+        echo "  docker      $(docker version --format '{{.Server.Version}}' 2>/dev/null)"
+        local n
+        n=$(docker ps --filter "label=io.x-k8s.kind.cluster" --format '{{.Names}}' 2>/dev/null | wc -l)
+        # Must be an `if`, not `[ ] && echo`: as the last statement in this
+        # function the latter returns 1 when the count is zero, and `set -e`
+        # then kills the caller. That is the fresh-machine case — no clusters
+        # yet — so the bug only ever shows up where it does most harm.
+        if [ "$n" -gt 0 ]; then
+            echo "  - $n kind node container(s) already running; see 'make cluster list'"
+        fi
+    else
+        echo "  ! docker cli present but the daemon is unreachable"
+        MANUAL+=("Start Docker, or add yourself to the docker group:
+      sudo usermod -aG docker \"\$USER\"   # then log out and back in")
+    fi
+}
+
+# kind and Tilt both watch large trees. WSL ships defaults (8192/128) far too low,
+# and the failure mode is silent: Tilt simply stops noticing file changes.
+detect_inotify() {
+    local w i
+    w=$(cat /proc/sys/fs/inotify/max_user_watches 2>/dev/null || echo 0)
+    i=$(cat /proc/sys/fs/inotify/max_user_instances 2>/dev/null || echo 0)
+    echo "  inotify     watches=$w instances=$i"
+
+    if [ "$w" -lt 524288 ] || [ "$i" -lt 512 ]; then
+        echo "  ! inotify limits are low — Tilt will silently stop noticing file changes"
+        MANUAL+=("Raise inotify limits (needs root on the host):
+      echo -e 'fs.inotify.max_user_watches=524288\\nfs.inotify.max_user_instances=512' \\
+        | sudo tee /etc/sysctl.d/99-rig.conf
+      sudo sysctl --system")
+    fi
+}
+
+# ── fetch ──────────────────────────────────────────────────────────────────
+
+# Resolve where a given artifact comes from, honouring DEPS_SOURCE.
+resolve_url() {
+    local upstream="$1"
+    case "$DEPS_SOURCE" in
+        upstream)    echo "$upstream" ;;
+        artifactory)
+            if [ -z "$DEPS_ARTIFACTORY_URL" ]; then
+                echo "DEPS_SOURCE=artifactory but DEPS_ARTIFACTORY_URL is empty" >&2
+                exit 1
+            fi
+            echo "${DEPS_ARTIFACTORY_URL%/}/$(basename "$upstream")"
+            ;;
+        *) echo "unsupported DEPS_SOURCE '$DEPS_SOURCE' for a download" >&2; exit 1 ;;
+    esac
+}
+
+verify() {
+    local file="$1" want="$2" name="$3" got
+    got=$(sha256sum "$file" | awk '{print $1}')
+    if [ "$got" != "$want" ]; then
+        echo "checksum mismatch for $name" >&2
+        echo "  expected $want" >&2
+        echo "  got      $got" >&2
+        exit 1
+    fi
+}
+
+# fetch_bin <name> <url> <sha256> <dest-dir>  — a bare binary
+fetch_bin() {
+    local name="$1" url="$2" sha="$3" dest="$4"
+    local tmp="$dest/.$name.tmp"
+    echo "  fetching $name"
+    curl -fsSL --retry 3 -o "$tmp" "$(resolve_url "$url")"
+    verify "$tmp" "$sha" "$name"
+    mv "$tmp" "$dest/$name"
+    chmod +x "$dest/$name"
+}
+
+# fetch_tgz <name> <url> <sha256> <dest-dir> <path-inside-archive> <strip>
+# Archive layouts differ — tilt's is flat (the binary at the root, strip=0),
+# others nest it a directory down — so the caller says which.
+fetch_tgz() {
+    local name="$1" url="$2" sha="$3" dest="$4" inner="$5" strip="$6"
+    local tmp="$dest/.$name.tgz"
+    echo "  fetching $name"
+    curl -fsSL --retry 3 -o "$tmp" "$(resolve_url "$url")"
+    verify "$tmp" "$sha" "$name"
+    # --no-same-owner: extracting as root would otherwise restore the uid/gid
+    # baked into the archive (some ship as uid 1001), leaving a binary the host
+    # user does not own.
+    tar -xzf "$tmp" -C "$dest" --strip-components="$strip" --no-same-owner "$inner"
+    rm -f "$tmp"
+    chmod +x "$dest/$name"
+}
+
+# The installer runs as root so it can reach the docker socket, which means
+# everything it writes into a mounted volume lands root-owned and unusable from
+# the host. Hand it back to whoever owns the mount point (the host user created
+# that directory before mounting it).
+fix_ownership() {
+    local dir="$1"
+    [ -d "$dir" ] || return 0
+    local owner="${HOST_UID:-}:${HOST_GID:-}"
+    if [ "$owner" = ":" ]; then
+        owner=$(stat -c '%u:%g' "$dir")
+    fi
+    [ "$owner" = "0:0" ] && return 0
+    chown -R "$owner" "$dir" 2>/dev/null || true
+}
+
+# Two tiers, because not every machine should get cluster tooling.
+#
+#   core  kubectl, jq — talk to a cluster someone else runs. Nothing that
+#         creates one. Appropriate on a managed or corporate-issued machine
+#         where development tools are not wanted by default.
+#   dev   core plus kind and tilt — build clusters and hot-reload into them.
+#
+# The split exists because "install the toolchain" is not one decision: on a
+# managed workspace the right answer is kubectl and nothing else.
+CORE_TOOLS="kubectl jq"
+# No helm: every addon installs with `kubectl apply -f <url>`, so nothing here
+# has ever invoked it. Add it back the day something actually needs a chart.
+#
+# ctlptl is 'dev' rather than 'core' for the same reason kind is: core is "talk
+# to a cluster someone else runs", and ctlptl builds them. It earns its place
+# because it is what wires a cluster to a local registry — without one, an
+# unqualified image name resolves to docker.io/library/<name> and there is
+# nothing structural stopping a push there.
+#
+# docker-compose is 'dev' for the same reason, and is here because the distro
+# docker packages ship the daemon and CLI but frequently not the compose
+# plugin — so `docker compose up` fails with "unknown command" on an otherwise
+# working Docker, and nothing about that message names the missing piece.
+DEV_TOOLS="kind tilt ctlptl docker-compose"
+
+# ── what is already on this machine ───────────────────────────────────────
+#
+# A tool already on PATH at its pinned version is left where it is. Without
+# this, install downloads a second copy into OUT_BIN and then reports the first
+# one as shadowed — noise, and wrong, when both are the same version. That is
+# the normal state of any machine someone set up by hand: the AWS Workspace
+# keeps its toolchain in ~/wdir/bin, all five at exactly these pins.
+
+pin_of() {
+    case "$1" in
+        kubectl) echo "$KUBECTL_VERSION" ;;
+        jq)      echo "$JQ_VERSION" ;;
+        kind)    echo "$KIND_VERSION" ;;
+        tilt)    echo "$TILT_VERSION" ;;
+        ctlptl)  echo "$CTLPTL_VERSION" ;;
+        docker-compose) echo "$COMPOSE_VERSION" ;;
+    esac
+}
+
+# The version string a binary reports. Each tool spells the question
+# differently, and kubectl has to be told --client or it goes looking for a
+# server to ask.
+reported_version() {
+    local tool="$1" path="$2"
+    case "$tool" in
+        kubectl) "$path" version --client 2>/dev/null ;;
+        jq)      "$path" --version 2>/dev/null ;;
+        *)       "$path" version 2>/dev/null ;;
+    esac
+}
+
+# Does the binary at PATH report PIN? Matched as a whole version token, so
+# 0.37.6 never matches 10.37.60, with the leading v optional either side: kind
+# says v0.32.0, jq says jq-1.8.2, and tilt says v0.37.6 against a pin of 0.37.6.
+#
+# Bash's own regex rather than grep, deliberately. grep is not the same program
+# on every machine — some builds reject patterns that others accept — and a
+# failed grep inside a count reads exactly like a zero.
+version_matches() {
+    local tool="$1" path="$2" pin="$3" out v re
+    out=$(reported_version "$tool" "$path") || return 1
+    v="${pin#v}"
+    v="${v//./\\.}"
+    re="(^|[^0-9.])v?${v}([^0-9.]|\$)"
+    [[ $out =~ $re ]]
+}
+
+# DEPS_ONLY narrows a fetch to the tools it names. Unset means the whole tier,
+# which is what an explicit `deps.sh fetch` always gets: "download these into
+# DIR" must not quietly skip something because this machine happens to have it.
+# Only install() sets it, to what detect_toolchain found missing or mismatched.
+want() { [ -z "${DEPS_ONLY:-}" ] || [[ " $DEPS_ONLY " == *" $1 "* ]]; }
+
+# Every tool in the tier with its state, probed once and reported once. What
+# still needs fetching is left in TOOLCHAIN_NEED for install() to act on.
+TOOLCHAIN_NEED=""
+detect_toolchain() {
+    local tier="${TIER:-dev}" b pin path found
+    TOOLCHAIN_NEED=""
+    echo
+    echo "toolchain  (pinned, tier '$tier')"
+    for b in $(tier_tools "$tier"); do
+        pin=$(pin_of "$b")
+        path=$(command -v "$b" 2>/dev/null || true)
+        # compose is the one tool that is normally NOT a binary on PATH. It is a
+        # docker CLI plugin, so a machine where `docker compose` works perfectly
+        # has no `docker-compose` to find — and probing only PATH would report it
+        # missing and re-download a copy that is already there. That is the exact
+        # noise the version-aware skip exists to prevent, so ask docker instead.
+        if [ "$b" = docker-compose ] && [ -z "$path" ]; then
+            if found=$(docker compose version --short 2>/dev/null) && [ -n "$found" ]; then
+                if [ "${found#v}" = "${pin#v}" ]; then
+                    printf "    %-8s %-9s %s\n" "$b" "$pin" "docker cli plugin"
+                else
+                    printf "  ! %-8s wants %s, the docker cli plugin reports '%s'\n" \
+                        "$b" "$pin" "$found"
+                    TOOLCHAIN_NEED+="$b "
+                fi
+                continue
+            fi
+        fi
+        if [ -z "$path" ]; then
+            printf "  - %-8s %-9s not found\n" "$b" "$pin"
+            TOOLCHAIN_NEED+="$b "
+        elif version_matches "$b" "$path" "$pin"; then
+            printf "    %-8s %-9s %s\n" "$b" "$pin" "$path"
+        else
+            found=$(reported_version "$b" "$path" 2>/dev/null | head -1 || true)
+            printf "  ! %-8s wants %s, %s reports '%s'\n" "$b" "$pin" "$path" "$found"
+            TOOLCHAIN_NEED+="$b "
+        fi
+    done
+    if [ -z "$TOOLCHAIN_NEED" ]; then
+        echo "  every pinned tool is already on PATH — nothing to fetch"
+    else
+        echo "  'make deps' fetches only: ${TOOLCHAIN_NEED% }"
+    fi
+}
+
+fetch() {
+    local dest="$OUT_BIN" tier="${TIER:-dev}"
+    while [ $# -gt 0 ]; do
+        case "$1" in
+            --to)      dest="$2"; shift 2 ;;
+            core|dev)  tier="$1"; shift ;;
+            *)         echo "unknown argument: $1" >&2; exit 1 ;;
+        esac
+    done
+    dest="$(abspath "$dest")"
+    mkdir -p "$dest"
+    TIER="$tier"
+
+    if [ "$DEPS_SOURCE" = "baked" ]; then
+        echo "installing baked binaries from $BAKED_BIN"
+        cp -a "$BAKED_BIN"/. "$dest"/
+        fix_ownership "$dest"
+        return
+    fi
+
+    if [ -n "${DEPS_ONLY:-}" ]; then
+        echo "fetching ${DEPS_ONLY% } (source: $DEPS_SOURCE)"
+    else
+        echo "fetching '$tier' toolchain (source: $DEPS_SOURCE)"
+    fi
+    if want kubectl; then fetch_bin kubectl "$KUBECTL_URL" "$KUBECTL_SHA256" "$dest"; fi
+    if want jq;      then fetch_bin jq      "$JQ_URL"      "$JQ_SHA256"      "$dest"; fi
+    if [ "$tier" = "dev" ]; then
+        if want kind;   then fetch_bin kind   "$KIND_URL"   "$KIND_SHA256"   "$dest"; fi
+        if want tilt;   then fetch_tgz tilt   "$TILT_URL"   "$TILT_SHA256"   "$dest" tilt   0; fi
+        if want ctlptl; then fetch_tgz ctlptl "$CTLPTL_URL" "$CTLPTL_SHA256" "$dest" ctlptl 0; fi
+        if want docker-compose; then
+            fetch_bin docker-compose "$COMPOSE_URL" "$COMPOSE_SHA256" "$dest"
+        fi
+    fi
+
+    fix_ownership "$dest"
+    # kind writes the kubeconfig as root too; hand that back as well when it's
+    # a mounted host directory rather than container-local state.
+    fix_ownership "${KUBE_DIR:-/out/kube}"
+}
+
+# ── install ────────────────────────────────────────────────────────────────
+
+report_manual() {
+    echo
+    if [ ${#MANUAL[@]} -eq 0 ]; then
+        echo "nothing left to do by hand."
+        return
+    fi
+    echo "host actions this cannot perform (${#MANUAL[@]}):"
+    echo
+    local n=1
+    for m in "${MANUAL[@]}"; do
+        echo "  $n. $m"
+        echo
+        n=$((n + 1))
+    done
+}
+
+# Installing into a directory that sits early in PATH silently replaces whatever
+# the machine was already using — which on a shared or client machine can break
+# unrelated work (kubectl more than one minor away from a cluster is the common
+# one). Say so; never decide it for them.
+tier_tools() { [ "$1" = "core" ] && echo "$CORE_TOOLS" || echo "$CORE_TOOLS $DEV_TOOLS"; }
+
+warn_shadowing() {
+    local b existing shadowed="" tier="${1:-dev}"
+    for b in $(tier_tools "$tier"); do
+        [ -x "$OUT_BIN/$b" ] || continue
+        # Where would this resolve if OUT_BIN weren't in the way?
+        existing=$(PATH=$(echo "$PATH" | tr ':' '\n' | grep -vx "$OUT_BIN" | paste -sd:) \
+                   command -v "$b" 2>/dev/null || true)
+        [ -n "$existing" ] || continue
+        [ "$existing" = "$OUT_BIN/$b" ] && continue
+        # The same version in both places is not a conflict: nothing changes for
+        # any other project whichever copy PATH happens to find first.
+        if version_matches "$b" "$existing" "$(pin_of "$b")"; then continue; fi
+        shadowed+="      $b  $existing"$'\n'
+    done
+
+    [ -n "$shadowed" ] || return 0
+
+    case ":${PATH}:" in
+        *":$OUT_BIN:"*) ;;
+        *) return 0 ;;   # not on PATH yet, so nothing is being shadowed
+    esac
+
+    echo
+    echo "  ! these were already installed elsewhere and are now shadowed by $OUT_BIN:"
+    printf '%s' "$shadowed"
+    echo "    Other projects on this machine will pick up the new versions."
+    MANUAL+=("Decide which toolchain wins. To keep the previous one, remove what
+  was just installed:
+      rm -f $(for b in $(tier_tools "$tier"); do printf '%s ' "$OUT_BIN/$b"; done)
+  Or install somewhere private instead:
+      OUT_BIN=\$PWD/def/bin make deps   # then put that dir first in PATH")
+}
+
+# A copy in OUT_BIN only gives you `docker-compose`. That hyphenated form is the
+# retired v1 spelling; every compose file written in the last few years assumes
+# `docker compose`, which resolves plugins BY NAME out of a plugin directory.
+# So the binary is fetched like any other and then linked, in your own home —
+# no root, and nothing outside it.
+install_compose_plugin() {
+    local src="$OUT_BIN/docker-compose" dir="$HOME/.docker/cli-plugins"
+    [ -x "$src" ] || return 0
+    mkdir -p "$dir"
+    # Something else already owns that name — docker-desktop and some distro
+    # packages install a real file there. Overwriting it would take the plugin
+    # away from whatever put it there, so say so and let the user decide.
+    if [ -e "$dir/docker-compose" ] && [ ! -L "$dir/docker-compose" ]; then
+        MANUAL+=("Something already installs the compose plugin at
+      $dir/docker-compose
+  To use rig's pinned build instead:
+      ln -sf $src $dir/docker-compose")
+        return 0
+    fi
+    ln -sfn "$src" "$dir/docker-compose"
+    echo "  compose plugin -> $dir/docker-compose"
+    return 0
+}
+
+install() {
+    local tier="${1:-dev}" b
+    TIER="$tier"
+    detect
+
+    # detect_toolchain has already probed PATH. Fetch only what it found missing
+    # or at the wrong version; a tool already present at its pin stays where it is.
+    if [ -n "$TOOLCHAIN_NEED" ]; then
+        echo
+        DEPS_ONLY="$TOOLCHAIN_NEED" fetch "$tier"
+        echo
+        echo "installed to $OUT_BIN ($tier):"
+        for b in $TOOLCHAIN_NEED; do
+            if [ -x "$OUT_BIN/$b" ]; then echo "  $b"; fi
+        done
+        if [ "$tier" = "core" ]; then
+            echo "  (no kind/tilt — 'make deps dev' adds them)"
+        fi
+        # Only when compose was one of the things fetched: linking a binary
+        # that is already satisfied elsewhere on PATH would point the plugin at
+        # a copy rig did not install.
+        case " $TOOLCHAIN_NEED " in
+            *" docker-compose "*) install_compose_plugin ;;
+        esac
+
+        # Only worth saying when something actually landed in OUT_BIN. When every
+        # tool was satisfied elsewhere, OUT_BIN may reasonably be off PATH, and
+        # telling the user to add it would be advice to fix nothing.
+        case ":${PATH}:" in
+            *":$OUT_BIN:"*) ;;
+            *) MANUAL+=("Put the toolchain on your PATH — add to ~/.bashrc:
+      export PATH=\"${OUT_BIN}:\$PATH\"") ;;
+        esac
+    fi
+    warn_shadowing "$tier"
+
+    report_manual
+}
+
+# ── main ───────────────────────────────────────────────────────────────────
+
+require_linux
+
+case "${1:-install}" in
+    detect)  detect; report_manual ;;
+    fetch)   shift; fetch "$@" ;;
+    install) shift; install "${1:-dev}" ;;
+    *)       echo "usage: $0 [detect|fetch|install]" >&2; exit 1 ;;
+esac

rig/ctrl/env.d/client.env

@@ -19,7 +19,7 @@ DNS_MODE=hosts
 # Opt in to the real ports below only when this is the ONLY environment and
 # nothing else owns :80. They fail to bind otherwise, and docker reports it as an
 # opaque "failed to bind host port 0.0.0.0:80/tcp: address already in use"
-# halfway through cluster creation. `make station` checks before you spend the
+# halfway through cluster creation. `make check` checks before you spend the
 # time. Uncommenting also means only one environment can exist at a time.
 # HTTP_PORT=80
 # HTTPS_PORT=443

rig/ctrl/env.d/data.env

@@ -1,11 +1,10 @@
-# data — a cluster with the dependency containers a soleprint room asks for.
+# data — a cluster with the cabinets an environment asks for.
 #
-# The point of this profile is that a room declares what it needs once, in
-# cfg/<room>/data/cabinets.json, and gets it on either target: `build.py`
-# composes those services into docker-compose.yml for a laptop, and the addons
-# below install the same ones here. The names match deliberately —
-# soleprint/station/cabinets/<name>/cabinet.json carries a `rig_addon` field
-# pointing at ctrl/addons/<name>.sh.
+# A cabinet is a public service dropped in as-is — the upstream image,
+# unmodified, reachable at a known address. It is declared once and installs on
+# either target: a `service.yml` composes it for a laptop, and the addons below
+# install the same one here. The names match deliberately — each cabinet.json
+# carries a `rig_addon` field pointing at ctrl/addons/<name>.sh.
 #
 # Everything lands in the `data` namespace (DATA_NAMESPACE to move it), so
 # `make cluster reset` on the app namespace leaves the databases alone.
@@ -30,8 +29,8 @@ DATA_NAMESPACE=data
 # Postgres identity. The password is not here: postgres.sh generates one on
 # first install and keeps it across re-runs, so re-running the addon never
 # rotates the credential out from under whatever is already connected.
-POSTGRES_DB=soleprint
-POSTGRES_USER=soleprint
+POSTGRES_DB=app
+POSTGRES_USER=app
 POSTGRES_STORAGE=2Gi
 
 AIRFLOW_ADMIN_USER=admin

rig/ctrl/env.d/offline.env

@@ -1,5 +1,5 @@
 # offline — air-gapped. Everything comes from a local registry that was loaded
-# ahead of time; nothing reaches the internet. Pair with the wizard-full image
+# ahead of time; nothing reaches the internet. Pair with the deps-full image
 # (DEPS_SOURCE=baked) so the toolchain install is offline too.
 #
 # The heavier addons are left out to keep first boot viable.

rig/ctrl/k8s/README.md

@@ -1,8 +1,7 @@
 # `ctrl/k8s` — cluster shape, and what runs on it
 
-Same layout as every other project here (`unt`, `nvi`, `eth`, `mpr`, and
-soleprint's generated rooms): a kind config, a kustomize `base/`, and an
-`overlays/dev/` that patches it. See ALL `projects/templates/conventions.md`.
+Same layout as every other project here: a kind config, a kustomize `base/`,
+and an `overlays/dev/` that patches it.
 
 ```
 kind-config*.yaml.tpl   the cluster itself — nodes, ports, audit

rig/ctrl/k8s/kind-config.audit.yaml.tpl

@@ -44,7 +44,7 @@ nodes:
   - role: control-plane
     image: ${NODE_IMAGE}
     # hostPath is resolved by the HOST dockerd, so this must be a host path even
-    # when cluster.sh runs inside the wizard container. HOST_WORKDIR says where
+    # when cluster.sh runs inside the installer container. HOST_WORKDIR says where
     # this rig lives on the host; bare on a host it is just the repo root.
     extraMounts:
       - hostPath: ${HOST_WORKDIR}/ctrl/k8s/audit-policy.yaml

rig/ctrl/lib/config.sh

@@ -21,9 +21,14 @@
 # NODES and AUDIT are deliberately NOT here: they are properties of the chosen
 # ctrl/k8s/kind-config*.yaml.tpl and are read back out of it below, so there is
 # one place that decides the shape of the cluster rather than two that can drift.
+#
+# REGISTRY_PORT and MANIFESTS_DIR were missing here while ctrl/.env set them, so
+# the caller's env silently LOST to the file for those two — the one precedence
+# rule this header states. Both are now listed; the other twelve are unchanged.
 CONFIG_OVERRIDABLE="PROFILE CLUSTER K8S_VERSION KIND_CONFIG ADDONS
                     REGISTRY_MODE INGRESS_MODE DNS_MODE TILT_PORT
-                    SOURCE ARCH DEPS_SOURCE HTTP_PORT HTTPS_PORT"
+                    SOURCE ARCH DEPS_SOURCE HTTP_PORT HTTPS_PORT
+                    REGISTRY_PORT MANIFESTS_DIR"
 
 # The containing folder's name, reduced to something kind accepts as a cluster
 # name (a DNS label: lowercase alphanumerics and dashes). Run from ctrl/, so the
@@ -93,6 +98,12 @@ load_config() {
     TILT_PORT="${TILT_PORT:-$((base + 2))}"
     REGISTRY_PORT="${REGISTRY_PORT:-$((base + 3))}"
 
+    # Where the workload's manifests live, repo-root relative. Defaulted here so
+    # it is always resolved rather than sometimes-set: it is the seam that lets
+    # the real manifests be versioned away from the installer, and a consumer
+    # should not have to know whether anyone filled it in. See k8s/README.md.
+    MANIFESTS_DIR="${MANIFESTS_DIR:-ctrl/k8s/overlays/dev}"
+
     # Profiles name a k8s minor (v1_36); versions.env holds the pinned digest.
     local var="NODE_IMAGE_${K8S_VERSION}"
     NODE_IMAGE="${!var:-}"
@@ -103,16 +114,26 @@ load_config() {
 
     # The cluster's shape is a file in ctrl/k8s/, named by the profile. Adding a
     # shape is adding a file; there is no dispatcher to edit.
+    #
+    # A host that needs its own shape — extra port mappings, more nodes — passes
+    # an absolute path instead, and rig renders it exactly like one of its own:
+    # ${CLUSTER} and ${NODE_IMAGE} are substituted either way. The shape stays in
+    # the host's tree, because what a host's cluster needs is the host's business;
+    # rig only knows how to build whatever it is handed.
     KIND_CONFIG="${KIND_CONFIG:-kind-config.yaml.tpl}"
-    KIND_CONFIG_PATH="./k8s/${KIND_CONFIG}"
+    case "$KIND_CONFIG" in
+        /*) KIND_CONFIG_PATH="$KIND_CONFIG";          KIND_CONFIG_SHOWN="$KIND_CONFIG" ;;
+        *)  KIND_CONFIG_PATH="./k8s/${KIND_CONFIG}";  KIND_CONFIG_SHOWN="ctrl/k8s/${KIND_CONFIG}" ;;
+    esac
     if [ ! -f "$KIND_CONFIG_PATH" ]; then
-        echo "no such cluster shape: ctrl/k8s/${KIND_CONFIG}" >&2
-        echo "available: $(ls k8s/kind-config*.yaml.tpl 2>/dev/null | xargs -n1 basename | tr '\n' ' ')" >&2
+        echo "no such cluster shape: ${KIND_CONFIG_SHOWN}" >&2
+        echo "rig's own: $(ls k8s/kind-config*.yaml.tpl 2>/dev/null | xargs -n1 basename | tr '\n' ' ')" >&2
+        echo "or pass an absolute path to a shape of your own" >&2
         exit 1
     fi
 
     # Read the shape back out of the YAML rather than trusting a profile to
-    # restate it. station.sh sizes the memory warning on NODES, and cluster.sh
+    # restate it. check.sh sizes the memory warning on NODES, and cluster.sh
     # prints AUDIT before spending minutes building something that cannot be
     # changed afterwards — both would mislead if the numbers drifted.
     NODES=$(grep -c '^  - role:' "$KIND_CONFIG_PATH")
@@ -125,7 +146,7 @@ load_config() {
 # depending on something the caller does not set.
 #
 # hostPath entries are resolved by the HOST dockerd, so HOST_WORKDIR must stay a
-# host path even when this runs inside the wizard container.
+# host path even when this runs inside the installer container.
 render_kind_config() {
     local host_workdir="${HOST_WORKDIR:-$(cd .. && pwd)}"
     sed -e "s|\${CLUSTER}|${CLUSTER}|g" \

rig/ctrl/mem.sh

@@ -0,0 +1,233 @@
+#!/usr/bin/env bash
+# What memory this machine has, what is left, and — where there is one — what
+# cap is holding it there.
+#
+# Runs on native Linux and under WSL, because rig is developed on one and used
+# on the other. The difference is not cosmetic: on WSL the memory you see is a
+# VM allocation that can be raised, and the commonest failure is raising it
+# without restarting, so the number on disk and the number in /proc disagree.
+# On native Linux there is no such cap and pretending otherwise sends you to a
+# file that does not exist.
+#
+# This reports and instructs. It never writes a .wslconfig — applying one costs
+# a full VM restart that takes every shell, mount and container with it, and
+# choosing that moment is yours.
+#
+# `backup` exists so `restore` has something to read: back up, hand-edit
+# following the printed instruction, restore if it goes wrong. Both are
+# WSL-only, because .wslconfig is the only thing here worth backing up.
+#
+# Usage: mem.sh status | backup | restore
+set -euo pipefail
+
+# Windows outside WSL — Git Bash, MSYS, Cygwin — looks close enough to work and
+# then fails in a pile of confusing ways: no /proc, no docker socket, none of
+# the tooling. Detectable, so name it instead.
+require_linux() {
+    case "$(uname -s)" in
+        MINGW*|MSYS*|CYGWIN*)
+            cat >&2 <<'EOF'
+This has to run inside WSL, not Git Bash / MSYS / Cygwin.
+
+If WSL is not installed yet, from an elevated PowerShell or Command Prompt:
+
+    wsl --install
+
+That enables Windows features and needs a reboot, so it is not something this
+script will do for you. Afterwards, open the Linux shell it installs and run
+this from there.
+
+See "Starting from plain Windows" in README.md.
+EOF
+            exit 1 ;;
+    esac
+}
+
+is_wsl() { grep -qi microsoft /proc/version 2>/dev/null; }
+
+require_wsl() {
+    if ! is_wsl; then
+        echo "$1 acts on .wslconfig, which only exists under WSL." >&2
+        echo "This is native Linux — there is no VM allocation to save or roll back." >&2
+        echo "Use 'mem.sh status' to see what the machine actually has." >&2
+        exit 1
+    fi
+}
+
+mb() { echo $(( $(awk "/^$1:/{print \$2}" /proc/meminfo) / 1024 )); }
+
+# /mnt/c/Users can hold several real accounts — a renamed login leaves the old
+# directory behind — so picking the first alphabetically is a coin toss. Ask
+# Windows, then fall back to whichever profile actually owns a config.
+wslconfig_path() {
+    local profile winpath found
+    profile=$(cmd.exe /c "echo %USERPROFILE%" 2>/dev/null | tr -d "\r\n" || true)
+    case "$profile" in
+        ""|*%*) ;;
+        *) winpath=$(wslpath -u "$profile" 2>/dev/null || true)
+           if [ -n "$winpath" ] && [ -d "$winpath" ]; then
+               echo "$winpath/.wslconfig"; return
+           fi ;;
+    esac
+
+    found=$(ls -d /mnt/c/Users/*/.wslconfig 2>/dev/null | head -1 || true)
+    if [ -n "$found" ]; then echo "$found"; return; fi
+
+    echo "cannot tell which Windows profile owns .wslconfig. Candidates:" >&2
+    ls -d /mnt/c/Users/*/ 2>/dev/null \
+      | grep -viE "/(All Users|Default|Default User|Public)/$" | sed "s/^/  /" >&2
+    exit 1
+}
+
+configured_memory() {
+    [ -r "$1" ] || { echo ""; return; }
+    sed -n 's/^[[:space:]]*memory[[:space:]]*=[[:space:]]*//p' "$1" | tail -1 | tr -d '[:space:]'
+}
+
+# "9GB" / "8192MB" / "9G" -> MB, so it can be compared with /proc/meminfo.
+to_mb() {
+    local v="${1^^}" n
+    n=$(echo "$v" | tr -dc '0-9')
+    [ -n "$n" ] || { echo ""; return; }
+    case "$v" in
+        *GB|*G) echo $(( n * 1024 )) ;;
+        *MB|*M) echo "$n" ;;
+        *)      echo $(( n / 1024 / 1024 )) ;;
+    esac
+}
+
+hogs() {
+    echo "holding the most:"
+    ps -eo rss,comm --sort=-rss 2>/dev/null | awk 'NR>1 && NR<=6 {printf "  %6.0f MB  %s\n", $1/1024, $2}'
+}
+
+status() {
+    local total avail swap_total swap_free
+    total=$(mb MemTotal); avail=$(mb MemAvailable)
+    swap_total=$(mb SwapTotal); swap_free=$(mb SwapFree)
+
+    if is_wsl; then
+        local cfg conf conf_mb
+        cfg=$(wslconfig_path)
+        conf=$(configured_memory "$cfg")
+        echo "platform   WSL"
+        echo "config     $cfg"
+        if [ -n "$conf" ]; then
+            conf_mb=$(to_mb "$conf")
+            echo "configured $conf  (${conf_mb} MB)"
+        else
+            conf_mb=""
+            echo "configured (no memory= set — WSL defaults to 50% of host RAM, or 8GB, whichever is less)"
+        fi
+        echo "booted     ${total} MB"
+        echo "available  ${avail} MB"
+        echo "swap       ${swap_total} MB  ($(( swap_total - swap_free )) MB used)"
+
+        if [ -n "$conf_mb" ]; then
+            # The VM reports a little less than allocated; 15% covers the kernel
+            # without calling every healthy machine a mismatch.
+            if [ "$total" -lt $(( conf_mb * 85 / 100 )) ]; then
+                echo
+                echo "! configured ${conf_mb} MB but booted ${total} MB."
+                echo "  The change has not been applied. From a WINDOWS terminal:"
+                echo
+                echo "      wsl --shutdown"
+                echo
+                echo "  then start the distro again."
+            fi
+        else
+            echo
+            echo "To raise it, add to $cfg on the Windows side:"
+            echo
+            echo "    [wsl2]"
+            echo "    memory=8GB"
+            echo
+            echo "then, from a WINDOWS terminal:  wsl --shutdown"
+        fi
+
+        local n
+        n=$(ls "$cfg".*.bak 2>/dev/null | wc -l)
+        [ "$n" -gt 0 ] && echo "backups    $n  (newest: $(ls -t "$cfg".*.bak 2>/dev/null | head -1))"
+    else
+        echo "platform   native linux"
+        echo "total      ${total} MB"
+        echo "available  ${avail} MB"
+        echo "swap       ${swap_total} MB  ($(( swap_total - swap_free )) MB used)"
+        echo
+        echo "No VM allocation to raise here — this is the machine's own memory."
+        echo "If it is tight the levers are freeing something or adding swap."
+    fi
+
+    # Under a fifth left is worth naming wherever you are running.
+    if [ "$avail" -lt $(( total / 5 )) ]; then
+        echo
+        hogs
+    fi
+    return 0
+}
+
+backup() {
+    require_wsl backup
+    local cfg dest
+    cfg=$(wslconfig_path)
+    [ -r "$cfg" ] || { echo "nothing to back up: $cfg does not exist" >&2; exit 1; }
+    # Timestamped and never overwritten: a backup that can destroy itself on a
+    # second run is not a backup.
+    dest="${cfg}.$(date +%Y%m%d-%H%M%S).bak"
+    cp "$cfg" "$dest"
+    echo "backed up  $dest"
+    echo
+    echo "Edit $cfg by hand, then from a WINDOWS terminal:  wsl --shutdown"
+}
+
+restore() {
+    require_wsl restore
+    local cfg newest count
+    cfg=$(wslconfig_path)
+    newest=$(ls -t "$cfg".*.bak 2>/dev/null | head -1 || true)
+    [ -n "$newest" ] || { echo "no backups found beside $cfg" >&2; exit 1; }
+
+    echo "restoring  $newest"
+    echo "        -> $cfg"
+    echo
+
+    # Newest is the right default — undo the last edit — but if you backed up
+    # *after* editing, the state you want is older. Show the rest so a no-op
+    # restore is obviously a no-op rather than a mystery.
+    count=$(ls "$cfg".*.bak 2>/dev/null | wc -l)
+    if [ "$count" -gt 1 ]; then
+        echo "$count backups exist, newest first:"
+        ls -t "$cfg".*.bak | sed 's/^/  /'
+        echo "  (restoring the newest; copy another by hand to pick an older one)"
+        echo
+    fi
+
+    if [ -r "$cfg" ]; then
+        echo "what changes:"
+        if diff "$cfg" "$newest" > /tmp/mem.diff 2>&1 && [ ! -s /tmp/mem.diff ]; then
+            echo "  nothing — that backup is identical to the current config"
+        else
+            sed 's/^/  /' /tmp/mem.diff
+        fi
+        rm -f /tmp/mem.diff
+        echo
+    fi
+
+    printf "proceed? [y/N] "
+    read -r reply
+    case "$reply" in
+        y|Y|yes|Yes) ;;
+        *) echo "left alone"; return 0 ;;
+    esac
+    cp "$newest" "$cfg"
+    echo "restored. From a WINDOWS terminal:  wsl --shutdown"
+}
+
+require_linux
+
+case "${1:-status}" in
+    status)  status ;;
+    backup)  backup ;;
+    restore) restore ;;
+    *)       echo "usage: $0 [status|backup|restore]" >&2; exit 1 ;;
+esac

rig/ctrl/newbox.sh

@@ -58,10 +58,13 @@ require_wsl() {
         cat >&2 <<'EOF'
 newbox is WSL-only for now.
 
+If WSL is not installed, run `wsl --install` from an elevated Windows prompt
+first — see "Starting from plain Windows" in README.md.
+
 On native Linux you do not need it: rig already isolates environments by
 directory (own cluster, context, images and port block), so a second copy in a
 second directory is the clean slate. To validate the installer itself against a
-bare system, run the wizard against a stock Debian container instead.
+bare system, run ctrl/deps.sh against a stock Debian container instead.
 EOF
         exit 1
     fi
@@ -250,7 +253,7 @@ create() {
     echo
     echo "next:"
     echo "  make newbox shell        # a shell inside it"
-    echo "  then: cd ~/rig && make station && make deps && make cluster up"
+    echo "  then: cd ~/rig && make check && make deps && make cluster up"
     echo
     echo "For a browser on Windows to resolve the hostnames, paste this into"
     echo "C:\\Windows\\System32\\drivers\\etc\\hosts (it has no wildcard support):"

rig/ctrl/pins.sh

@@ -0,0 +1,57 @@
+#!/usr/bin/env bash
+# Do the standalone scripts still install what rig pins?
+#
+# standalone/rigdeps.sh carries its toolchain pins inline, because it exists for
+# a machine that will never have ctrl/versions.env. That makes two copies of the
+# same versions and checksums, and two copies drift the day one is edited and
+# the other forgotten. This is the check that notices.
+#
+# ctrl/versions.env is the source of truth. Only the keys rigdeps.sh itself
+# defines are compared: versions.env also pins addon images (cert-manager,
+# metallb, metrics-server) that rigdeps.sh never installs, and demanding those
+# would make this fail forever for no reason.
+#
+# Exits non-zero on any mismatch — unlike the host checks, this one is a test.
+#
+# Usage: pins.sh
+set -euo pipefail
+cd "$(dirname "$0")"
+
+SOURCE=./versions.env
+COPY=../standalone/rigdeps.sh
+
+[ -r "$COPY" ] || { echo "no $COPY to compare" >&2; exit 1; }
+
+# KEY=value for the pin keys a file defines, quotes stripped. awk rather than a
+# grep regex, which is not the same program everywhere.
+pins() {
+    awk -F= '/^[A-Z_]+_(VERSION|SHA256)=/ {
+                 v = substr($0, index($0, "=") + 1); gsub(/^["\x27]|["\x27]$/, "", v)
+                 print $1 "=" v }' "$1"
+}
+
+echo "pins: standalone/rigdeps.sh against ctrl/versions.env"
+bad=0
+while IFS='=' read -r key copy_val; do
+    [ -n "$key" ] || continue
+    src_val=$(pins "$SOURCE" | sed -n "s/^${key}=//p" | head -1)
+    if [ -z "$src_val" ]; then
+        printf "  ! %-16s in rigdeps.sh but not in versions.env\n" "$key"
+        bad=1
+    elif [ "$src_val" = "$copy_val" ]; then
+        printf "    %-16s %s\n" "$key" "$( [ ${#src_val} -gt 20 ] && echo "${src_val:0:12}…" || echo "$src_val" )"
+    else
+        printf "  ! %-16s versions.env %s\n" "$key" "$src_val"
+        printf "    %-16s rigdeps.sh   %s\n" "" "$copy_val"
+        bad=1
+    fi
+done < <(pins "$COPY")
+
+echo
+if [ "$bad" -eq 0 ]; then
+    echo "in step — rigdeps.sh installs exactly what rig pins."
+else
+    echo "DRIFT. versions.env is the source of truth: copy the differing lines from it"
+    echo "into standalone/rigdeps.sh, taking checksums from the publisher's release list."
+    exit 1
+fi

rig/ctrl/ports.sh

@@ -19,7 +19,7 @@
 # written into ctrl/.env, so it becomes pinned, visible and editable rather than
 # a number that appears from nowhere. Anything already in ctrl/.env wins.
 #
-# Usage: ports.sh show | derive | persist
+# Usage: ports.sh show | active | derive | persist
 set -euo pipefail
 cd "$(dirname "$0")"
 
@@ -38,6 +38,33 @@ derive() {
     DERIVED_REGISTRY=$((base + 3))
 }
 
+# The resolved facts a consumer outside bash needs, machine-readable:
+#
+#   CLUSTER KUBECONTEXT HTTP HTTPS TILT REGISTRY MANIFESTS_DIR
+#
+# Identity and ports together, because they are one fact set — the header above
+# says so: both derive from the directory name so that copies never collide. A
+# consumer needs all of them or none, and fetching them separately is how two
+# end up disagreeing. MANIFESTS_DIR rides along because the one consumer that
+# needs the addressing is the one that needs to know what to deploy.
+#
+# Space-separated, so MANIFESTS_DIR must not contain spaces. Everything else in
+# rig already assumes that of paths — kind, docker and kubectl all do.
+#
+# `derive` answers a DIFFERENT question — what the directory name alone implies
+# — and deliberately ignores ctrl/.env. Configuring anything from it would
+# silently contradict this file's own rule that "anything already in ctrl/.env
+# wins". `active` is what anything downstream should read.
+#
+# Why this exists at all: the cluster name is not the bare directory name.
+# default_cluster_name() lowercases it and replaces every character outside
+# [a-z0-9-], because it has to be a DNS label. Re-deriving that in another
+# language is how a copy in `My_Project/` ends up guarding the wrong context.
+active() {
+    load_config
+    echo "$CLUSTER $KUBECONTEXT $HTTP_PORT $HTTPS_PORT $TILT_PORT $REGISTRY_PORT $MANIFESTS_DIR"
+}
+
 show() {
     derive
     echo "environment  $CLUSTER"
@@ -98,6 +125,7 @@ persist() {
 case "${1:-show}" in
     show)    show ;;
     derive)  derive; echo "$DERIVED_HTTP $DERIVED_HTTPS $DERIVED_TILT $DERIVED_REGISTRY" ;;
+    active)  active ;;
     persist) persist ;;
-    *)       echo "usage: $0 [show|derive|persist]" >&2; exit 1 ;;
+    *)       echo "usage: $0 [show|active|derive|persist]" >&2; exit 1 ;;
 esac

rig/ctrl/registry.sh

@@ -43,7 +43,7 @@ K="kubectl --context ${KUBECONTEXT}"
 #   2. every kind node's containerd — nodes do NOT inherit host trust
 #   3. anything doing HTTPS from inside the cluster, in its own trust store
 #
-# We handle (2) here because it's ours to handle. (1) is reported by station.sh
+# We handle (2) here because it's ours to handle. (1) is reported by check.sh
 # since it needs root. (3) belongs to the workload.
 install_ca_into_nodes() {
     [ -n "${REGISTRY_CA_FILE:-}" ] || return 0

rig/ctrl/selftest.sh

@@ -0,0 +1,193 @@
+#!/usr/bin/env bash
+# What rig has settled, written down as assertions.
+#
+# These are documentation that runs. Each check is ONE decision that has already
+# been made, with the reason above it — not coverage, and deliberately not an
+# exhaustive sweep of use cases. rig's own index says a rule without its reason
+# gets overridden the first time it is inconvenient; a rule nobody can restate
+# is worse. So the test says what was decided, and failing it should read as
+# "you are about to undo this" rather than "something broke".
+#
+# Scope, on purpose:
+#   - no cluster, no docker, no network. It must be cheap enough to actually run.
+#   - it asserts about RIG. `make check` asserts about the MACHINE and never
+#     fails; this exits 1, the way `make pins` does.
+#   - what actually deploys is not testable here. `tilt ci` stays a manual step.
+#
+# Usage: make selftest   (or: bash ctrl/selftest.sh)
+set -uo pipefail          # NOT -e: one failing check must not abort the rest
+cd "$(dirname "$0")"
+
+source ./lib/config.sh
+
+rc=0
+passed=0
+
+check() {   # name, expected, actual
+    if [ "$2" = "$3" ]; then
+        printf '  ok    %s\n' "$1"
+        passed=$((passed + 1))
+    else
+        printf '  FAIL  %s\n        expected: %s\n        got:      %s\n' "$1" "$2" "$3"
+        rc=1
+    fi
+}
+
+note() { printf '\n%s\n' "$1"; }
+
+# Resolve one key the way every rig script does, in a clean shell so the
+# caller's exported value is the only thing in play.
+resolved() {
+    bash -c 'source ./lib/config.sh; load_config >/dev/null 2>&1; printf "%s" "${!1}"' _ "$1"
+}
+
+
+note "the ports.sh active contract"
+# ports.sh active is read POSITIONALLY by two other files — the Makefile takes
+# $(word 2) and $(word 5), the Tiltfile takes _facts[0]..[6]. Insert a field in
+# the middle and nothing errors: Tilt simply guards on the wrong context or
+# binds the wrong port. The field count and order are the contract, so they are
+# pinned here rather than left to whoever edits ports.sh next.
+FACTS="$(bash ports.sh active)"
+check "active: exactly 7 fields"          "7" "$(printf '%s' "$FACTS" | wc -w)"
+read -r F_CLUSTER F_CTX F_HTTP F_HTTPS F_TILT F_REG F_MANIFESTS <<< "$FACTS"
+check "active: field 2 is kind-<cluster>" "kind-$F_CLUSTER" "$F_CTX"
+check "active: fields 3-6 are numeric"    "yes" \
+    "$([[ "$F_HTTP$F_HTTPS$F_TILT$F_REG" =~ ^[0-9]+$ ]] && echo yes || echo no)"
+check "active: field 7 is a path"         "yes" \
+    "$([ -n "$F_MANIFESTS" ] && [ "${F_MANIFESTS#-}" = "$F_MANIFESTS" ] && echo yes || echo no)"
+# derive answers a different question and must keep its own shape: it reports
+# what the directory name implies, ignoring ctrl/.env, so nothing should
+# configure itself from it.
+check "derive: still 4 fields, not 7"     "4" "$(bash ports.sh derive | wc -w)"
+
+
+note "the caller's env beats the files"
+# lib/config.sh states one precedence rule: versions.env < env.d/<profile> <
+# ctrl/.env < the caller's env. It is enforced by CONFIG_OVERRIDABLE, a
+# hand-maintained list — and a key missing from it loses to the file SILENTLY.
+# REGISTRY_PORT and MANIFESTS_DIR were both missing on 2026-09-13 and were found
+# by accident.
+#
+# So this loop is generated FROM the list: add a key to CONFIG_OVERRIDABLE and
+# this test starts asking about it without anyone remembering to come here.
+# Three keys name something that must exist and are validated at load, so they
+# get a real alternative rather than a sentinel.
+test_value() {
+    case "$1" in
+        PROFILE)        echo "client" ;;                       # env.d/client.env exists
+        K8S_VERSION)    echo "v1_35" ;;                        # NODE_IMAGE_v1_35 is pinned
+        KIND_CONFIG)    echo "kind-config.client.yaml.tpl" ;;  # the shape must exist
+        *_PORT)         echo "19999" ;;
+        CLUSTER)        echo "selftest-name" ;;
+        MANIFESTS_DIR)  echo "../elsewhere/overlays/dev" ;;
+        ADDONS)         echo "metallb" ;;
+        *)              echo "selftest-sentinel" ;;
+    esac
+}
+for key in $CONFIG_OVERRIDABLE; do
+    [ -n "$key" ] || continue
+    want="$(test_value "$key")"
+    if [ -z "$want" ]; then
+        check "precedence: $key has a test value" "yes" "no — add one to test_value()"
+        continue
+    fi
+    got="$(export "$key=$want"; resolved "$key")"
+    check "precedence: caller's $key wins" "$want" "$got"
+done
+
+
+note "one derivation, not three"
+# The Makefile used to compute the cluster name itself and sed TILT_PORT out of
+# ctrl/.env — a second derivation of values lib/config.sh already owns, which
+# could disagree with it after `make ports persist`. It now reads ports.sh
+# active. Nothing structurally prevents the sed coming back, so the agreement is
+# asserted against the real `make -n` output rather than against the source.
+# --no-print-directory and a grep, not `tail -1`: run from `make selftest` this
+# is a RECURSIVE make, and the "Entering/Leaving directory" lines go to STDOUT.
+# tail -1 then reads "make[1]: Leaving directory ..." and both checks below fail
+# — but only when invoked through make, never when the script is run directly.
+# A test that passes one way and fails the other is worse than no test.
+MK="$(cd .. && make --no-print-directory -n tilt 2>/dev/null | grep -m1 'tilt ')"
+check "Makefile: --context comes from active" "$F_CTX" \
+    "$(printf '%s' "$MK" | sed -n 's/.*--context \([^ ]*\).*/\1/p')"
+check "Makefile: --port comes from active"    "$F_TILT" \
+    "$(printf '%s' "$MK" | sed -n 's/.*--port \([^ ]*\).*/\1/p')"
+
+
+note "identity follows the folder, safely"
+# The cluster name is NOT the bare directory name: kind needs a DNS label, so
+# default_cluster_name lowercases it and replaces everything outside [a-z0-9-].
+# Re-deriving that anywhere else is how a copy ends up guarding the wrong
+# context — which is exactly why the Tiltfile asks instead of computing.
+TMP="$(mktemp -d)"
+trap 'rm -rf "$TMP"' EXIT
+mkdir -p "$TMP/My_Proj"
+cp -r . "$TMP/My_Proj/ctrl"
+# A pinned CLUSTER in .env would be an override, not a derivation, and this
+# check is about the derivation.
+sed -i '/^CLUSTER=/d' "$TMP/My_Proj/ctrl/.env" 2>/dev/null
+COPY="$(cd "$TMP/My_Proj/ctrl" && bash ports.sh active)"
+check "a dir named My_Proj derives a DNS label" "my-proj"      "$(awk '{print $1}' <<< "$COPY")"
+check "and a context to match"                  "kind-my-proj" "$(awk '{print $2}' <<< "$COPY")"
+check "a renamed copy gets a DIFFERENT block"   "different" \
+    "$([ "$(awk '{print $3}' <<< "$COPY")" != "$F_HTTP" ] && echo different || echo COLLIDES)"
+
+
+note "ports are stable across versions"
+# Not a change-detector. The block is derived, never stored, so if the
+# derivation shifts then every EXISTING environment's ports move underneath it —
+# a running cluster keeps its old ports while rig starts reporting new ones, and
+# `make ports` stops describing reality. Anchored to three known names.
+check "derive_port_base rig"     "20310" "$(derive_port_base rig)"
+check "derive_port_base foo"     "21690" "$(derive_port_base foo)"
+check "derive_port_base my-proj" "21030" "$(derive_port_base my-proj)"
+
+
+note "rig stays standalone"
+# rig sits inside a host project's tree but must be copyable straight out of it:
+# no imports, no paths, no assumption the host is there. This grep is the whole
+# test of that claim, and until now it lived only in prose and in whoever
+# remembered to run it.
+#
+# The pattern is assembled from fragments so this file does not match ITSELF.
+# Writing it literally would fail forever; excluding this file instead would put
+# a blind spot in the one check that guards the boundary.
+HOST_PAT="$(printf '%s' 'sole' 'print' '|\b' 'sp' 'r\b')"
+check "no host-project references" "0" \
+    "$(cd .. && grep -rIl -iE "$HOST_PAT" . --exclude-dir=def 2>/dev/null | wc -l)"
+
+
+note "the Tiltfile hardcodes nothing"
+# Every other Tiltfile on this machine writes its slug in five or six times by
+# hand, so a copied project deploys into the original's cluster until someone
+# edits all of them. rig's asks ports.sh. A literal kind-<name> here would mean
+# that has been undone.
+check "no literal kind-<name>"  "0" "$(grep -cE "['\"]kind-[a-z0-9]" Tiltfile)"
+check "guards on the variable"  "1" "$(grep -c 'allow_k8s_contexts(CTX)' Tiltfile)"
+check "asks ports.sh for facts" "1" "$(grep -c "local('bash ports.sh active'" Tiltfile)"
+
+
+note "optional — needs tilt and this rig's cluster"
+# Parsing the Tiltfile for real is the only way to know it still evaluates, but
+# Tilt snapshots a kubectl context before parsing, so it cannot run without a
+# cluster. Skipped rather than failed when there is none, the same way docgen
+# skips its graphgen section.
+if ! command -v tilt >/dev/null; then
+    printf '  skip  tilt is not installed\n'
+elif ! kubectl config get-contexts -o name 2>/dev/null | grep -qx "$F_CTX"; then
+    printf "  skip  no %s context — run 'make cluster up' to include this\n" "$F_CTX"
+else
+    out="$(tilt alpha tiltfile-result --context "$F_CTX" 2>&1)"
+    check "Tiltfile evaluates" "yes" \
+        "$(printf '%s' "$out" | grep -q '"Manifests"' && echo yes || echo "no: $(printf '%s' "$out" | tail -1)")"
+fi
+
+
+printf '\n'
+if [ "$rc" -eq 0 ]; then
+    printf '%d checks passed — rig still does what it says\n' "$passed"
+else
+    printf 'FAILED — a decision above has drifted; read the comment next to it\n' >&2
+fi
+exit "$rc"

rig/ctrl/setup.sh

@@ -73,11 +73,11 @@ record() {
 
 step_host() {
     local out
-    if ! out=$(bash ./wizard.sh detect 2>&1); then
+    if ! out=$(bash ./deps.sh detect 2>&1); then
         record host fail "detection failed"
         return
     fi
-    # Anything the wizard flagged with '!' needs a human; surface the count here
+    # Anything flagged with '!' needs a human; surface the count here
     # and the detail below rather than burying it.
     local warns; warns=$(echo "$out" | grep -c '^\s*!' || true)
     HOST_DETAIL="$out"
@@ -102,7 +102,7 @@ step_toolchain() {
         return
     fi
 
-    if bash ./wizard.sh install "$TIER" >/tmp/rig-deps.$$ 2>&1; then
+    if bash ./deps.sh install "$TIER" >/tmp/rig-deps.$$ 2>&1; then
         local still=""
         for b in $want; do
             [ -x "${OUT_BIN:-$HOME/.local/bin}/$b" ] || still="$still $b"

rig/ctrl/station.sh

@@ -1,106 +0,0 @@
-#!/usr/bin/env bash
-# Station check: is this workstation ready to run rig?
-#
-# Reports and instructs; never silently fixes anything. Everything it finds is
-# either already fine, or something a human has to decide on.
-#
-# Runs the wizard's host detection in a container when Docker is the only thing
-# installed, or directly when the toolchain is already present. Then adds the
-# checks that need this repo's config: profile sanity, CA trust, port clashes.
-set -euo pipefail
-cd "$(dirname "$0")"
-
-WIZARD_IMAGE="${WIZARD_IMAGE:-$(basename "$(cd .. && pwd)")-wizard}"
-
-# Host detection. Prefer running it bare — it needs no dependencies beyond
-# coreutils — and fall back to the container only if this shell can't.
-bash ./wizard.sh detect
-
-# ── repo-level checks ──────────────────────────────────────────────────────
-
-source ./lib/config.sh
-load_config
-
-echo
-echo "config"
-echo "  profile     ${PROFILE_NAME}  (nodes=${NODES} audit=${AUDIT})"
-echo "  cluster     ${CLUSTER}  (context ${KUBECONTEXT})"
-echo "  registry    ${REGISTRY_MODE}"
-echo "  ingress     ${INGRESS_MODE}"
-
-if [ ! -f ./.env ]; then
-    echo "  ! ctrl/.env missing — copy it:  cp ctrl/.env.example ctrl/.env"
-fi
-
-# A 3-node profile on a box that's already full is the most common first
-# failure, and it presents as pods stuck Pending rather than anything obvious.
-avail=$(awk '/^MemAvailable:/{printf "%d", $2/1024/1024}' /proc/meminfo)
-need=$((NODES * 2))
-if [ "$avail" -lt "$need" ]; then
-    echo "  ! profile '${PROFILE_NAME}' wants ~${need} GB, ${avail} GB available"
-    echo "    'make cluster list' shows what else is running; 'make cluster free' stops it"
-fi
-
-# The CA reaches three places and only one of them is ours. Report the other two.
-if [ -n "${REGISTRY_CA_FILE:-}" ]; then
-    echo
-    echo "registry CA"
-    if [ ! -r "$REGISTRY_CA_FILE" ]; then
-        echo "  ! REGISTRY_CA_FILE not readable: $REGISTRY_CA_FILE"
-    else
-        echo "  file        $REGISTRY_CA_FILE"
-        host="${REGISTRY_REMOTE_URL#*://}"; host="${host%%/*}"
-        if [ -n "$host" ] && [ ! -f "/etc/docker/certs.d/${host}/ca.crt" ]; then
-            echo "  ! the HOST docker daemon does not trust it yet:"
-            echo "      sudo mkdir -p /etc/docker/certs.d/${host}"
-            echo "      sudo cp ${REGISTRY_CA_FILE} /etc/docker/certs.d/${host}/ca.crt"
-            echo "    (kind nodes are handled by registry.sh; in-cluster clients are the workload's job)"
-        fi
-    fi
-fi
-
-# Host ports this environment will try to bind. Checked before cluster creation
-# because docker reports a clash halfway through, as an opaque
-# "failed to bind host port ...: address already in use".
-echo
-echo "ports  (block derived from the directory name — see 'make ports')"
-
-port_busy() {
-    if command -v ss >/dev/null 2>&1; then
-        ss -ltn "sport = :$1" 2>/dev/null | grep -q LISTEN && return 0 || return 1
-    fi
-    # iproute2 is absent from a minimal Debian, so fall back to procfs rather
-    # than silently reporting everything as free.
-    local hex; hex=$(printf ':%04X' "$1")
-    grep -qi "^ *[0-9]*: [0-9A-F]*$hex " /proc/net/tcp /proc/net/tcp6 2>/dev/null
-}
-
-# A port held by THIS environment's own cluster is not a clash — it is the thing
-# working. Reporting it as a problem every time the cluster is up would train
-# people to ignore this section, which is the opposite of the point.
-# Extract with a second grep rather than `tr -d ':->'`: in tr, ':->' is the
-# character RANGE ':' to '>', which does not contain '-', so the trailing dash
-# survives and nothing ever matches.
-ours=$(docker ps --filter "label=io.x-k8s.kind.cluster=${CLUSTER}" \
-        --format '{{.Ports}}' 2>/dev/null | tr ',' '\n' \
-        | grep -oE ':[0-9]+->' | grep -oE '[0-9]+' || true)
-
-clash=0
-for entry in "HTTP:${HTTP_PORT}" "HTTPS:${HTTPS_PORT}" \
-             "TILT:${TILT_PORT}" "REGISTRY:${REGISTRY_PORT}"; do
-    name="${entry%%:*}"; p="${entry#*:}"
-    [ -n "$p" ] || continue
-    if ! port_busy "$p"; then
-        printf "    %-9s %-6s free\n" "$name" "$p"
-    elif echo "$ours" | grep -qx "$p"; then
-        printf "    %-9s %-6s in use by this environment's cluster\n" "$name" "$p"
-    else
-        printf "  ! %-9s %-6s IN USE by something else\n" "$name" "$p"
-        clash=1
-    fi
-done
-
-if [ "$clash" -eq 1 ]; then
-    echo "    override the clashing one in ctrl/.env, e.g.  HTTP_PORT=21080"
-    echo "    (or rename this directory — the whole block follows the name)"
-fi

rig/ctrl/versions.env

@@ -1,4 +1,4 @@
-# Pinned toolchain — the single manifest the wizard installs from.
+# Pinned toolchain — the single manifest ctrl/deps.sh installs from.
 # Every entry is a single binary; none of them needs an apt repo.
 #   kubectl  fully static
 #   kind     libc only
@@ -6,8 +6,19 @@
 #   jq       upstream static build (Debian's is linked against libjq/libonig)
 #
 # Checksums are the upstream-published SHA256 of the linux/amd64 artifact.
-# To bump: change the version, then re-run `bash ctrl/versions-refresh.sh` and
-# commit the result — never hand-edit a checksum.
+#
+# To bump: change the version, then take the checksum from the release's own
+# published list — never hand-edit or hand-copy one from a download you did.
+# For anything hosted on GitHub releases that is:
+#
+#   curl -sSL https://github.com/<org>/<repo>/releases/download/<tag>/checksums.txt \
+#     | grep linux.x86_64
+#
+# (kubectl publishes its own instead: <KUBECTL_URL>.sha256.)
+#
+# There was a `ctrl/versions-refresh.sh` named here that has never existed. If
+# bumping stops being rare enough to do by hand, write it — but a comment
+# pointing at a missing script is worse than no comment.
 
 KIND_VERSION=v0.32.0
 KIND_SHA256=50030de23cf40a18505f20426f6a8506bedf13c6e509244bd1fa9463721b0f54
@@ -21,10 +32,27 @@ TILT_VERSION=0.37.6
 TILT_SHA256=e9672b8a18d43501f35dcfe98465969a7db0e436b36cf0c50c7e6f8d40de5fe6
 TILT_URL=https://github.com/tilt-dev/tilt/releases/download/v${TILT_VERSION}/tilt.${TILT_VERSION}.linux.x86_64.tar.gz
 
+# ctlptl — creates a kind cluster WITH a local registry wired in, which is what
+# keeps images off docker.io (an unqualified name means docker.io/library/<name>).
+# Same publisher and same archive shape as tilt: binary at the archive root, so
+# fetch_tgz handles it with strip=0 and no special case.
+CTLPTL_VERSION=0.9.4
+CTLPTL_SHA256=c63a1ec28e60bc3faf6becb76f53355c5cf5e0143dafdd27ad85db5584fa6b1e
+CTLPTL_URL=https://github.com/tilt-dev/ctlptl/releases/download/v${CTLPTL_VERSION}/ctlptl.${CTLPTL_VERSION}.linux.x86_64.tar.gz
+
 JQ_VERSION=1.8.2
 JQ_SHA256=b1c22172dd303f3be49e935aa56aa48a8b7a46e0bc838b4997d3bb451495870f
 JQ_URL=https://github.com/jqlang/jq/releases/download/jq-${JQ_VERSION}/jq-linux-amd64
 
+# docker compose — the distro docker packages ship the daemon and the CLI but
+# frequently not this, so `docker compose up` fails with "unknown command" on an
+# otherwise working Docker. It is a CLI plugin, found by NAME in a plugin
+# directory, so a copy in the bin dir alone only gives you the retired
+# `docker-compose` v1 spelling; deps.sh links it into ~/.docker/cli-plugins.
+COMPOSE_VERSION=5.5.1
+COMPOSE_SHA256=db1889184726840f75c4f9c001048430d4f25b3be3cb084d3ddd762bc0aed576
+COMPOSE_URL=https://github.com/docker/compose/releases/download/v${COMPOSE_VERSION}/docker-compose-linux-x86_64
+
 # Node images shipped with KIND_VERSION above, pinned by digest so a kind upgrade
 # can never silently move the k8s version. Profiles select one via K8S_VERSION.
 # Older entries are kept deliberately: running a trailing-edge control plane is
@@ -44,9 +72,9 @@ CERT_MANAGER_VERSION=v1.21.1
 METRICS_SERVER_VERSION=v0.9.0
 METALLB_VERSION=v0.16.0
 
-# Dependency containers. These mirror soleprint's cabinets
-# (soleprint/station/cabinets/), so a room that declares postgres gets the same
-# thing whether it runs on compose or in the cluster. Pinned by tag rather than
+# Cabinets — public services dropped in as-is, the upstream image unmodified.
+# The same declaration installs on compose or in the cluster, so a dependency is
+# named once and works either way. Pinned by tag rather than
 # digest because they are ordinary upstream images with no supply chain claim
 # attached — bump freely, and preload them for the offline profile.
 POSTGRES_IMAGE=postgres:16-alpine

rig/ctrl/wizard.sh

@@ -1,381 +0,0 @@
-#!/usr/bin/env bash
-# The installation wizard: detect the host, install a pinned toolchain onto it,
-# then report what it could not do. It never runs the cluster and never mutates
-# the host outside the directories mounted into it.
-#
-# Usage (normally via `make station` / `make deps`, or directly):
-#   wizard.sh detect            # report host facts only, change nothing
-#   wizard.sh fetch [core|dev] [--to DIR]   # download + verify into DIR
-#   wizard.sh install [core|dev]            # detect, fetch, install, report
-#
-# Tiers: 'core' is kubectl + jq (talk to a cluster); 'dev' adds kind and tilt
-# Default is dev.
-#
-# Runs both inside the wizard container and bare on a host. Inside the
-# container, host files are read through $HOST_ROOT (mount / as :ro); bare, it
-# falls back to /.
-
-set -euo pipefail
-
-# Keep the caller's cwd so a relative --to resolves where the user expects,
-# not against ctrl/ once we've moved.
-INVOKED_FROM="$PWD"
-cd "$(dirname "$0")"
-
-source ./versions.env
-
-# Resolve a possibly-relative path against the caller's original directory.
-abspath() {
-    case "$1" in
-        /*) echo "$1" ;;
-        *)  echo "$INVOKED_FROM/$1" ;;
-    esac
-}
-
-OUT_BIN="${OUT_BIN:-$HOME/.local/bin}"
-HOST_ROOT="${HOST_ROOT:-/}"
-DEPS_SOURCE="${DEPS_SOURCE:-upstream}"
-DEPS_ARTIFACTORY_URL="${DEPS_ARTIFACTORY_URL:-}"
-BAKED_BIN="${BAKED_BIN:-/opt/rig/bin}"
-
-# Collected by detect(), printed by report_manual() at the very end.
-MANUAL=()
-
-# Host FILES (/etc/..., /mnt/c/...) must be read through the mount. Kernel-level
-# facts (kernel version, meminfo, inotify) are shared with the container, so the
-# container's own view is already the host's.
-host_file() {
-    local p="${1#/}"
-    if [ "$HOST_ROOT" != "/" ] && [ -e "$HOST_ROOT/$p" ]; then
-        echo "$HOST_ROOT/$p"
-    else
-        echo "/$p"
-    fi
-}
-
-# ── detect ─────────────────────────────────────────────────────────────────
-
-is_wsl() { grep -qi microsoft /proc/version 2>/dev/null; }
-
-detect() {
-    echo "host"
-    echo "  kernel      $(uname -r)"
-
-    local osr; osr=$(host_file /etc/os-release)
-    [ -r "$osr" ] && echo "  distro      $(sed -n 's/^PRETTY_NAME="\(.*\)"/\1/p' "$osr")"
-
-    local total_kb avail_kb
-    total_kb=$(awk '/^MemTotal:/{print $2}' /proc/meminfo)
-    avail_kb=$(awk '/^MemAvailable:/{print $2}' /proc/meminfo)
-    printf "  memory      %d GB total, %d GB available\n" \
-        $((total_kb / 1024 / 1024)) $((avail_kb / 1024 / 1024))
-
-    if [ $((avail_kb / 1024 / 1024)) -lt 4 ]; then
-        echo "  ! under 4 GB available — a multi-node profile will struggle."
-        echo "    'make cluster list' shows the others; 'make cluster free' stops them."
-    fi
-
-    detect_wsl
-    detect_docker
-    detect_inotify
-}
-
-detect_wsl() {
-    if ! is_wsl; then
-        echo "  platform    native linux"
-        return
-    fi
-
-    echo "  platform    WSL"
-
-    # systemd is off by default in WSL, and the ingress/DNS paths that use a
-    # host service need it. Enabling it requires a Windows-side restart, which
-    # cannot be issued from inside the distro.
-    local wc; wc=$(host_file /etc/wsl.conf)
-    if [ -r "$wc" ] && grep -qE '^\s*systemd\s*=\s*true' "$wc"; then
-        echo "  systemd     enabled in wsl.conf"
-    else
-        echo "  ! systemd not enabled in /etc/wsl.conf"
-        MANUAL+=("Enable systemd — add to /etc/wsl.conf:
-      [boot]
-      systemd=true
-  then from a WINDOWS terminal (not this shell):  wsl --shutdown")
-    fi
-
-    # WSL regenerates /etc/resolv.conf on every boot, which silently reverts any
-    # local DNS setup.
-    if [ -r "$wc" ] && grep -qE '^\s*generateResolvConf\s*=\s*false' "$wc"; then
-        echo "  resolv.conf pinned (generateResolvConf=false)"
-    else
-        echo "  - resolv.conf is WSL-generated; DNS_MODE=dnsmasq would be reverted on reboot"
-    fi
-
-    local wcfg
-    wcfg=$(ls "$HOST_ROOT"/mnt/c/Users/*/.wslconfig 2>/dev/null | head -1 || true)
-    if [ -n "$wcfg" ] && grep -qE '^\s*memory\s*=' "$wcfg"; then
-        echo "  wslconfig   memory set: $(grep -E '^\s*memory\s*=' "$wcfg" | tr -d ' ')"
-    else
-        MANUAL+=("Cap/raise the WSL VM memory — in %USERPROFILE%\\.wslconfig on Windows:
-      [wsl2]
-      memory=8GB
-  then from a WINDOWS terminal:  wsl --shutdown")
-    fi
-}
-
-detect_docker() {
-    # Reachability of the daemon is the real question, and the CLI is only how
-    # we ask it. Note that when this runs inside the wizard container, Docker
-    # necessarily exists on the host — otherwise nothing would be executing —
-    # so a missing CLI in here is a wizard packaging bug, not a host problem.
-    if ! command -v docker >/dev/null 2>&1; then
-        if [ -S /var/run/docker.sock ]; then
-            echo "  docker      socket present (no cli in this context)"
-        else
-            echo "  ! docker not found and no socket at /var/run/docker.sock"
-            MANUAL+=("Install Docker — the one true prerequisite:
-      sudo apt-get install -y docker.io && sudo usermod -aG docker \"\$USER\"
-  then log out and back in.")
-        fi
-        return
-    fi
-    if docker info >/dev/null 2>&1; then
-        echo "  docker      $(docker version --format '{{.Server.Version}}' 2>/dev/null)"
-        local n
-        n=$(docker ps --filter "label=io.x-k8s.kind.cluster" --format '{{.Names}}' 2>/dev/null | wc -l)
-        # Must be an `if`, not `[ ] && echo`: as the last statement in this
-        # function the latter returns 1 when the count is zero, and `set -e`
-        # then kills the caller. That is the fresh-machine case — no clusters
-        # yet — so the bug only ever shows up where it does most harm.
-        if [ "$n" -gt 0 ]; then
-            echo "  - $n kind node container(s) already running; see 'make cluster list'"
-        fi
-    else
-        echo "  ! docker cli present but the daemon is unreachable"
-        MANUAL+=("Start Docker, or add yourself to the docker group:
-      sudo usermod -aG docker \"\$USER\"   # then log out and back in")
-    fi
-}
-
-# kind and Tilt both watch large trees. WSL ships defaults (8192/128) far too low,
-# and the failure mode is silent: Tilt simply stops noticing file changes.
-detect_inotify() {
-    local w i
-    w=$(cat /proc/sys/fs/inotify/max_user_watches 2>/dev/null || echo 0)
-    i=$(cat /proc/sys/fs/inotify/max_user_instances 2>/dev/null || echo 0)
-    echo "  inotify     watches=$w instances=$i"
-
-    if [ "$w" -lt 524288 ] || [ "$i" -lt 512 ]; then
-        echo "  ! inotify limits are low — Tilt will silently stop noticing file changes"
-        MANUAL+=("Raise inotify limits (needs root on the host):
-      echo -e 'fs.inotify.max_user_watches=524288\\nfs.inotify.max_user_instances=512' \\
-        | sudo tee /etc/sysctl.d/99-rig.conf
-      sudo sysctl --system")
-    fi
-}
-
-# ── fetch ──────────────────────────────────────────────────────────────────
-
-# Resolve where a given artifact comes from, honouring DEPS_SOURCE.
-resolve_url() {
-    local upstream="$1"
-    case "$DEPS_SOURCE" in
-        upstream)    echo "$upstream" ;;
-        artifactory)
-            if [ -z "$DEPS_ARTIFACTORY_URL" ]; then
-                echo "DEPS_SOURCE=artifactory but DEPS_ARTIFACTORY_URL is empty" >&2
-                exit 1
-            fi
-            echo "${DEPS_ARTIFACTORY_URL%/}/$(basename "$upstream")"
-            ;;
-        *) echo "unsupported DEPS_SOURCE '$DEPS_SOURCE' for a download" >&2; exit 1 ;;
-    esac
-}
-
-verify() {
-    local file="$1" want="$2" name="$3" got
-    got=$(sha256sum "$file" | awk '{print $1}')
-    if [ "$got" != "$want" ]; then
-        echo "checksum mismatch for $name" >&2
-        echo "  expected $want" >&2
-        echo "  got      $got" >&2
-        exit 1
-    fi
-}
-
-# fetch_bin <name> <url> <sha256> <dest-dir>  — a bare binary
-fetch_bin() {
-    local name="$1" url="$2" sha="$3" dest="$4"
-    local tmp="$dest/.$name.tmp"
-    echo "  fetching $name"
-    curl -fsSL --retry 3 -o "$tmp" "$(resolve_url "$url")"
-    verify "$tmp" "$sha" "$name"
-    mv "$tmp" "$dest/$name"
-    chmod +x "$dest/$name"
-}
-
-# fetch_tgz <name> <url> <sha256> <dest-dir> <path-inside-archive> <strip>
-# Archive layouts differ — tilt's is flat (the binary at the root, strip=0),
-# others nest it a directory down — so the caller says which.
-fetch_tgz() {
-    local name="$1" url="$2" sha="$3" dest="$4" inner="$5" strip="$6"
-    local tmp="$dest/.$name.tgz"
-    echo "  fetching $name"
-    curl -fsSL --retry 3 -o "$tmp" "$(resolve_url "$url")"
-    verify "$tmp" "$sha" "$name"
-    # --no-same-owner: extracting as root would otherwise restore the uid/gid
-    # baked into the archive (some ship as uid 1001), leaving a binary the host
-    # user does not own.
-    tar -xzf "$tmp" -C "$dest" --strip-components="$strip" --no-same-owner "$inner"
-    rm -f "$tmp"
-    chmod +x "$dest/$name"
-}
-
-# The wizard runs as root so it can reach the docker socket, which means
-# everything it writes into a mounted volume lands root-owned and unusable from
-# the host. Hand it back to whoever owns the mount point (the host user created
-# that directory before mounting it).
-fix_ownership() {
-    local dir="$1"
-    [ -d "$dir" ] || return 0
-    local owner="${HOST_UID:-}:${HOST_GID:-}"
-    if [ "$owner" = ":" ]; then
-        owner=$(stat -c '%u:%g' "$dir")
-    fi
-    [ "$owner" = "0:0" ] && return 0
-    chown -R "$owner" "$dir" 2>/dev/null || true
-}
-
-# Two tiers, because not every machine should get cluster tooling.
-#
-#   core  kubectl, jq — talk to a cluster someone else runs. Nothing that
-#         creates one. Appropriate on a managed or corporate-issued machine
-#         where development tools are not wanted by default.
-#   dev   core plus kind and tilt — build clusters and hot-reload into them.
-#
-# The split exists because "install the toolchain" is not one decision: on a
-# managed workspace the right answer is kubectl and nothing else.
-CORE_TOOLS="kubectl jq"
-# No helm: every addon installs with `kubectl apply -f <url>`, so nothing here
-# has ever invoked it. Add it back the day something actually needs a chart.
-DEV_TOOLS="kind tilt"
-
-fetch() {
-    local dest="$OUT_BIN" tier="${TIER:-dev}"
-    while [ $# -gt 0 ]; do
-        case "$1" in
-            --to)      dest="$2"; shift 2 ;;
-            core|dev)  tier="$1"; shift ;;
-            *)         echo "unknown argument: $1" >&2; exit 1 ;;
-        esac
-    done
-    dest="$(abspath "$dest")"
-    mkdir -p "$dest"
-    TIER="$tier"
-
-    if [ "$DEPS_SOURCE" = "baked" ]; then
-        echo "installing baked binaries from $BAKED_BIN"
-        cp -a "$BAKED_BIN"/. "$dest"/
-        fix_ownership "$dest"
-        return
-    fi
-
-    echo "fetching '$tier' toolchain (source: $DEPS_SOURCE)"
-    fetch_bin kubectl "$KUBECTL_URL" "$KUBECTL_SHA256" "$dest"
-    fetch_bin jq      "$JQ_URL"      "$JQ_SHA256"      "$dest"
-    if [ "$tier" = "dev" ]; then
-        fetch_bin kind "$KIND_URL" "$KIND_SHA256" "$dest"
-        fetch_tgz tilt "$TILT_URL" "$TILT_SHA256" "$dest" tilt 0
-    fi
-
-    fix_ownership "$dest"
-    # kind writes the kubeconfig as root too; hand that back as well when it's
-    # a mounted host directory rather than container-local state.
-    fix_ownership "${KUBE_DIR:-/out/kube}"
-}
-
-# ── install ────────────────────────────────────────────────────────────────
-
-report_manual() {
-    echo
-    if [ ${#MANUAL[@]} -eq 0 ]; then
-        echo "nothing left to do by hand."
-        return
-    fi
-    echo "host actions the wizard cannot perform (${#MANUAL[@]}):"
-    echo
-    local n=1
-    for m in "${MANUAL[@]}"; do
-        echo "  $n. $m"
-        echo
-        n=$((n + 1))
-    done
-}
-
-# Installing into a directory that sits early in PATH silently replaces whatever
-# the machine was already using — which on a shared or client machine can break
-# unrelated work (kubectl more than one minor away from a cluster is the common
-# one). Say so; never decide it for them.
-tier_tools() { [ "$1" = "core" ] && echo "$CORE_TOOLS" || echo "$CORE_TOOLS $DEV_TOOLS"; }
-
-warn_shadowing() {
-    local b existing shadowed="" tier="${1:-dev}"
-    for b in $(tier_tools "$tier"); do
-        [ -x "$OUT_BIN/$b" ] || continue
-        # Where would this resolve if OUT_BIN weren't in the way?
-        existing=$(PATH=$(echo "$PATH" | tr ':' '\n' | grep -vx "$OUT_BIN" | paste -sd:) \
-                   command -v "$b" 2>/dev/null || true)
-        [ -n "$existing" ] || continue
-        [ "$existing" = "$OUT_BIN/$b" ] && continue
-        shadowed+="      $b  $existing"$'\n'
-    done
-
-    [ -n "$shadowed" ] || return 0
-
-    case ":${PATH}:" in
-        *":$OUT_BIN:"*) ;;
-        *) return 0 ;;   # not on PATH yet, so nothing is being shadowed
-    esac
-
-    echo
-    echo "  ! these were already installed elsewhere and are now shadowed by $OUT_BIN:"
-    printf '%s' "$shadowed"
-    echo "    Other projects on this machine will pick up the new versions."
-    MANUAL+=("Decide which toolchain wins. To keep the previous one, remove what
-  was just installed:
-      rm -f $(for b in $(tier_tools "$tier"); do printf '%s ' "$OUT_BIN/$b"; done)
-  Or install somewhere private instead:
-      OUT_BIN=\$PWD/def/bin make deps   # then put that dir first in PATH")
-}
-
-install() {
-    local tier="${1:-dev}"
-    detect
-    echo
-    fetch "$tier"
-    echo
-    echo "installed to $OUT_BIN ($tier):"
-    for b in $(tier_tools "$tier"); do
-        [ -x "$OUT_BIN/$b" ] && echo "  $b"
-    done
-    if [ "$tier" = "core" ]; then
-        echo "  (no kind/tilt — 'make deps dev' adds them)"
-    fi
-    warn_shadowing "$tier"
-
-    case ":${PATH}:" in
-        *":$OUT_BIN:"*) ;;
-        *) MANUAL+=("Put the toolchain on your PATH — add to ~/.bashrc:
-      export PATH=\"${OUT_BIN}:\$PATH\"") ;;
-    esac
-
-    report_manual
-}
-
-# ── main ───────────────────────────────────────────────────────────────────
-
-case "${1:-install}" in
-    detect)  detect; report_manual ;;
-    fetch)   shift; fetch "$@" ;;
-    install) shift; install "${1:-dev}" ;;
-    *)       echo "usage: $0 [detect|fetch|install]" >&2; exit 1 ;;
-esac

rig/docs/graphs/01-install.dot

@@ -20,13 +20,13 @@ digraph rig_install {
         bin     [label="~/.local/bin\nkind · kubectl · tilt\njq" fillcolor="#121829" shape=cylinder]
     }
 
-    subgraph cluster_wizard {
+    subgraph cluster_installer {
         label="Installer container (transient)"
         style=dashed
         color="#1e2a4a"
         fontcolor="#8892a8"
 
-        wizard  [label="wizard\ncurl · jq · python · graphviz" fillcolor="#121829"]
+        installer [label="deps installer\ncurl · jq · python · graphviz" fillcolor="#121829"]
         detect  [label="detect host\nWSL · memory · inotify · docker" fillcolor="#121829"]
         fetch   [label="fetch + verify\nSHA256, pinned versions" fillcolor="#121829"]
     }
@@ -34,14 +34,14 @@ digraph rig_install {
     upstream [label="upstream\nreleases / corporate mirror" fillcolor="#1a3a1a" fontcolor="#00c853" shape=octagon]
     report   [label="report what it\nCANNOT do" fillcolor="#3a1a1a" fontcolor="#ffc107"]
 
-    docker  -> wizard  [label="docker run"]
-    wizard  -> detect
+    docker  -> installer  [label="docker run"]
+    installer  -> detect
     detect  -> fetch
     fetch   -> upstream [label="pinned + checksummed" color="#00c853"]
     fetch   -> bin      [label="install"]
     detect  -> report   [style=dashed label="sudo / Windows-side steps" color="#ffc107"]
 
     // The container is gone after this; nothing depends on it at run time.
-    wizard -> gone [style=dotted label="exits"]
+    installer -> gone [style=dotted label="exits"]
     gone [label="(container discarded)" fillcolor="#0a0e17" fontcolor="#4a5568" color="#1e2a4a" style="filled,dashed"]
 }

rig/docs/graphs/01-install.svg

@@ -16,7 +16,7 @@
 <text xml:space="preserve" text-anchor="middle" x="1011.15" y="-170.8" font-family="Helvetica,sans-Serif" font-size="16.00" fill="#8892a8">Your machine</text>
 </g>
 <g id="clust2" class="cluster">
-<title>cluster_wizard</title>
+<title>cluster_installer</title>
 <polygon fill="#0a0e17" stroke="#1e2a4a" stroke-dasharray="5,2" points="8,-95 8,-175 745.5,-175 745.5,-95 8,-95"/>
 <text xml:space="preserve" text-anchor="middle" x="376.75" y="-155.8" font-family="Helvetica,sans-Serif" font-size="16.00" fill="#8892a8">Installer container (transient)</text>
 </g>
@@ -27,16 +27,16 @@
 <text xml:space="preserve" text-anchor="middle" x="1011.15" y="-46.05" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#0066ff">Docker</text>
 <text xml:space="preserve" text-anchor="middle" x="1011.15" y="-32.55" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#0066ff">(the one prerequisite)</text>
 </g>
-<!-- wizard -->
+<!-- installer -->
 <g id="node3" class="node">
-<title>wizard</title>
+<title>installer</title>
 <polygon fill="#121829" stroke="#1e2a4a" points="181.25,-139 16,-139 16,-103 181.25,-103 181.25,-139"/>
-<text xml:space="preserve" text-anchor="middle" x="98.62" y="-124.05" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">wizard</text>
+<text xml:space="preserve" text-anchor="middle" x="98.62" y="-124.05" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">deps installer</text>
 <text xml:space="preserve" text-anchor="middle" x="98.62" y="-110.55" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">curl · jq · python · graphviz</text>
 </g>
-<!-- docker&#45;&gt;wizard -->
+<!-- docker&#45;&gt;installer -->
 <g id="edge1" class="edge">
-<title>docker&#45;&gt;wizard</title>
+<title>docker&#45;&gt;installer</title>
 <path fill="none" stroke="#4a5568" d="M906.12,-45.7C757.47,-50.51 475.98,-63.12 238.25,-94 223.38,-95.93 207.7,-98.48 192.45,-101.24"/>
 <polygon fill="#4a5568" stroke="#4a5568" points="192.13,-97.74 182.93,-103.01 193.4,-104.63 192.13,-97.74"/>
 <text xml:space="preserve" text-anchor="middle" x="506.62" y="-74.39" font-family="Helvetica,sans-Serif" font-size="9.00" fill="#8892a8">docker run</text>
@@ -57,9 +57,9 @@
 <text xml:space="preserve" text-anchor="middle" x="333.62" y="-124.05" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">detect host</text>
 <text xml:space="preserve" text-anchor="middle" x="333.62" y="-110.55" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">WSL · memory · inotify · docker</text>
 </g>
-<!-- wizard&#45;&gt;detect -->
+<!-- installer&#45;&gt;detect -->
 <g id="edge2" class="edge">
-<title>wizard&#45;&gt;detect</title>
+<title>installer&#45;&gt;detect</title>
 <path fill="none" stroke="#4a5568" d="M181.44,-121C195.97,-121 211.28,-121 226.37,-121"/>
 <polygon fill="#4a5568" stroke="#4a5568" points="226.33,-124.5 236.33,-121 226.33,-117.5 226.33,-124.5"/>
 </g>
@@ -69,9 +69,9 @@
 <polygon fill="#0a0e17" stroke="#1e2a4a" stroke-dasharray="5,2" points="400.5,-47 266.75,-47 266.75,-11 400.5,-11 400.5,-47"/>
 <text xml:space="preserve" text-anchor="middle" x="333.62" y="-25.3" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#4a5568">(container discarded)</text>
 </g>
-<!-- wizard&#45;&gt;gone -->
+<!-- installer&#45;&gt;gone -->
 <g id="edge7" class="edge">
-<title>wizard&#45;&gt;gone</title>
+<title>installer&#45;&gt;gone</title>
 <path fill="none" stroke="#4a5568" stroke-dasharray="1,5" d="M119.29,-102.62C138.26,-86 168.54,-62.29 199.25,-49.75 216.76,-42.6 236.49,-37.9 255.29,-34.81"/>
 <polygon fill="#4a5568" stroke="#4a5568" points="255.67,-38.29 265.04,-33.35 254.64,-31.37 255.67,-38.29"/>
 <text xml:space="preserve" text-anchor="middle" x="209.75" y="-52.45" font-family="Helvetica,sans-Serif" font-size="9.00" fill="#8892a8">exits</text>

rig/docs/graphs/02-environment.dot

@@ -20,7 +20,7 @@ digraph rig_environment {
 
         cname   [label="cluster name\nacmebank" fillcolor="#121829"]
         ctx     [label="kubectl context\nkind-acmebank" fillcolor="#121829"]
-        img     [label="image tag\nacmebank-wizard" fillcolor="#121829"]
+        img     [label="image tag\nacmebank-deps" fillcolor="#121829"]
         ports   [label="port block\n21300–21309" fillcolor="#121829"]
         reg     [label="registry container\nacmebank-registry" fillcolor="#121829"]
     }

rig/docs/graphs/02-environment.svg

@@ -4,28 +4,28 @@
 <!-- Generated by graphviz version 14.1.2 (0)
  -->
 <!-- Title: rig_environment Pages: 1 -->
-<svg width="971pt" height="481pt"
- viewBox="0.00 0.00 971.00 481.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
+<svg width="962pt" height="481pt"
+ viewBox="0.00 0.00 962.00 481.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
 <g id="graph0" class="graph" transform="scale(1 1) rotate(0) translate(4 476.83)">
 <title>rig_environment</title>
-<polygon fill="#0a0e17" stroke="none" points="-4,4 -4,-476.83 967,-476.83 967,4 -4,4"/>
-<text xml:space="preserve" text-anchor="middle" x="481.5" y="-453.63" font-family="Helvetica,sans-Serif" font-size="16.00" fill="#0066ff">One environment per directory — copies never collide</text>
+<polygon fill="#0a0e17" stroke="none" points="-4,4 -4,-476.83 958,-476.83 958,4 -4,4"/>
+<text xml:space="preserve" text-anchor="middle" x="477" y="-453.63" font-family="Helvetica,sans-Serif" font-size="16.00" fill="#0066ff">One environment per directory — copies never collide</text>
 <g id="clust1" class="cluster">
 <title>cluster_derived</title>
-<polygon fill="#0a0e17" stroke="#1e2a4a" stroke-dasharray="5,2" points="8,-65 8,-144.5 605,-144.5 605,-65 8,-65"/>
-<text xml:space="preserve" text-anchor="middle" x="306.5" y="-125.3" font-family="Helvetica,sans-Serif" font-size="16.00" fill="#8892a8">Everything below is derived from it</text>
+<polygon fill="#0a0e17" stroke="#1e2a4a" stroke-dasharray="5,2" points="8,-65 8,-144.5 596,-144.5 596,-65 8,-65"/>
+<text xml:space="preserve" text-anchor="middle" x="302" y="-125.3" font-family="Helvetica,sans-Serif" font-size="16.00" fill="#8892a8">Everything below is derived from it</text>
 </g>
 <g id="clust2" class="cluster">
 <title>cluster_config</title>
-<polygon fill="#0a0e17" stroke="#1e2a4a" stroke-dasharray="5,2" points="613,-65 613,-437.33 955,-437.33 955,-65 613,-65"/>
-<text xml:space="preserve" text-anchor="middle" x="784" y="-418.13" font-family="Helvetica,sans-Serif" font-size="16.00" fill="#8892a8">Configuration — weakest first, later wins</text>
+<polygon fill="#0a0e17" stroke="#1e2a4a" stroke-dasharray="5,2" points="604,-65 604,-437.33 946,-437.33 946,-65 604,-65"/>
+<text xml:space="preserve" text-anchor="middle" x="775" y="-418.13" font-family="Helvetica,sans-Serif" font-size="16.00" fill="#8892a8">Configuration — weakest first, later wins</text>
 </g>
 <!-- dirname -->
 <g id="node1" class="node">
 <title>dirname</title>
-<polygon fill="#1f6feb" stroke="#1e2a4a" points="375.11,-197.44 375.11,-219.63 329.94,-235.33 266.06,-235.33 220.89,-219.63 220.89,-197.44 266.06,-181.75 329.94,-181.75 375.11,-197.44"/>
-<text xml:space="preserve" text-anchor="middle" x="298" y="-211.59" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#ffffff">directory name</text>
-<text xml:space="preserve" text-anchor="middle" x="298" y="-198.09" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#ffffff">e.g. acmebank/</text>
+<polygon fill="#1f6feb" stroke="#1e2a4a" points="370.11,-197.44 370.11,-219.63 324.94,-235.33 261.06,-235.33 215.89,-219.63 215.89,-197.44 261.06,-181.75 324.94,-181.75 370.11,-197.44"/>
+<text xml:space="preserve" text-anchor="middle" x="293" y="-211.59" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#ffffff">directory name</text>
+<text xml:space="preserve" text-anchor="middle" x="293" y="-198.09" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#ffffff">e.g. acmebank/</text>
 </g>
 <!-- cname -->
 <g id="node2" class="node">
@@ -37,8 +37,8 @@
 <!-- dirname&#45;&gt;cname -->
 <g id="edge1" class="edge">
 <title>dirname&#45;&gt;cname</title>
-<path fill="none" stroke="#4a5568" d="M232.76,-193.05C195.81,-183.01 149.78,-167.28 113,-144.5 101.37,-137.29 90.31,-127.09 81.33,-117.6"/>
-<polygon fill="#4a5568" stroke="#4a5568" points="84.04,-115.37 74.73,-110.3 78.85,-120.07 84.04,-115.37"/>
+<path fill="none" stroke="#4a5568" d="M228.68,-192.51C192.88,-182.36 148.52,-166.7 113,-144.5 101.4,-137.25 90.34,-127.04 81.36,-117.55"/>
+<polygon fill="#4a5568" stroke="#4a5568" points="84.07,-115.32 74.76,-110.26 78.88,-120.02 84.07,-115.32"/>
 </g>
 <!-- ctx -->
 <g id="node3" class="node">
@@ -50,120 +50,120 @@
 <!-- dirname&#45;&gt;ctx -->
 <g id="edge2" class="edge">
 <title>dirname&#45;&gt;ctx</title>
-<path fill="none" stroke="#4a5568" d="M269.64,-181.32C248.71,-161.98 220.42,-135.83 199.86,-116.83"/>
-<polygon fill="#4a5568" stroke="#4a5568" points="202.4,-114.41 192.68,-110.19 197.65,-119.55 202.4,-114.41"/>
+<path fill="none" stroke="#4a5568" d="M265.77,-181.32C245.77,-162.07 218.77,-136.06 199.05,-117.08"/>
+<polygon fill="#4a5568" stroke="#4a5568" points="201.55,-114.63 191.91,-110.21 196.69,-119.67 201.55,-114.63"/>
 </g>
 <!-- img -->
 <g id="node4" class="node">
 <title>img</title>
-<polygon fill="#121829" stroke="#1e2a4a" points="354,-109 242,-109 242,-73 354,-73 354,-109"/>
-<text xml:space="preserve" text-anchor="middle" x="298" y="-94.05" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">image tag</text>
-<text xml:space="preserve" text-anchor="middle" x="298" y="-80.55" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">acmebank&#45;wizard</text>
+<polygon fill="#121829" stroke="#1e2a4a" points="344.12,-109 241.88,-109 241.88,-73 344.12,-73 344.12,-109"/>
+<text xml:space="preserve" text-anchor="middle" x="293" y="-94.05" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">image tag</text>
+<text xml:space="preserve" text-anchor="middle" x="293" y="-80.55" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">acmebank&#45;deps</text>
 </g>
 <!-- dirname&#45;&gt;img -->
 <g id="edge3" class="edge">
 <title>dirname&#45;&gt;img</title>
-<path fill="none" stroke="#4a5568" d="M298,-181.32C298,-163.19 298,-139.07 298,-120.47"/>
-<polygon fill="#4a5568" stroke="#4a5568" points="301.5,-120.67 298,-110.67 294.5,-120.67 301.5,-120.67"/>
+<path fill="none" stroke="#4a5568" d="M293,-181.32C293,-163.19 293,-139.07 293,-120.47"/>
+<polygon fill="#4a5568" stroke="#4a5568" points="296.5,-120.67 293,-110.67 289.5,-120.67 296.5,-120.67"/>
 </g>
 <!-- ports -->
 <g id="node5" class="node">
 <title>ports</title>
-<polygon fill="#121829" stroke="#1e2a4a" points="460.38,-109 371.62,-109 371.62,-73 460.38,-73 460.38,-109"/>
-<text xml:space="preserve" text-anchor="middle" x="416" y="-94.05" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">port block</text>
-<text xml:space="preserve" text-anchor="middle" x="416" y="-80.55" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">21300–21309</text>
+<polygon fill="#121829" stroke="#1e2a4a" points="451.38,-109 362.62,-109 362.62,-73 451.38,-73 451.38,-109"/>
+<text xml:space="preserve" text-anchor="middle" x="407" y="-94.05" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">port block</text>
+<text xml:space="preserve" text-anchor="middle" x="407" y="-80.55" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">21300–21309</text>
 </g>
 <!-- dirname&#45;&gt;ports -->
 <g id="edge4" class="edge">
 <title>dirname&#45;&gt;ports</title>
-<path fill="none" stroke="#4a5568" d="M324.94,-181.53C336.66,-170.19 350.54,-156.71 363,-144.5 372.11,-135.57 382.06,-125.74 390.85,-117.02"/>
-<polygon fill="#4a5568" stroke="#4a5568" points="393.15,-119.67 397.79,-110.14 388.22,-114.7 393.15,-119.67"/>
+<path fill="none" stroke="#4a5568" d="M318.87,-181.32C337.78,-162.15 363.29,-136.3 382,-117.34"/>
+<polygon fill="#4a5568" stroke="#4a5568" points="384.47,-119.81 389,-110.24 379.49,-114.9 384.47,-119.81"/>
 </g>
 <!-- reg -->
 <g id="node6" class="node">
 <title>reg</title>
-<polygon fill="#121829" stroke="#1e2a4a" points="597.38,-109 478.62,-109 478.62,-73 597.38,-73 597.38,-109"/>
-<text xml:space="preserve" text-anchor="middle" x="538" y="-94.05" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">registry container</text>
-<text xml:space="preserve" text-anchor="middle" x="538" y="-80.55" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">acmebank&#45;registry</text>
+<polygon fill="#121829" stroke="#1e2a4a" points="588.38,-109 469.62,-109 469.62,-73 588.38,-73 588.38,-109"/>
+<text xml:space="preserve" text-anchor="middle" x="529" y="-94.05" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">registry container</text>
+<text xml:space="preserve" text-anchor="middle" x="529" y="-80.55" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">acmebank&#45;registry</text>
 </g>
 <!-- dirname&#45;&gt;reg -->
 <g id="edge5" class="edge">
 <title>dirname&#45;&gt;reg</title>
-<path fill="none" stroke="#4a5568" d="M356.87,-190.63C390.74,-179.74 433.49,-163.98 469,-144.5 483.3,-136.65 497.86,-126.04 509.9,-116.41"/>
-<polygon fill="#4a5568" stroke="#4a5568" points="511.88,-119.31 517.39,-110.26 507.44,-113.9 511.88,-119.31"/>
+<path fill="none" stroke="#4a5568" d="M350.86,-190.3C383.87,-179.35 425.43,-163.64 460,-144.5 474.27,-136.6 488.83,-125.98 500.87,-116.36"/>
+<polygon fill="#4a5568" stroke="#4a5568" points="502.85,-119.26 508.36,-110.21 498.41,-113.84 502.85,-119.26"/>
 </g>
 <!-- cluster -->
 <g id="node11" class="node">
 <title>cluster</title>
-<polygon fill="#1a1a3a" stroke="#1e2a4a" points="469.81,-10.54 469.81,-25.46 438.29,-36 393.71,-36 362.19,-25.46 362.19,-10.54 393.71,0 438.29,0 469.81,-10.54"/>
-<text xml:space="preserve" text-anchor="middle" x="416" y="-14.3" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#0066ff">kind cluster</text>
+<polygon fill="#1a1a3a" stroke="#1e2a4a" points="460.81,-10.54 460.81,-25.46 429.29,-36 384.71,-36 353.19,-25.46 353.19,-10.54 384.71,0 429.29,0 460.81,-10.54"/>
+<text xml:space="preserve" text-anchor="middle" x="407" y="-14.3" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#0066ff">kind cluster</text>
 </g>
 <!-- cname&#45;&gt;cluster -->
 <g id="edge9" class="edge">
 <title>cname&#45;&gt;cluster</title>
-<path fill="none" stroke="#4a5568" d="M93.06,-72.62C99.54,-69.72 106.38,-67.02 113,-65 192.78,-40.7 288.45,-28.91 350.65,-23.42"/>
-<polygon fill="#4a5568" stroke="#4a5568" points="350.69,-26.93 360.36,-22.6 350.1,-19.96 350.69,-26.93"/>
+<path fill="none" stroke="#4a5568" d="M93.35,-72.51C99.75,-69.67 106.48,-67 113,-65 189.55,-41.49 281.19,-29.58 341.58,-23.85"/>
+<polygon fill="#4a5568" stroke="#4a5568" points="341.71,-27.35 351.35,-22.96 341.07,-20.38 341.71,-27.35"/>
 </g>
 <!-- ports&#45;&gt;cluster -->
 <g id="edge10" class="edge">
 <title>ports&#45;&gt;cluster</title>
-<path fill="none" stroke="#4a5568" d="M416,-72.81C416,-65.23 416,-56.1 416,-47.54"/>
-<polygon fill="#4a5568" stroke="#4a5568" points="419.5,-47.54 416,-37.54 412.5,-47.54 419.5,-47.54"/>
+<path fill="none" stroke="#4a5568" d="M407,-72.81C407,-65.23 407,-56.1 407,-47.54"/>
+<polygon fill="#4a5568" stroke="#4a5568" points="410.5,-47.54 407,-37.54 403.5,-47.54 410.5,-47.54"/>
 </g>
 <!-- versions -->
 <g id="node7" class="node">
 <title>versions</title>
-<polygon fill="#121829" stroke="#1e2a4a" points="759.38,-401.83 652.62,-401.83 652.62,-365.83 759.38,-365.83 759.38,-401.83"/>
-<text xml:space="preserve" text-anchor="middle" x="706" y="-386.88" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">versions.env</text>
-<text xml:space="preserve" text-anchor="middle" x="706" y="-373.38" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">pinned toolchain</text>
+<polygon fill="#121829" stroke="#1e2a4a" points="750.38,-401.83 643.62,-401.83 643.62,-365.83 750.38,-365.83 750.38,-401.83"/>
+<text xml:space="preserve" text-anchor="middle" x="697" y="-386.88" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">versions.env</text>
+<text xml:space="preserve" text-anchor="middle" x="697" y="-373.38" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">pinned toolchain</text>
 </g>
 <!-- profile -->
 <g id="node8" class="node">
 <title>profile</title>
-<polygon fill="#121829" stroke="#1e2a4a" points="790.5,-318.58 621.5,-318.58 621.5,-282.58 790.5,-282.58 790.5,-318.58"/>
-<text xml:space="preserve" text-anchor="middle" x="706" y="-303.63" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">env.d/&lt;profile&gt;.env</text>
-<text xml:space="preserve" text-anchor="middle" x="706" y="-290.13" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">nodes · CNI · audit · addons</text>
+<polygon fill="#121829" stroke="#1e2a4a" points="781.5,-318.58 612.5,-318.58 612.5,-282.58 781.5,-282.58 781.5,-318.58"/>
+<text xml:space="preserve" text-anchor="middle" x="697" y="-303.63" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">env.d/&lt;profile&gt;.env</text>
+<text xml:space="preserve" text-anchor="middle" x="697" y="-290.13" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">nodes · CNI · audit · addons</text>
 </g>
 <!-- versions&#45;&gt;profile -->
 <g id="edge6" class="edge">
 <title>versions&#45;&gt;profile</title>
-<path fill="none" stroke="#4a5568" d="M706,-365.59C706,-355.32 706,-342.03 706,-330.21"/>
-<polygon fill="#4a5568" stroke="#4a5568" points="709.5,-330.58 706,-320.58 702.5,-330.58 709.5,-330.58"/>
-<text xml:space="preserve" text-anchor="middle" x="737.5" y="-339.28" font-family="Helvetica,sans-Serif" font-size="9.00" fill="#8892a8">overridden by</text>
+<path fill="none" stroke="#4a5568" d="M697,-365.59C697,-355.32 697,-342.03 697,-330.21"/>
+<polygon fill="#4a5568" stroke="#4a5568" points="700.5,-330.58 697,-320.58 693.5,-330.58 700.5,-330.58"/>
+<text xml:space="preserve" text-anchor="middle" x="728.5" y="-339.28" font-family="Helvetica,sans-Serif" font-size="9.00" fill="#8892a8">overridden by</text>
 </g>
 <!-- localenv -->
 <g id="node9" class="node">
 <title>localenv</title>
-<polygon fill="#121829" stroke="#1e2a4a" points="761.88,-226.54 646.12,-226.54 646.12,-190.54 761.88,-190.54 761.88,-226.54"/>
-<text xml:space="preserve" text-anchor="middle" x="704" y="-211.59" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">ctrl/.env</text>
-<text xml:space="preserve" text-anchor="middle" x="704" y="-198.09" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">secrets, overrides</text>
+<polygon fill="#121829" stroke="#1e2a4a" points="752.88,-226.54 637.12,-226.54 637.12,-190.54 752.88,-190.54 752.88,-226.54"/>
+<text xml:space="preserve" text-anchor="middle" x="695" y="-211.59" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">ctrl/.env</text>
+<text xml:space="preserve" text-anchor="middle" x="695" y="-198.09" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#e8eaf0">secrets, overrides</text>
 </g>
 <!-- profile&#45;&gt;localenv -->
 <g id="edge7" class="edge">
 <title>profile&#45;&gt;localenv</title>
-<path fill="none" stroke="#4a5568" d="M705.61,-282.22C705.34,-269.76 704.96,-252.69 704.64,-238.23"/>
-<polygon fill="#4a5568" stroke="#4a5568" points="708.14,-238.28 704.42,-228.36 701.14,-238.43 708.14,-238.28"/>
-<text xml:space="preserve" text-anchor="middle" x="736.68" y="-256.03" font-family="Helvetica,sans-Serif" font-size="9.00" fill="#8892a8">overridden by</text>
+<path fill="none" stroke="#4a5568" d="M696.61,-282.22C696.34,-269.76 695.96,-252.69 695.64,-238.23"/>
+<polygon fill="#4a5568" stroke="#4a5568" points="699.14,-238.28 695.42,-228.36 692.14,-238.43 699.14,-238.28"/>
+<text xml:space="preserve" text-anchor="middle" x="727.68" y="-256.03" font-family="Helvetica,sans-Serif" font-size="9.00" fill="#8892a8">overridden by</text>
 </g>
 <!-- shell -->
 <g id="node10" class="node">
 <title>shell</title>
-<polygon fill="#1a3a1a" stroke="#1e2a4a" points="772.38,-109 623.62,-109 623.62,-73 772.38,-73 772.38,-109"/>
-<text xml:space="preserve" text-anchor="middle" x="698" y="-94.05" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#00c853">the environment</text>
-<text xml:space="preserve" text-anchor="middle" x="698" y="-80.55" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#00c853">PROFILE=client make …</text>
+<polygon fill="#1a3a1a" stroke="#1e2a4a" points="763.38,-109 614.62,-109 614.62,-73 763.38,-73 763.38,-109"/>
+<text xml:space="preserve" text-anchor="middle" x="689" y="-94.05" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#00c853">the environment</text>
+<text xml:space="preserve" text-anchor="middle" x="689" y="-80.55" font-family="Helvetica,sans-Serif" font-size="11.00" fill="#00c853">PROFILE=client make …</text>
 </g>
 <!-- localenv&#45;&gt;shell -->
 <g id="edge8" class="edge">
 <title>localenv&#45;&gt;shell</title>
-<path fill="none" stroke="#00c853" d="M703.11,-190.49C702.16,-172.16 700.63,-142.72 699.49,-120.79"/>
-<polygon fill="#00c853" stroke="#00c853" points="703,-120.81 698.99,-111.01 696.01,-121.18 703,-120.81"/>
-<text xml:space="preserve" text-anchor="middle" x="733.21" y="-155.2" font-family="Helvetica,sans-Serif" font-size="9.00" fill="#8892a8">overridden by</text>
+<path fill="none" stroke="#00c853" d="M694.11,-190.49C693.16,-172.16 691.63,-142.72 690.49,-120.79"/>
+<polygon fill="#00c853" stroke="#00c853" points="694,-120.81 689.99,-111.01 687.01,-121.18 694,-120.81"/>
+<text xml:space="preserve" text-anchor="middle" x="724.21" y="-155.2" font-family="Helvetica,sans-Serif" font-size="9.00" fill="#8892a8">overridden by</text>
 </g>
 <!-- shell&#45;&gt;cluster -->
 <g id="edge11" class="edge">
 <title>shell&#45;&gt;cluster</title>
-<path fill="none" stroke="#4a5568" stroke-dasharray="5,2" d="M637.29,-72.6C627.83,-69.99 618.17,-67.38 609,-65 562.72,-52.99 509.89,-40.48 471.22,-31.54"/>
-<polygon fill="#4a5568" stroke="#4a5568" points="472.2,-28.18 461.67,-29.35 470.63,-35 472.2,-28.18"/>
+<path fill="none" stroke="#4a5568" stroke-dasharray="5,2" d="M628.29,-72.6C618.83,-69.99 609.17,-67.38 600,-65 553.72,-52.99 500.89,-40.48 462.22,-31.54"/>
+<polygon fill="#4a5568" stroke="#4a5568" points="463.2,-28.18 452.67,-29.35 461.63,-35 463.2,-28.18"/>
 </g>
 </g>
 </svg>

rig/docs/index.html

@@ -265,11 +265,11 @@
       <h3>The only prerequisite</h3>
       <p><b>Docker.</b> No curl, no jq, no python, no apt repositories to configure.</p>
 <pre><code><span class="c"># then, in the environment directory:</span>
-make station     <span class="c"># is this workstation ready? reports, never fixes</span>
+make check     <span class="c"># is this machine ready? reports, never fixes</span>
 make deps        <span class="c"># install the pinned toolchain</span>
 make cluster up  <span class="c"># build the cluster for the active profile</span>
 </code></pre>
-      <p>Read <code>make station</code> before <code>make deps</code>. It never changes
+      <p>Read <code>make check</code> before <code>make deps</code>. It never changes
       anything — it prints what it found and, at the end, the steps it cannot perform
       for you.</p>
     </div>
@@ -280,13 +280,13 @@ make cluster up  <span class="c"># build the cluster for the active profile</spa
     <p class="lede">Start to finish, in order, with what each one actually does.</p>
     <div class="prose">
 
-      <h3>1 &middot; make station</h3>
-      <p>Asks whether this workstation is ready. It <b>changes nothing</b> — it
+      <h3>1 &middot; make check</h3>
+      <p>Asks whether this machine is ready. It <b>changes nothing</b> — it
       reports what it found and, at the end, the things only a human can do
       (anything needing <code>sudo</code>, or a Windows-side restart). Read it
       before installing anything; it is faster than discovering the same problems
       one failure at a time.</p>
-<pre><code>make station</code></pre>
+<pre><code>make check</code></pre>
 
       <h3>2 &middot; make setup</h3>
       <p>Does the preparation that can be automated: installs the pinned
@@ -381,8 +381,8 @@ make setup core  <span class="c"># same distinction, via setup</span>
       wants only Docker.</p>
 
       <h3>Air-gapped</h3>
-<pre><code>make wizard full                       <span class="c"># bakes every binary into the image</span>
-docker save …-wizard:full | gzip &gt; rig.tgz
+<pre><code>make deps-image full                       <span class="c"># bakes every binary into the image</span>
+docker save …-deps:full | gzip &gt; rig.tgz
 <span class="c"># carry that one file in, then:</span>
 docker load &lt; rig.tgz &amp;&amp; make cluster up PROFILE=offline
 </code></pre>
@@ -485,7 +485,7 @@ docker load &lt; rig.tgz &amp;&amp; make cluster up PROFILE=offline
       registry is usually behind an internal CA, and trust has to reach
       <b>three</b> places: the host Docker daemon, every cluster node's containerd
       (nodes do <i>not</i> inherit host trust), and any in-cluster client. Set
-      <code>REGISTRY_CA_FILE</code> and <code>make station</code> reports which is
+      <code>REGISTRY_CA_FILE</code> and <code>make check</code> reports which is
       still missing. The symptom otherwise is an opaque
       <code>x509: certificate signed by unknown authority</code>.</p></div>
 
@@ -528,11 +528,11 @@ docker load &lt; rig.tgz &amp;&amp; make cluster up PROFILE=offline
       <h3>Tilt stops noticing file changes</h3>
       <p>Almost always <code>inotify</code> limits, and it fails <i>silently</i> —
       nothing errors, changes just stop being picked up. Defaults on WSL are far too
-      low. <code>make station</code> reports it and prints the fix.</p>
+      low. <code>make check</code> reports it and prints the fix.</p>
 
       <h3>Cluster creation dies halfway with a port error</h3>
       <p>Docker reports <code>failed to bind host port … address already in use</code>
-      partway through creating the cluster. Run <code>make station</code> first — it
+      partway through creating the cluster. Run <code>make check</code> first — it
       checks every port in this environment's block before anything is built.</p>
 
       <h3>Every node stays NotReady</h3>

rig/sample-rig/.gitignore

@@ -1,9 +0,0 @@
-# NOTE: generated/ is deliberately NOT ignored. The artifact IS the deliverable —
-# the whole point is a folder you copy, apply and boot without generating
-# anything first. Regenerate it with `make manifest` after editing bundle.json
-# or app/serve.py, and commit the result.
-#
-# (This differs from rig's ctrl/k8s/generated, which is a local build artifact.)
-
-__pycache__/
-*.pyc

rig/sample-rig/Makefile

@@ -1,50 +0,0 @@
-# Thin control Makefile — one target per ctrl/ script, subcommand as an
-# argument. Same shape as rig's, for the same reason: the logic lives in the
-# script, never here.
-#
-#   make up        ->  ctrl/bundle.sh up
-#   make bundle down
-#
-# This directory is a BUNDLE, not an installer. It needs a cluster, which rig
-# owns:
-#
-#   cd .. && make cluster up     # kind cluster for this environment
-#   make up                      # then deploy this bundle into it
-#
-# Start with:  make up
-
-ARGS := $(wordlist 2,$(words $(MAKECMDGOALS)),$(MAKECMDGOALS))
-ifneq ($(ARGS),)
-$(eval $(ARGS):;@:)
-endif
-
-.DEFAULT_GOAL := help
-.PHONY: help bundle manifest up down status url list dev
-
-help:                          ## list targets
-	@grep -hE '^[a-z]+:.*?##' $(MAKEFILE_LIST) | sed 's/:.*##/\t/' | expand -t16
-
-bundle:                        ## the bundle  [manifest|up|down|status|url|list|dev]
-	bash ctrl/bundle.sh $(or $(ARGS),status)
-
-# Shorthands for the ones used constantly.
-manifest:                      ## regenerate generated/<slug>.yaml — no cluster needed
-	bash ctrl/bundle.sh manifest
-
-up:                            ## deploy this rig (installs MetalLB if absent)
-	bash ctrl/bundle.sh up
-
-down:                          ## remove this rig (leaves cluster, MetalLB, siblings)
-	bash ctrl/bundle.sh down
-
-status:                        ## what is deployed for this rig
-	bash ctrl/bundle.sh status
-
-url:                           ## the address MetalLB assigned
-	bash ctrl/bundle.sh url
-
-list:                          ## every rig in this cluster, with addresses
-	bash ctrl/bundle.sh list
-
-dev:                           ## run the UI locally with vite — no cluster needed
-	bash ctrl/bundle.sh dev

rig/sample-rig/README.md

@@ -1,140 +0,0 @@
-# sample-rig
-
-A minimal, non-sensitive bundle that proves an installation works and shows what
-shipped. Copy it, rename it, and you have another rig.
-
-```bash
-make manifest    # generate the artifact — no cluster, no kubectl needed
-make up          # deploy it into the local cluster
-make list        # every rig in this cluster, with addresses
-make dev         # run the UI locally with vite, no cluster at all
-```
-
-`make up` prints an address. Open it and the page says **IT WORKS**, then lists
-the tools and rigs in the bundle.
-
-## What it is for
-
-Three jobs, in the order you hit them:
-
-1. **Prove the install.** kind is there, a cluster exists, MetalLB hands out an
-   address, a `type: LoadBalancer` Service actually resolves, and a pod serves.
-   If all of that works, the environment is sound.
-2. **Say what shipped.** The page renders [`bundle.json`](bundle.json) —
-   standalone tools and rigs, flat, with none of soleprint's internal hierarchy.
-   Editing that file is the only step needed to change the listing.
-3. **Stand in for the real thing.** Nothing here is sensitive. The real
-   architecture connects separately, against a setup already known to work.
-
-## The UI is a complement, not the product
-
-`rig-ui/` is just a vite app. It complements a rig; a rig is complete and useful
-without it, and nothing depends on it being there. It is deliberately **not**
-generated by kind or tilt — you copy the folder into a rig after that rig is
-pulled, and apply one manifest:
-
-```bash
-kubectl apply -n <namespace> -f rig-ui/k8s.yaml
-```
-
-That file is the whole integration: one Pod running `npm run dev` on
-`node:22-alpine`, one Service. A bare Pod rather than a Deployment because this
-is a dev-loop convenience, not a workload to keep alive.
-
-The app and `bundle.json` arrive as a ConfigMap, so nothing is baked into an
-image and editing the bundle is the entire update cycle. The container runs
-`npm install` at start, which needs egress to a registry — on a locked-down
-cluster point npm at the internal one, or bake an image instead. Nothing else
-changes if you do.
-
-## One artifact, two destinations
-
-`ctrl/manifest.py` emits `generated/<slug>.yaml` — namespace, the app and
-bundle embedded in a ConfigMap, Pod, Service. It is self-contained and applies
-unmodified anywhere:
-
-```bash
-kubectl apply -f generated/sample-rig.yaml     # local kind, or an external cluster
-```
-
-`make up` applies **that same file**. There is no separate local path, so what
-works here cannot quietly differ from what is applied elsewhere.
-
-This is what `type: LoadBalancer` buys. MetalLB answers it on kind; the AWS load
-balancer controller answers it on EKS. NodePort would not survive the trip — it
-is a single cluster-wide port range, so two rigs would have to negotiate numbers.
-
-**VPC-agnostic on purpose.** The target is EKS, but the Service carries no
-annotations — no `aws-load-balancer-subnets`, no security groups, no `-scheme`,
-no `-type: nlb`. Each of those encodes a specific network layout, and one of them
-appearing here would pin the artifact to the account and VPC it was written
-against, which is precisely what stops it also working on kind. Subnet discovery
-is the cluster's business: EKS resolves it from the tags its own subnets carry.
-
-That leaves one thing genuinely environment-specific — internal versus
-internet-facing. A bare `LoadBalancer` provisions internet-facing, which a
-regulated account will usually refuse, and should. That belongs in a
-per-environment overlay applied on top, never inlined into this artifact.
-
-**MetalLB only — no ingress-nginx.** Its controller supports a narrow window of
-Kubernetes versions, so depending on it constrains which k8s a rig can be built
-with. That undercuts running trailing-edge control planes to model a legacy
-estate, which is the reason `versions.env` pins v1_33..v1_36. MetalLB carries no
-such constraint, so reachability costs nothing in version coverage.
-
-## Several rigs, one cluster
-
-Identity follows the **folder name**, the same rule rig uses for cluster
-identity. The namespace is the folder; resource names are generic, and names only
-have to be unique within a namespace.
-
-```bash
-cp -r sample-rig corporate-rig
-cd corporate-rig && make up      # its own namespace, its own address
-```
-
-No edits, no collisions, both in the same local cluster. `make list` shows them
-together. `make down` removes only this one — siblings, MetalLB and the cluster
-are left alone.
-
-Client rigs are gitignored (`*-rig/`, with `sample-rig/` the deliberate
-exception): a rig's k8s files spell out a real architecture, and that is exactly
-what must not land in this repo.
-
-## Staging workstations
-
-`ctrl/manifest.py` is stdlib-only on purpose: it runs on a bare machine before
-anything is installed. The toolchain itself is rig's job — `make deps` installs
-the pinned kind and tilt binaries, which is what makes a staging AWS workspace
-reachable from the same commands as a laptop.
-
-## Layout
-
-```
-sample-rig/
-├── Makefile            # thin — one target per ctrl/ script
-├── bundle.json         # what shipped; the UI renders THIS
-├── rig-ui/             # the vite app — optional, copied into a rig to enable it
-│   ├── k8s.yaml        #   how to plug it in: one Pod, one Service
-│   ├── index.html
-│   ├── package.json
-│   ├── vite.config.js
-│   └── src/{main.js,style.css}
-├── ctrl/
-│   ├── manifest.py     # emits the artifact
-│   └── bundle.sh       # generate / deploy / inspect
-└── generated/          # the artifact — committed, this is the deliverable
-```
-
-Editing `bundle.json` or anything in `rig-ui/` means re-running `make manifest`.
-The ConfigMap carries a checksum of everything embedded, so a stale deployment is
-visible rather than silent.
-
-## Not built, but not foreclosed
-
-Everything derives from `bundle.json` plus a target namespace. A Pulumi or
-Terraform emitter would sit beside `ctrl/manifest.py` consuming the same inputs;
-nothing above it assumes the artifact is YAML.
-
-Licence terms for the compiled UI component belong in the soleprint-generated
-bundle, not here — this sample carries no proprietary component.

rig/sample-rig/bundle.json

@@ -1,51 +0,0 @@
-{
-  "_comment": "What this bundle contains. Single source of truth — the landing page renders THIS file, so adding an entry here is the only edit needed. Deliberately FLAT: standalone tools and rigs, with none of soleprint's internal hierarchy (no artery/atlas/station layering). Nothing here is sensitive; the real architecture connects separately.",
-  "bundle": {
-    "name": "sample-rig",
-    "description": "Non-sensitive sample bundle. Proves the kind install works and shows what ships.",
-    "sensitive": false
-  },
-  "tools": [
-    {
-      "name": "modelgen",
-      "summary": "Generate models from config",
-      "standalone": true
-    },
-    {
-      "name": "datagen",
-      "summary": "Generate test data from rig-owned generators",
-      "standalone": true
-    },
-    {
-      "name": "graphgen",
-      "summary": "Generate navigable model graphs",
-      "standalone": true
-    },
-    {
-      "name": "tester",
-      "summary": "HTTP contract test runner — one suite, any environment",
-      "standalone": true
-    },
-    {
-      "name": "databrowse",
-      "summary": "SQL data browser",
-      "standalone": true
-    },
-    {
-      "name": "sbwrapper",
-      "summary": "Sandbox wrapper",
-      "standalone": true
-    }
-  ],
-  "rigs": [
-    {
-      "name": "sample-rig",
-      "summary": "This bundle — a minimal, copyable environment",
-      "active": true
-    }
-  ],
-  "next": [
-    "Point MANIFESTS_DIR at the real manifests to connect the actual architecture.",
-    "Real k8s files are versioned separately and are not part of this bundle."
-  ]
-}

rig/sample-rig/cluster.mock.json

@@ -1,40 +0,0 @@
-{
-  "_comment": "MOCKED cluster state. Nothing here is read from a live cluster — it exists so the UI can be shown when there is no cluster at all (a locked-down machine, a laptop with no memory to spare, a demo where kind will not start). The page labels it as mocked; a demo that looks live but is not is worse than one that says so. When a real cluster is present the same shapes come from kubectl.",
-
-  "mocked": true,
-
-  "cluster": {
-    "name": "sample-rig",
-    "context": "kind-sample-rig",
-    "provider": "kind",
-    "profile": "minimal",
-    "k8s": "v1.36.1",
-    "nodes": 1
-  },
-
-  "workloads": [
-    {
-      "name": "rig-ui",
-      "summary": "Pod · node:22-alpine · vite on :5173",
-      "state": "Running"
-    },
-    {
-      "name": "metallb-system/controller",
-      "summary": "Deployment · assigns LoadBalancer addresses",
-      "state": "Running"
-    },
-    {
-      "name": "metallb-system/speaker",
-      "summary": "DaemonSet · answers ARP in layer 2 mode",
-      "state": "Running"
-    }
-  ],
-
-  "services": [
-    {
-      "name": "rig-ui",
-      "summary": "LoadBalancer · 80 -> 5173 · no annotations, so it resolves on kind and on EKS alike",
-      "state": "172.18.255.200"
-    }
-  ]
-}

rig/sample-rig/ctrl/bundle.sh

@@ -1,223 +0,0 @@
-#!/usr/bin/env bash
-# The rig bundle: generate it, deploy it, tear it down, find it.
-#
-# Usage: bundle.sh manifest | up | down | status | url | list | dev
-#
-# What `up` proves, in order: kind installed and a cluster exists, MetalLB can
-# hand out an address, a Service of type LoadBalancer actually resolves, and a
-# pod serves the bundle listing. If all of that works the installation is sound,
-# and the only thing missing is the real architecture.
-#
-# ONE ARTIFACT
-# `up` applies generated/<slug>.yaml — the same self-contained file you would
-# hand to an external cluster. There is no separate local path, so what works
-# here cannot quietly differ from the master deployment applied elsewhere.
-#
-# ONE CLUSTER, SEVERAL RIGS
-# Identity follows the FOLDER NAME, exactly as rig's cluster identity does. This
-# directory deploys into a namespace named after itself, so copying it to
-# corporate-rig/ yields a second rig in the SAME local cluster with no edits and
-# no collisions — different namespace, its own MetalLB address. `list` shows all
-# of them. The cluster itself is rig's business; this only ever owns a namespace.
-#
-# MetalLB is installed by calling rig's own addon script rather than
-# reimplementing it — deriving the pool from the kind Docker network is the
-# fiddly part and there should be exactly one copy of it.
-set -euo pipefail
-cd "$(dirname "$0")/.."
-
-BUNDLE_ROOT="$(pwd)"
-RIG_CTRL="$(cd .. && pwd)/ctrl"
-
-# The containing folder's name, reduced to a DNS label (same rule as rig's
-# default_cluster_name and ctrl/manifest.py, so all three agree on the slug).
-slug() {
-    local n
-    n=$(basename "$BUNDLE_ROOT")
-    n=$(echo "$n" | tr '[:upper:]' '[:lower:]' | tr -c 'a-z0-9-' '-')
-    n=$(echo "$n" | sed 's/^-*//; s/-*$//')
-    echo "${n:-rig-bundle}"
-}
-NS="$(slug)"
-ARTIFACT="generated/${NS}.yaml"
-
-# Resolved lazily, not at load time: `manifest` and `dev` deliberately work
-# with no cluster and no kubectl at all, and a top-level check would break that.
-#
-# Follows whatever context rig's cluster.sh selected, so this bundle works in a
-# copied-and-renamed environment without being told which cluster it is in.
-init_kube() {
-    KUBECONTEXT="${KUBECONTEXT:-$(kubectl config current-context 2>/dev/null || true)}"
-    if [ -z "$KUBECONTEXT" ]; then
-        echo "no kubectl context — bring a cluster up first:  (cd .. && make cluster up)" >&2
-        exit 1
-    fi
-    KCTX="kubectl --context ${KUBECONTEXT}"
-    K="kubectl --context ${KUBECONTEXT} --namespace ${NS}"
-}
-
-require_cluster() {
-    if ! $KCTX cluster-info >/dev/null 2>&1; then
-        echo "context '$KUBECONTEXT' does not reach a cluster" >&2
-        echo "bring one up:  (cd .. && make cluster up)" >&2
-        exit 1
-    fi
-}
-
-ensure_metallb() {
-    if $KCTX get deployment -n metallb-system controller >/dev/null 2>&1; then
-        echo "metallb: present"
-        return 0
-    fi
-
-    # Only kind needs it. On a real cluster the cloud load balancer answers a
-    # `type: LoadBalancer` Service, and installing MetalLB there would be wrong.
-    case "$KUBECONTEXT" in
-        kind-*) ;;
-        *)
-            echo "metallb: skipped — '$KUBECONTEXT' is not a kind context"
-            echo "  (a cloud load balancer answers LoadBalancer services there)"
-            return 0
-            ;;
-    esac
-
-    if [ ! -f "$RIG_CTRL/addons/metallb.sh" ]; then
-        echo "metallb is not installed and rig's addon script was not found at" >&2
-        echo "  $RIG_CTRL/addons/metallb.sh" >&2
-        echo "a Service of type LoadBalancer will sit at <pending> without it." >&2
-        exit 1
-    fi
-
-    # rig's addons derive their target cluster from RIG'S OWN folder name via
-    # load_config, so left alone this bundle would install into `kind-rig` —
-    # a cluster that need not exist — while deploying everything else into the
-    # context actually selected. CLUSTER is in load_config's overridable set,
-    # so passing it here points the addon at the same cluster we are using.
-    local target="${KUBECONTEXT#kind-}"
-    echo "metallb: installing via rig's addon into '$target'"
-    CLUSTER="$target" bash "$RIG_CTRL/addons/metallb.sh"
-}
-
-# Regenerate the artifact. No cluster and no kubectl required — this is the step
-# a staging workstation runs before anything is installed.
-manifest() {
-    mkdir -p generated
-    python3 ctrl/manifest.py "$NS" > "$ARTIFACT"
-    echo "wrote $ARTIFACT ($(wc -l < "$ARTIFACT") lines)"
-    echo "  applies as-is anywhere:  kubectl apply -f ${BUNDLE_ROOT}/${ARTIFACT}"
-}
-
-up() {
-    manifest
-    init_kube
-    require_cluster
-    ensure_metallb
-
-    echo
-    echo "applying '${NS}' to context '${KUBECONTEXT}'"
-    $KCTX apply -f "$ARTIFACT"
-
-    # `rollout status` does not work on a bare Pod — it only understands
-    # Deployments, StatefulSets and DaemonSets. Wait on the condition instead.
-    # This is the slow step: the container npm-installs before vite serves.
-    echo "waiting for the pod to be ready (npm install runs first)..."
-    $K wait --for=condition=Ready pod/rig-ui --timeout=300s
-    echo
-    url
-}
-
-down() {
-    init_kube
-    # Delete the namespace and everything in it goes with it. Scoped to THIS
-    # rig — a sibling rig in the same cluster is untouched.
-    $KCTX delete namespace "$NS" --ignore-not-found
-    echo "'${NS}' removed (cluster, metallb and any sibling rig are left alone)"
-}
-
-status() {
-    init_kube
-    require_cluster
-    if ! $KCTX get namespace "$NS" >/dev/null 2>&1; then
-        echo "'${NS}' is not deployed — run: make up"
-        return 0
-    fi
-    $K get pod,svc,configmap -o wide
-}
-
-# Every rig in this cluster, not just this one — the point of the namespace
-# split is that several coexist, so there has to be a way to see them together.
-list() {
-    init_kube
-    require_cluster
-    local names
-    names=$($KCTX get namespace -l rig.bundle/name \
-            -o jsonpath='{.items[*].metadata.name}' 2>/dev/null || true)
-    if [ -z "$names" ]; then
-        echo "no rigs deployed in context '${KUBECONTEXT}'"
-        return 0
-    fi
-    printf "%-20s %-16s %s\n" RIG ADDRESS ""
-    local n ip
-    for n in $names; do
-        ip=$($KCTX -n "$n" get svc rig-ui \
-             -o jsonpath='{.status.loadBalancer.ingress[0].ip}' 2>/dev/null || true)
-        printf "%-20s %-16s %s\n" "$n" "${ip:-<pending>}" \
-            "$([ "$n" = "$NS" ] && echo '<- this one')"
-    done
-}
-
-# The address MetalLB (or a cloud load balancer) assigned. <pending> here is the
-# classic silent failure: everything reports healthy and nothing is reachable.
-url() {
-    init_kube
-    local ip
-    ip=$($K get svc rig-ui \
-         -o jsonpath='{.status.loadBalancer.ingress[0].ip}' 2>/dev/null || true)
-    if [ -z "$ip" ]; then
-        ip=$($K get svc rig-ui \
-             -o jsonpath='{.status.loadBalancer.ingress[0].hostname}' 2>/dev/null || true)
-    fi
-    if [ -z "$ip" ]; then
-        echo "no external address yet — nothing has assigned one."
-        echo "on kind:  kubectl --context $KUBECONTEXT -n metallb-system get pods"
-        return 1
-    fi
-    echo "IT WORKS -> http://${ip}/"
-    echo "  bundle    http://${ip}/bundle.json"
-}
-
-# Run the UI locally with no cluster at all — the fast way to iterate on
-# bundle.json. Same vite command the pod runs, so what you see here is what
-# gets served there.
-dev() {
-    if ! command -v npm >/dev/null 2>&1; then
-        echo "npm not found — the UI needs node locally for this." >&2
-        echo "(in-cluster it runs on the node:22-alpine image instead)" >&2
-        exit 1
-    fi
-    # bundle.json lives one level up so it stays the rig's data rather than the
-    # app's; vite serves public/ at the root, which is where the app fetches it.
-    mkdir -p rig-ui/public
-    cp bundle.json rig-ui/public/bundle.json
-
-    # The mocked cluster is a DEMO asset and is deliberately not embedded in the
-    # deployed artifact — on a real rig the UI would then show canned values
-    # beside a live cluster, which is precisely the lie its banner warns about.
-    # It is served here, and in the static build for the public UI-only page.
-    cp cluster.mock.json rig-ui/public/cluster.mock.json
-
-    cd rig-ui
-    [ -d node_modules ] || npm install --no-audit --no-fund
-    VITE_RIG_NAME="$NS" npm run dev
-}
-
-case "${1:-status}" in
-    manifest) manifest ;;
-    up)       up ;;
-    down)     down ;;
-    status)   status ;;
-    url)      url ;;
-    list)     list ;;
-    dev)      dev ;;
-    *)        echo "usage: $0 [manifest|up|down|status|url|list|dev]" >&2; exit 1 ;;
-esac

rig/sample-rig/ctrl/manifest.py

@@ -1,141 +0,0 @@
-#!/usr/bin/env python3
-"""Emit the complete, self-contained deployment for this rig.
-
-    python3 ctrl/manifest.py [namespace] > generated/<slug>.yaml
-
-The output is the ARTIFACT. It carries everything — namespace, the vite app and
-bundle.json embedded in a ConfigMap, the Pod and the Service — so it applies
-unmodified to any cluster:
-
-    kubectl apply -f generated/sample-rig.yaml
-
-On kind, MetalLB answers the `type: LoadBalancer` Service. On a real external
-cluster the cloud load balancer does. Same file, no edits, no branch — which is
-the point: what runs locally is byte-identical to the deployment applied
-elsewhere, so local success actually means something.
-
-`ctrl/bundle.sh up` applies this same generated output rather than a separate
-code path, so the local convenience wrapper can never drift from the artifact.
-
-Stdlib only, deliberately: this must run on a bare staging workstation before
-anything is installed, so it cannot depend on PyYAML or a template engine.
-
-Open seam — not built: everything here derives from bundle.json plus a target
-namespace. A Pulumi or Terraform emitter would sit beside this file consuming the
-same inputs; nothing above it assumes the artifact is YAML.
-"""
-
-import json
-import re
-import sys
-import zlib
-from pathlib import Path
-
-ROOT = Path(__file__).resolve().parent.parent
-UI = ROOT / "rig-ui"
-
-# Files embedded into the ConfigMap, mounted read-only at /src in the pod and
-# copied into vite's layout at start (see rig-ui/k8s.yaml). Flat on purpose:
-# ConfigMap keys cannot contain '/'.
-EMBEDDED = {
-    "bundle.json": ROOT / "bundle.json",
-    "package.json": UI / "package.json",
-    "vite.config.js": UI / "vite.config.js",
-    "index.html": UI / "index.html",
-    "main.js": UI / "src" / "main.js",
-    "style.css": UI / "src" / "style.css",
-}
-
-
-def slug(name: str) -> str:
-    """Reduce a folder name to a DNS label, matching ctrl/bundle.sh's rule."""
-    out = re.sub(r"[^a-z0-9-]", "-", name.lower()).strip("-")
-    return out or "rig-bundle"
-
-
-def block(text: str, indent: int) -> str:
-    """Indent a file's contents for a YAML literal block scalar.
-
-    Blank lines are emitted truly empty rather than as whitespace: trailing
-    spaces on an otherwise blank line are legal YAML but show up as diff noise
-    in a committed artifact.
-    """
-    pad = " " * indent
-    return "\n".join(pad + line if line.strip() else "" for line in text.splitlines())
-
-
-def checksum(parts: list[str]) -> str:
-    """Stable content hash of everything embedded, stamped as a label.
-
-    A mounted ConfigMap updates in place without restarting anything, so without
-    a visible change nothing signals that the pod is serving stale content.
-    """
-    return str(zlib.crc32("".join(parts).encode()) & 0xFFFFFFFF)
-
-
-def build(namespace: str) -> str:
-    contents = {}
-    for key, path in EMBEDDED.items():
-        if not path.exists():
-            sys.exit(f"missing input: {path}")
-        contents[key] = path.read_text()
-
-    # Fail loudly here rather than shipping an artifact that renders an error.
-    try:
-        json.loads(contents["bundle.json"])
-    except json.JSONDecodeError as exc:
-        sys.exit(f"bundle.json is not valid JSON: {exc}")
-
-    app = (UI / "k8s.yaml").read_text()
-    app = app.replace("__RIG_NAME__", namespace)
-
-    data = "\n".join(
-        f"  {key}: |\n{block(text, 4)}" for key, text in sorted(contents.items())
-    )
-
-    return f"""# GENERATED by ctrl/manifest.py — do not edit.
-# Regenerate with:  make manifest
-#
-# Self-contained: applies as-is to any cluster, local kind or external.
-#   kubectl apply -f this-file.yaml
-#
-# Namespace carries the identity, so several rigs coexist in one cluster.
-apiVersion: v1
-kind: Namespace
-metadata:
-  name: {namespace}
-  labels:
-    rig.bundle/name: {namespace}
----
-apiVersion: v1
-kind: ConfigMap
-metadata:
-  name: rig-ui
-  namespace: {namespace}
-  labels:
-    rig.bundle/checksum: "{checksum(list(contents.values()))}"
-data:
-{data}
----
-{_namespaced(app, namespace)}
-"""
-
-
-def _namespaced(doc: str, namespace: str) -> str:
-    """Add `namespace:` to each resource so the artifact applies without -n.
-
-    rig-ui/k8s.yaml omits it on purpose — applied by hand it should land in
-    whatever namespace you choose. Pinning it belongs to the generated artifact,
-    which has to be self-contained.
-    """
-    return re.sub(
-        r"^(metadata:\n(?:[ \t]+.*\n)*?)([ \t]+)(name: rig-ui)$",
-        lambda m: f"{m.group(1)}{m.group(2)}{m.group(3)}\n{m.group(2)}namespace: {namespace}",
-        doc.strip(),
-        flags=re.MULTILINE,
-    )
-
-
-if __name__ == "__main__":
-    target = sys.argv[1] if len(sys.argv) > 1 else slug(ROOT.name)
-    sys.stdout.write(build(target))

rig/sample-rig/generated/sample-rig.yaml

@@ -1,406 +0,0 @@
-# GENERATED by ctrl/manifest.py — do not edit.
-# Regenerate with:  make manifest
-#
-# Self-contained: applies as-is to any cluster, local kind or external.
-#   kubectl apply -f this-file.yaml
-#
-# Namespace carries the identity, so several rigs coexist in one cluster.
-apiVersion: v1
-kind: Namespace
-metadata:
-  name: sample-rig
-  labels:
-    rig.bundle/name: sample-rig
----
-apiVersion: v1
-kind: ConfigMap
-metadata:
-  name: rig-ui
-  namespace: sample-rig
-  labels:
-    rig.bundle/checksum: "2074194964"
-data:
-  bundle.json: |
-    {
-      "_comment": "What this bundle contains. Single source of truth — the landing page renders THIS file, so adding an entry here is the only edit needed. Deliberately FLAT: standalone tools and rigs, with none of soleprint's internal hierarchy (no artery/atlas/station layering). Nothing here is sensitive; the real architecture connects separately.",
-      "bundle": {
-        "name": "sample-rig",
-        "description": "Non-sensitive sample bundle. Proves the kind install works and shows what ships.",
-        "sensitive": false
-      },
-      "tools": [
-        {
-          "name": "modelgen",
-          "summary": "Generate models from config",
-          "standalone": true
-        },
-        {
-          "name": "datagen",
-          "summary": "Generate test data from rig-owned generators",
-          "standalone": true
-        },
-        {
-          "name": "graphgen",
-          "summary": "Generate navigable model graphs",
-          "standalone": true
-        },
-        {
-          "name": "tester",
-          "summary": "HTTP contract test runner — one suite, any environment",
-          "standalone": true
-        },
-        {
-          "name": "databrowse",
-          "summary": "SQL data browser",
-          "standalone": true
-        },
-        {
-          "name": "sbwrapper",
-          "summary": "Sandbox wrapper",
-          "standalone": true
-        }
-      ],
-      "rigs": [
-        {
-          "name": "sample-rig",
-          "summary": "This bundle — a minimal, copyable environment",
-          "active": true
-        }
-      ],
-      "next": [
-        "Point MANIFESTS_DIR at the real manifests to connect the actual architecture.",
-        "Real k8s files are versioned separately and are not part of this bundle."
-      ]
-    }
-  index.html: |
-    <!doctype html>
-    <html lang="en">
-      <head>
-        <meta charset="utf-8" />
-        <meta name="viewport" content="width=device-width, initial-scale=1" />
-        <title>IT WORKS</title>
-      </head>
-      <body>
-        <main id="app"></main>
-        <script type="module" src="/src/main.js"></script>
-      </body>
-    </html>
-  main.js: |
-    import "./style.css";
-
-    /* The IT WORKS page: renders bundle.json as the list of what shipped.
-     *
-     * Plain vite, no framework — this is a complement to the rig, not part of it,
-     * and it should stay small enough that nobody has to adopt a stack to read it.
-     *
-     * bundle.json is fetched at runtime rather than imported, so the same built app
-     * serves whatever rig it was copied into. Editing the ConfigMap changes the page
-     * without rebuilding.
-     *
-     * Styling is a handful of rules on purpose. The real visual identity lives in
-     * the soleprint UI package; nothing here should grow into a theme.
-     */
-
-    const esc = (s) =>
-      String(s).replace(/&/g, "&amp;").replace(/</g, "&lt;").replace(/>/g, "&gt;");
-
-    const tag = (text, on = false) =>
-      `<span class="tag${on ? " on" : ""}">${esc(text)}</span>`;
-
-    function items(list, activeKey) {
-      if (!list?.length) return `<li><span class="summary">nothing listed</span></li>`;
-      return list
-        .map((it) => {
-          const tags = [
-            it.standalone ? tag("standalone") : "",
-            it.state ? tag(it.state) : "",
-            activeKey && it[activeKey] ? tag("active", true) : "",
-          ].join("");
-          return `<li><span class="name">${esc(it.name ?? "?")}</span>
-            <span class="summary">${esc(it.summary ?? "")}</span>${tags}</li>`;
-        })
-        .join("");
-    }
-
-    /* Cluster state, when there is any to show.
-     *
-     * Fetched separately and allowed to fail: the bundle listing is the point, and a
-     * rig with no cluster reachable is a normal state, not an error. Renders nothing
-     * at all when absent.
-     *
-     * When the payload says `mocked`, say so loudly. This exists to demo the UI on a
-     * machine where kind will not run — and a demo that looks live but is not is
-     * worse than one that admits it. */
-    function clusterSection(c) {
-      if (!c) return "";
-      const m = c.cluster ?? {};
-      const banner = c.mocked
-        ? `<p class="mock">mocked — no cluster was queried; these are canned values</p>`
-        : "";
-      const meta = [m.context, m.k8s, m.profile ? `profile ${m.profile}` : "",
-                    m.nodes ? `${m.nodes} node${m.nodes > 1 ? "s" : ""}` : ""]
-        .filter(Boolean).join(" · ");
-
-      return `
-        <h2>Cluster${c.mocked ? " (mocked)" : ""}</h2>
-        ${banner}
-        ${meta ? `<p class="sub">${esc(meta)}</p>` : ""}
-        <ul>${items(c.workloads)}</ul>
-        <h2>Services (${c.services?.length ?? 0})</h2>
-        <ul>${items(c.services)}</ul>`;
-    }
-
-    function render(b, name, cluster) {
-      const meta = b.bundle ?? {};
-      const next = (b.next ?? []).map((n) => `<li>${esc(n)}</li>`).join("");
-      return `
-        <h1><span class="ok">IT WORKS</span> — ${esc(name || meta.name || "rig")}</h1>
-        <p class="sub">${esc(meta.description ?? "")}</p>
-
-        <h2>Tools (${b.tools?.length ?? 0})</h2>
-        <ul>${items(b.tools)}</ul>
-
-        <h2>Rigs (${b.rigs?.length ?? 0})</h2>
-        <ul>${items(b.rigs, "active")}</ul>
-
-        ${clusterSection(cluster)}
-
-        ${next ? `<div class="next"><ul>${next}</ul></div>` : ""}`;
-    }
-
-    const app = document.getElementById("app");
-
-    const json = (path, required) =>
-      fetch(path).then((r) => {
-        if (r.ok) return r.json();
-        if (required) throw new Error(`${path} -> HTTP ${r.status}`);
-        return null;            // optional: absent is a normal state, not an error
-      }).catch((err) => {
-        if (required) throw err;
-        return null;
-      });
-
-    Promise.all([json("/bundle.json", true), json("/cluster.mock.json", false)])
-      // RIG_NAME is injected by vite from the pod env, so two rigs sharing a
-      // cluster are distinguishable even if a copied bundle.json kept its old name.
-      .then(([b, cluster]) => {
-        app.innerHTML = render(b, import.meta.env.VITE_RIG_NAME, cluster);
-      })
-      .catch((err) => {
-        app.innerHTML = `<h1 class="err">bundle unavailable</h1>
-          <p class="sub">${esc(err.message)}</p>`;
-      });
-  package.json: |
-    {
-      "name": "rig-ui",
-      "private": true,
-      "version": "0.1.0",
-      "type": "module",
-      "scripts": {
-        "dev": "vite --host 0.0.0.0 --port 5173",
-        "build": "vite build",
-        "preview": "vite preview --host 0.0.0.0 --port 5173"
-      },
-      "devDependencies": {
-        "vite": "^6"
-      }
-    }
-  style.css: |
-    /* Minimal, self-contained. The real visual identity ships with the soleprint UI
-       package, which is a separate artifact — nothing here should grow into a theme. */
-
-    body {
-      margin: 0;
-      padding: 2.5rem 1.5rem;
-      background: #0d0d0f;
-      color: #e8e8f0;
-      font: 14px/1.6 ui-monospace, "JetBrains Mono", Menlo, monospace;
-    }
-    main { max-width: 52rem; margin: 0 auto; }
-
-    h1 { margin: 0; font-size: 1.6rem; letter-spacing: 0.02em; }
-    h1 .ok { color: #3ecf8e; }
-    h1.err { color: #f06565; }
-    .sub { color: #8888a0; margin: 0.35rem 0 2.25rem; }
-
-    h2 {
-      font-size: 0.75rem;
-      text-transform: uppercase;
-      letter-spacing: 0.12em;
-      color: #8888a0;
-      margin: 2rem 0 0.75rem;
-      font-weight: 600;
-    }
-
-    ul { list-style: none; margin: 0; padding: 0; }
-    li {
-      display: flex;
-      gap: 0.75rem;
-      align-items: baseline;
-      padding: 0.5rem 0.75rem;
-      border: 1px solid #2e2e38;
-      border-radius: 6px;
-      margin-bottom: 0.4rem;
-      background: #16161a;
-    }
-    .name { font-weight: 600; min-width: 9rem; }
-    .summary { color: #8888a0; flex: 1; }
-
-    .tag {
-      font-size: 0.7rem;
-      padding: 0.1rem 0.45rem;
-      border-radius: 3px;
-      background: #26262f;
-      color: #8888a0;
-      white-space: nowrap;
-    }
-    .tag.on { background: #3ecf8e; color: #0d0d0f; }
-
-    .next {
-      color: #555568;
-      font-size: 0.8rem;
-      margin-top: 2.5rem;
-      border-top: 1px solid #2e2e38;
-      padding-top: 1rem;
-    }
-    .next li {
-      display: list-item;
-      border: 0;
-      background: none;
-      padding: 0.15rem 0;
-      margin: 0 0 0 1.1rem;
-      list-style: disc;
-    }
-
-    /* Mocked-data banner. Deliberately loud: this only appears when the cluster
-       payload is canned, and a demo that looks live but is not is worse than one
-       that says so. */
-    .mock {
-      margin: 0 0 0.75rem;
-      padding: 0.4rem 0.75rem;
-      border: 1px dashed #f5a623;
-      border-radius: 6px;
-      color: #f5a623;
-      font-size: 0.8rem;
-    }
-  vite.config.js: |
-    import { defineConfig } from "vite";
-
-    /* Serves on 0.0.0.0 so the pod is reachable through the Service, and allows any
-     * Host header because the address is assigned at runtime (MetalLB locally, a
-     * cloud load balancer on EKS) and is never known at build time. */
-    export default defineConfig({
-      server: { host: "0.0.0.0", port: 5173, strictPort: true, allowedHosts: true },
-      preview: { host: "0.0.0.0", port: 5173, strictPort: true, allowedHosts: true },
-    });
----
-# How to plug the UI into whatever k8s you generated. THIS IS THE WHOLE THING:
-# one Pod running the vite app, one Service to reach it.
-#
-# Optional by design. The UI complements a rig; it is not part of the end
-# product, and a rig is complete and useful without it. Apply this only when you
-# want the listing:
-#
-#   kubectl apply -n <your-namespace> -f rig-ui/k8s.yaml
-#
-# A bare Pod, not a Deployment — this is a dev-loop convenience, not a workload
-# to keep alive. If it dies you re-apply it; nothing depends on it staying up.
-#
-# The app and bundle.json arrive as a ConfigMap named `rig-ui`, which
-# ctrl/manifest.py generates from the folder. Nothing is baked into an image, so
-# editing bundle.json and re-applying is the whole update cycle.
-apiVersion: v1
-kind: Pod
-metadata:
-  name: rig-ui
-  namespace: sample-rig
-  labels:
-    app: rig-ui
-spec:
-  containers:
-    - name: vite
-      image: node:22-alpine
-      workingDir: /app
-      # npm install at start: no image to build and no registry to publish to,
-      # which is the point of a minimal plug-in. It needs egress to a registry —
-      # on a locked-down cluster point npm at the internal one, or bake an image
-      # instead. Nothing else here changes if you do.
-      command: ["sh", "-c"]
-      # A ConfigMap mounts flat (keys cannot contain '/'), so the files are
-      # placed into vite's expected layout here. bundle.json goes to public/
-      # because that is what vite serves at /bundle.json, which is where the
-      # app fetches it.
-      args:
-        - |
-          mkdir -p /app/src /app/public &&
-          cp /src/package.json /src/vite.config.js /src/index.html /app/ &&
-          cp /src/main.js /src/style.css /app/src/ &&
-          cp /src/bundle.json /app/public/ &&
-          npm install --no-audit --no-fund &&
-          npm run dev
-      env:
-        # Rendered in the heading so two rigs sharing a cluster stay
-        # distinguishable. Set from the namespace by ctrl/manifest.py.
-        - name: VITE_RIG_NAME
-          value: sample-rig
-      ports:
-        - name: http
-          containerPort: 5173
-      volumeMounts:
-        # /src is read-only from the ConfigMap; the app is copied to a writable
-        # /app because npm install has to create node_modules.
-        - name: rig-ui
-          mountPath: /src
-        - name: app
-          mountPath: /app
-      readinessProbe:
-        httpGet: { path: /, port: 5173 }
-        # npm install decides how long this takes, and it is the slow part.
-        initialDelaySeconds: 15
-        periodSeconds: 5
-        failureThreshold: 30
-      resources:
-        requests: { memory: 128Mi, cpu: 50m }
-        limits: { memory: 512Mi }
-  volumes:
-    - name: rig-ui
-      configMap:
-        name: rig-ui
-    - name: app
-      emptyDir: {}
----
-apiVersion: v1
-kind: Service
-metadata:
-  name: rig-ui
-  namespace: sample-rig
-  labels:
-    app: rig-ui
-  # No annotations, deliberately — see k8s/app.yaml. The target is EKS but this
-  # stays VPC-agnostic: no subnets, no security groups, no -scheme, no -type.
-  # A bare LoadBalancer is what lets one manifest work on kind and on EKS.
-spec:
-  type: LoadBalancer
-  selector:
-    app: rig-ui
-  ports:
-    - name: http
-      port: 80
-      targetPort: 5173
-      protocol: TCP
-      # Pinned, because a LoadBalancer Service also allocates a NodePort and
-      # this is the only address that works everywhere.
-      #
-      # On WSL the MetalLB address is on a docker bridge INSIDE the Linux VM,
-      # and Windows has no route to it — the page looks broken while the
-      # cluster is perfectly healthy. 30080 is what rig's `hostport` ingress
-      # mode publishes to the host, so this is reachable at
-      # localhost:$HTTP_PORT from a Windows browser with nothing configured.
-      #
-      # Costs nothing elsewhere: MetalLB still assigns an external IP on Linux,
-      # and on EKS the load balancer targets this NodePort anyway. One Service,
-      # no per-environment branch.
-      #
-      # A pinned NodePort is cluster-unique, so two rigs must live in separate
-      # clusters — which is how they are run anyway.
-      nodePort: 30080

rig/sample-rig/rig-ui/.gitignore

@@ -1,3 +0,0 @@
-node_modules/
-public/
-dist/

rig/sample-rig/rig-ui/index.html

@@ -1,12 +0,0 @@
-<!doctype html>
-<html lang="en">
-  <head>
-    <meta charset="utf-8" />
-    <meta name="viewport" content="width=device-width, initial-scale=1" />
-    <title>IT WORKS</title>
-  </head>
-  <body>
-    <main id="app"></main>
-    <script type="module" src="/src/main.js"></script>
-  </body>
-</html>

rig/sample-rig/rig-ui/k8s.yaml

@@ -1,108 +0,0 @@
-# How to plug the UI into whatever k8s you generated. THIS IS THE WHOLE THING:
-# one Pod running the vite app, one Service to reach it.
-#
-# Optional by design. The UI complements a rig; it is not part of the end
-# product, and a rig is complete and useful without it. Apply this only when you
-# want the listing:
-#
-#   kubectl apply -n <your-namespace> -f rig-ui/k8s.yaml
-#
-# A bare Pod, not a Deployment — this is a dev-loop convenience, not a workload
-# to keep alive. If it dies you re-apply it; nothing depends on it staying up.
-#
-# The app and bundle.json arrive as a ConfigMap named `rig-ui`, which
-# ctrl/manifest.py generates from the folder. Nothing is baked into an image, so
-# editing bundle.json and re-applying is the whole update cycle.
-apiVersion: v1
-kind: Pod
-metadata:
-  name: rig-ui
-  labels:
-    app: rig-ui
-spec:
-  containers:
-    - name: vite
-      image: node:22-alpine
-      workingDir: /app
-      # npm install at start: no image to build and no registry to publish to,
-      # which is the point of a minimal plug-in. It needs egress to a registry —
-      # on a locked-down cluster point npm at the internal one, or bake an image
-      # instead. Nothing else here changes if you do.
-      command: ["sh", "-c"]
-      # A ConfigMap mounts flat (keys cannot contain '/'), so the files are
-      # placed into vite's expected layout here. bundle.json goes to public/
-      # because that is what vite serves at /bundle.json, which is where the
-      # app fetches it.
-      args:
-        - |
-          mkdir -p /app/src /app/public &&
-          cp /src/package.json /src/vite.config.js /src/index.html /app/ &&
-          cp /src/main.js /src/style.css /app/src/ &&
-          cp /src/bundle.json /app/public/ &&
-          npm install --no-audit --no-fund &&
-          npm run dev
-      env:
-        # Rendered in the heading so two rigs sharing a cluster stay
-        # distinguishable. Set from the namespace by ctrl/manifest.py.
-        - name: VITE_RIG_NAME
-          value: __RIG_NAME__
-      ports:
-        - name: http
-          containerPort: 5173
-      volumeMounts:
-        # /src is read-only from the ConfigMap; the app is copied to a writable
-        # /app because npm install has to create node_modules.
-        - name: rig-ui
-          mountPath: /src
-        - name: app
-          mountPath: /app
-      readinessProbe:
-        httpGet: { path: /, port: 5173 }
-        # npm install decides how long this takes, and it is the slow part.
-        initialDelaySeconds: 15
-        periodSeconds: 5
-        failureThreshold: 30
-      resources:
-        requests: { memory: 128Mi, cpu: 50m }
-        limits: { memory: 512Mi }
-  volumes:
-    - name: rig-ui
-      configMap:
-        name: rig-ui
-    - name: app
-      emptyDir: {}
----
-apiVersion: v1
-kind: Service
-metadata:
-  name: rig-ui
-  labels:
-    app: rig-ui
-  # No annotations, deliberately — see k8s/app.yaml. The target is EKS but this
-  # stays VPC-agnostic: no subnets, no security groups, no -scheme, no -type.
-  # A bare LoadBalancer is what lets one manifest work on kind and on EKS.
-spec:
-  type: LoadBalancer
-  selector:
-    app: rig-ui
-  ports:
-    - name: http
-      port: 80
-      targetPort: 5173
-      protocol: TCP
-      # Pinned, because a LoadBalancer Service also allocates a NodePort and
-      # this is the only address that works everywhere.
-      #
-      # On WSL the MetalLB address is on a docker bridge INSIDE the Linux VM,
-      # and Windows has no route to it — the page looks broken while the
-      # cluster is perfectly healthy. 30080 is what rig's `hostport` ingress
-      # mode publishes to the host, so this is reachable at
-      # localhost:$HTTP_PORT from a Windows browser with nothing configured.
-      #
-      # Costs nothing elsewhere: MetalLB still assigns an external IP on Linux,
-      # and on EKS the load balancer targets this NodePort anyway. One Service,
-      # no per-environment branch.
-      #
-      # A pinned NodePort is cluster-unique, so two rigs must live in separate
-      # clusters — which is how they are run anyway.
-      nodePort: 30080

rig/sample-rig/rig-ui/package.json

@@ -1,14 +0,0 @@
-{
-  "name": "rig-ui",
-  "private": true,
-  "version": "0.1.0",
-  "type": "module",
-  "scripts": {
-    "dev": "vite --host 0.0.0.0 --port 5173",
-    "build": "vite build",
-    "preview": "vite preview --host 0.0.0.0 --port 5173"
-  },
-  "devDependencies": {
-    "vite": "^6"
-  }
-}

rig/sample-rig/rig-ui/src/main.js

@@ -1,136 +0,0 @@
-import "./style.css";
-
-/* The IT WORKS page: renders bundle.json as the list of what shipped.
- *
- * Plain vite, no framework — this is a complement to the rig, not part of it,
- * and it should stay small enough that nobody has to adopt a stack to read it.
- *
- * Laid out like soleprint's templated vein pages, because it does the same job:
- * name each component, list what it exposes, show what comes back. Tool chrome
- * and output are styled apart on purpose (see style.css) — that separation is
- * what tells you whether you are reading the tool or its result.
- *
- * bundle.json is fetched at runtime rather than imported, so the same built app
- * serves whatever rig it was copied into. Editing the ConfigMap changes the page
- * without rebuilding.
- */
-
-const esc = (s) =>
-  String(s).replace(/&/g, "&amp;").replace(/</g, "&lt;").replace(/>/g, "&gt;");
-
-const tag = (text, on = false) =>
-  `<span class="tag${on ? " on" : ""}">${esc(text)}</span>`;
-
-/* Tool chrome: one bordered card per component. */
-function components(list, activeKey) {
-  if (!list?.length)
-    return `<div class="component"><p>nothing listed</p></div>`;
-  return list
-    .map((it) => {
-      const tags = [
-        it.standalone ? tag("standalone") : "",
-        it.state ? tag(it.state) : "",
-        activeKey && it[activeKey] ? tag("active", true) : "",
-      ].join("");
-      return `<div class="component">
-        <h4>${esc(it.name ?? "?")} ${tags}</h4>
-        <p>${esc(it.summary ?? "")}</p>
-      </div>`;
-    })
-    .join("");
-}
-
-/* Endpoint rows: path on the left, what it returns on the right. */
-function endpoints(list) {
-  return list
-    .map(
-      (e) => `<li><code>${esc(e.path)}</code>
-        <span class="desc">${esc(e.desc)}</span></li>`
-    )
-    .join("");
-}
-
-/* Output: what the endpoint above actually returns, so the page demonstrates
-   itself rather than describing what a demonstration would look like. */
-function example(bundle) {
-  const sample = {
-    bundle: bundle.bundle?.name,
-    tools: (bundle.tools ?? []).map((t) => t.name),
-    rigs: (bundle.rigs ?? []).map((r) => r.name),
-  };
-  return `<pre class="output">${esc(JSON.stringify(sample, null, 2))}</pre>`;
-}
-
-/* Cluster state, when there is any to show.
- *
- * Fetched separately and allowed to fail: the bundle listing is the point, and a
- * rig with no cluster reachable is a normal state, not an error. Renders nothing
- * at all when absent. When the payload says `mocked`, say so loudly. */
-function clusterSection(c) {
-  if (!c) return "";
-  const m = c.cluster ?? {};
-  const meta = [m.context, m.k8s, m.profile && `profile ${m.profile}`,
-                m.nodes && `${m.nodes} node${m.nodes > 1 ? "s" : ""}`]
-    .filter(Boolean).join(" · ");
-
-  return `
-    <h2>Cluster${c.mocked ? " (mocked)" : ""}</h2>
-    ${c.mocked ? `<p class="mock">mocked — no cluster was queried; these are canned values</p>` : ""}
-    ${meta ? `<p class="tagline">${esc(meta)}</p>` : ""}
-    <div class="components">${components(c.workloads)}</div>
-
-    <h2>Services</h2>
-    <div class="components">${components(c.services)}</div>`;
-}
-
-function render(b, name, cluster) {
-  const meta = b.bundle ?? {};
-  const next = (b.next ?? []).map((n) => `<li>${esc(n)}</li>`).join("");
-  return `
-    <h1><span class="ok">IT WORKS</span> — ${esc(name || meta.name || "rig")}</h1>
-    <p class="tagline">${esc(meta.description ?? "")}</p>
-
-    <h2>Tools (${b.tools?.length ?? 0})</h2>
-    <div class="components">${components(b.tools)}</div>
-
-    <h2>Rigs (${b.rigs?.length ?? 0})</h2>
-    <div class="components">${components(b.rigs, "active")}</div>
-
-    <h2>Endpoints</h2>
-    <ul class="endpoints">${endpoints([
-      { path: "/", desc: "this page" },
-      { path: "/bundle.json", desc: "the manifest it renders" },
-    ])}</ul>
-
-    <h2>Example — GET /bundle.json</h2>
-    ${example(b)}
-
-    ${clusterSection(cluster)}
-
-    ${next ? `<div class="next"><ul>${next}</ul></div>` : ""}`;
-}
-
-const app = document.getElementById("app");
-
-const json = (path, required) =>
-  fetch(path)
-    .then((r) => {
-      if (r.ok) return r.json();
-      if (required) throw new Error(`${path} -> HTTP ${r.status}`);
-      return null; // optional: absent is a normal state, not an error
-    })
-    .catch((err) => {
-      if (required) throw err;
-      return null;
-    });
-
-Promise.all([json("/bundle.json", true), json("/cluster.mock.json", false)])
-  // RIG_NAME is injected by vite from the pod env, so two rigs sharing a
-  // cluster are distinguishable even if a copied bundle.json kept its old name.
-  .then(([b, cluster]) => {
-    app.innerHTML = render(b, import.meta.env.VITE_RIG_NAME, cluster);
-  })
-  .catch((err) => {
-    app.innerHTML = `<h1 class="err">bundle unavailable</h1>
-      <p class="tagline">${esc(err.message)}</p>`;
-  });

rig/sample-rig/rig-ui/src/style.css

@@ -1,139 +0,0 @@
-/* Minimal and self-contained — no framework dependency.
- *
- * The visual language follows soleprint's templated vein pages, because this
- * page does the same job: say what a component is, list what it exposes, and
- * show what comes back. Two treatments, deliberately distinct:
- *
- *   TOOL CHROME  bordered cards on the darker background, accent-coloured
- *                titles, endpoint rows separated by rules.
- *   OUTPUT       a lighter raised block, monospace, pre-wrap and selectable —
- *                it is data, not furniture, and should read as a payload.
- *
- * Keeping them apart matters more than either looks: it is what tells you at a
- * glance whether you are reading the tool or the thing it produced. */
-
-:root {
-  --bg: #0d0d0f;
-  --surface: #16161a;
-  --surface-raised: #1e1e24;
-  --border: #2e2e38;
-  --border-strong: #3d3d4a;
-  --text: #e8e8f0;
-  --muted: #8888a0;
-  --accent: #3ecf8e;
-  --accent-dim: #f5a623;
-  --mono: "JetBrains Mono", "Cascadia Mono", Consolas, ui-monospace, monospace;
-}
-
-body {
-  margin: 0;
-  padding: 2.5rem 1.5rem;
-  background: var(--bg);
-  color: var(--text);
-  font: 14px/1.6 var(--mono);
-}
-main { max-width: 56rem; margin: 0 auto; }
-
-h1 { margin: 0; font-size: 1.6rem; letter-spacing: 0.02em; }
-h1 .ok { color: var(--accent); }
-h1.err { color: #f06565; }
-.tagline { color: var(--muted); margin: 0.35rem 0 2.25rem; }
-
-h2 {
-  font-size: 0.75rem;
-  text-transform: uppercase;
-  letter-spacing: 0.12em;
-  color: var(--muted);
-  margin: 2.25rem 0 0.75rem;
-  font-weight: 600;
-}
-
-/* ── tool chrome ────────────────────────────────────────────────────────── */
-
-.components {
-  display: grid;
-  grid-template-columns: repeat(auto-fit, minmax(200px, 1fr));
-  gap: 0.75rem;
-}
-
-.component {
-  background: var(--bg);
-  border: 1px solid var(--border-strong);
-  border-radius: 8px;
-  padding: 0.75rem;
-}
-.component h4 {
-  margin: 0 0 0.25rem;
-  font-size: 0.95rem;
-  color: var(--accent);
-}
-.component p {
-  margin: 0;
-  font-size: 0.85rem;
-  color: var(--muted);
-}
-
-.endpoints { list-style: none; margin: 0; padding: 0; }
-.endpoints li {
-  display: flex;
-  justify-content: space-between;
-  flex-wrap: wrap;
-  gap: 0.5rem;
-  padding: 0.6rem 0;
-  border-bottom: 1px solid var(--border-strong);
-}
-.endpoints li:last-child { border-bottom: none; }
-.endpoints code {
-  background: var(--surface);
-  color: var(--accent);
-  padding: 0.25rem 0.5rem;
-  border-radius: 4px;
-}
-.endpoints .desc { color: var(--muted); font-size: 0.9rem; }
-
-.tag {
-  font-size: 0.7rem;
-  padding: 0.1rem 0.45rem;
-  border-radius: 3px;
-  background: var(--border);
-  color: var(--muted);
-  white-space: nowrap;
-}
-.tag.on { background: var(--accent); color: var(--bg); }
-
-/* ── output ─────────────────────────────────────────────────────────────── */
-
-.output {
-  background: var(--surface);
-  border: 1px solid var(--border);
-  border-radius: 8px;
-  padding: 1rem;
-  font-size: 0.85rem;
-  white-space: pre-wrap;
-  word-break: break-word;
-  user-select: text;
-  color: var(--text);
-  margin: 0;
-}
-.output .k { color: var(--accent); }
-
-/* Mocked-data banner. Loud on purpose: it only appears when the payload is
-   canned, and a demo that looks live but is not is worse than one that says so. */
-.mock {
-  margin: 0 0 0.75rem;
-  padding: 0.4rem 0.75rem;
-  border: 1px dashed var(--accent-dim);
-  border-radius: 6px;
-  color: var(--accent-dim);
-  font-size: 0.8rem;
-}
-
-.next {
-  color: #555568;
-  font-size: 0.8rem;
-  margin-top: 2.5rem;
-  border-top: 1px solid var(--border);
-  padding-top: 1rem;
-}
-.next ul { margin: 0; padding: 0 0 0 1.1rem; }
-.next li { list-style: disc; padding: 0.15rem 0; }

rig/sample-rig/rig-ui/vite.config.js

@@ -1,9 +0,0 @@
-import { defineConfig } from "vite";
-
-/* Serves on 0.0.0.0 so the pod is reachable through the Service, and allows any
- * Host header because the address is assigned at runtime (MetalLB locally, a
- * cloud load balancer on EKS) and is never known at build time. */
-export default defineConfig({
-  server: { host: "0.0.0.0", port: 5173, strictPort: true, allowedHosts: true },
-  preview: { host: "0.0.0.0", port: 5173, strictPort: true, allowedHosts: true },
-});

rig/standalone/README.md

@@ -0,0 +1,37 @@
+# standalone — single files for a machine the full rig is not going to
+
+Each script here does one of rig's jobs without the rest of the tree. Copy one
+file onto a machine, run it, read the output. Nothing to clone, nothing to
+install first.
+
+| file | does | full-rig equivalent |
+| --- | --- | --- |
+| `rigdeps.sh` | installs kind, kubectl, tilt, ctlptl and jq at rig's pins, checksum-verified, no sudo | `make deps` (`ctrl/deps.sh`) |
+| `rigmini.sh` | reports how much memory the machine *advertises* and what caps it; `push` measures what it will actually *survive* | `make mem`, and the memory section of `make check` |
+
+**These are transitional.** Where the full rig is installed, use its own
+targets instead; they read `ctrl/versions.env` and the profile, which these
+cannot.
+
+## Why single files
+
+`rigdeps.sh` carries its pins inline, because `ctrl/versions.env` is not on the
+machine it is for. That makes two copies of the same versions and checksums.
+`make pins` compares them and fails on any difference — `ctrl/versions.env` is
+the source of truth.
+
+`rigmini.sh` exists because on a container or managed workspace `/proc/meminfo`
+reports the *host's* memory while a cgroup cap kills processes at a fraction of
+it. `status` reads the caps; `push` allocates until something stops it.
+
+## Use
+
+```bash
+bash rigdeps.sh detect          # report, change nothing
+bash rigdeps.sh install dev     # install into ~/.local/bin
+bash rigmini.sh status          # advertised memory and caps; safe
+bash rigmini.sh push            # allocates until it stops — not on a machine you need
+```
+
+`rigmini.sh push` deliberately consumes memory. Run `status` first, and only run
+`push` somewhere it is acceptable for other processes to be squeezed.

rig/standalone/rigdeps.sh

@@ -0,0 +1,620 @@
+#!/usr/bin/env bash
+# Put kind, tilt and kubectl on a machine that has none of them.
+#
+# The single file companion to rigmini.sh, for the same reason: rig installs its
+# toolchain from ctrl/deps.sh reading ctrl/versions.env, and neither of those is
+# going to a fresh AWS WorkSpace. The pins live inline here instead.
+#
+# What it will not do, deliberately:
+#
+#   * no sudo, no apt, no yum. It writes into $OUT_BIN (default ~/.local/bin)
+#     and, for compose only, a symlink under ~/.docker/cli-plugins — both in
+#     your own home. Everything needing root — installing Docker, joining the
+#     docker group, raising inotify limits — is REPORTED for you to decide on.
+#     That is what makes it safe to run on a machine that already works.
+#   * no unverified download. Every artifact is checked against a SHA256 taken
+#     from the publisher's own release list. A mismatch aborts.
+#   * no guessing at another architecture. See ARCHITECTURE below.
+#
+# Two tiers, because "install the toolchain" is not one decision:
+#
+#   core  kubectl, jq — talk to a cluster someone else runs. Nothing that
+#         creates one. The right answer on a managed or corporate machine.
+#   dev   core plus kind, tilt and ctlptl — build clusters and hot-reload
+#         into them. The default, and what you want on a workspace of your own.
+#
+# Usage:
+#   rigdeps.sh detect              report the host, change nothing
+#   rigdeps.sh list                the pinned versions and where they come from
+#   rigdeps.sh install [core|dev]  detect, download, verify, install, report
+#   rigdeps.sh fetch [core|dev] [--to DIR]   download + verify only
+#   rigdeps.sh verify              run what is installed and see if it works
+set -euo pipefail
+
+OUT_BIN="${OUT_BIN:-$HOME/.local/bin}"
+
+# ── the pinned toolchain ───────────────────────────────────────────────────
+#
+# ARCHITECTURE. These checksums are the upstream-published SHA256 of the
+# **linux/amd64** artifact and of nothing else. An arm64 WorkSpace bundle needs
+# a different binary with a different checksum, and this script refuses rather
+# than reusing these — a checksum that is merely plausible is worse than none,
+# because it turns a verified download into a ceremony.
+#
+# To bump a version, or to add arm64: take the checksum from the release's own
+# published list, never from a download you did.
+#
+#   curl -sSL https://github.com/<org>/<repo>/releases/download/<tag>/checksums.txt
+#
+# kubectl publishes its own instead, at <KUBECTL_URL>.sha256.
+
+KIND_VERSION=v0.32.0
+KIND_SHA256=50030de23cf40a18505f20426f6a8506bedf13c6e509244bd1fa9463721b0f54
+KIND_URL="https://github.com/kubernetes-sigs/kind/releases/download/${KIND_VERSION}/kind-linux-amd64"
+
+KUBECTL_VERSION=v1.36.3
+KUBECTL_SHA256=ebbd080e7c2e275093b55915722043257eb24004363e20acb3c4d71919f88336
+KUBECTL_URL="https://dl.k8s.io/release/${KUBECTL_VERSION}/bin/linux/amd64/kubectl"
+
+TILT_VERSION=0.37.6
+TILT_SHA256=e9672b8a18d43501f35dcfe98465969a7db0e436b36cf0c50c7e6f8d40de5fe6
+TILT_URL="https://github.com/tilt-dev/tilt/releases/download/v${TILT_VERSION}/tilt.${TILT_VERSION}.linux.x86_64.tar.gz"
+
+# ctlptl creates a kind cluster WITH a local registry wired in, which is what
+# keeps images off docker.io — an unqualified image name resolves to
+# docker.io/library/<name>, and there is nothing structural stopping a push there.
+CTLPTL_VERSION=0.9.4
+CTLPTL_SHA256=c63a1ec28e60bc3faf6becb76f53355c5cf5e0143dafdd27ad85db5584fa6b1e
+CTLPTL_URL="https://github.com/tilt-dev/ctlptl/releases/download/v${CTLPTL_VERSION}/ctlptl.${CTLPTL_VERSION}.linux.x86_64.tar.gz"
+
+# Upstream's static build. Debian's jq is linked against libjq/libonig, which is
+# fine on Debian and not portable anywhere else.
+JQ_VERSION=1.8.2
+JQ_SHA256=b1c22172dd303f3be49e935aa56aa48a8b7a46e0bc838b4997d3bb451495870f
+JQ_URL="https://github.com/jqlang/jq/releases/download/jq-${JQ_VERSION}/jq-linux-amd64"
+
+# Distro docker packages ship the daemon and CLI but frequently not this, so
+# `docker compose up` fails with "unknown command" on an otherwise working
+# Docker. It is a CLI plugin: the binary is found by name in a plugin directory,
+# which is why install_compose_plugin links it into ~/.docker/cli-plugins.
+COMPOSE_VERSION=5.5.1
+COMPOSE_SHA256=db1889184726840f75c4f9c001048430d4f25b3be3cb084d3ddd762bc0aed576
+COMPOSE_URL="https://github.com/docker/compose/releases/download/v${COMPOSE_VERSION}/docker-compose-linux-x86_64"
+
+CORE_TOOLS="kubectl jq"
+DEV_TOOLS="kind tilt ctlptl docker-compose"
+
+# No helm: every rig addon installs with `kubectl apply -f`, so nothing has ever
+# invoked it. Add it the day something actually needs a chart.
+
+# Collected as we go, printed by report_manual() at the very end. Anything that
+# needs root or a decision lands here instead of being done.
+MANUAL=()
+
+# ── platform ───────────────────────────────────────────────────────────────
+
+# Windows outside WSL — Git Bash, MSYS, Cygwin — looks close enough to work and
+# then fails in a pile of confusing ways: no /proc, no docker socket, none of
+# the tooling. Detectable, so name it instead.
+require_linux() {
+    case "$(uname -s)" in
+        MINGW*|MSYS*|CYGWIN*)
+            cat >&2 <<'EOF'
+This has to run inside WSL, not Git Bash / MSYS / Cygwin.
+
+If WSL is not installed yet, from an elevated PowerShell or Command Prompt:
+
+    wsl --install
+
+That enables Windows features and needs a reboot, so it is not something this
+script will do for you. Afterwards, open the Linux shell it installs and run
+this from there.
+EOF
+            exit 1 ;;
+        Linux) ;;
+        *)  echo "$(uname -s) is not Linux. These are linux binaries; nothing here" >&2
+            echo "would run even if it downloaded." >&2
+            exit 1 ;;
+    esac
+}
+
+arch() {
+    case "$(uname -m)" in
+        x86_64|amd64)  echo amd64 ;;
+        aarch64|arm64) echo arm64 ;;
+        *)             uname -m ;;
+    esac
+}
+
+# The pins above are amd64. Rather than download something that cannot execute
+# and let it fail as "cannot execute binary file: Exec format error", say so
+# here and hand over the commands that produce the right checksums.
+require_amd64() {
+    local a; a=$(arch)
+    [ "$a" = "amd64" ] && return 0
+    cat >&2 <<EOF
+This machine is ${a} ($(uname -m)); every pin in this script is linux/amd64.
+
+Nothing here would run, so it does not download. To make an ${a} version, the
+URLs need the ${a} artifact and the checksums need to come from each project's
+own published list — not from these values, and not from a download you did:
+
+    curl -sSL https://github.com/kubernetes-sigs/kind/releases/download/${KIND_VERSION}/checksums.txt
+    curl -sSL https://dl.k8s.io/release/${KUBECTL_VERSION}/bin/linux/${a}/kubectl.sha256
+    curl -sSL https://github.com/tilt-dev/tilt/releases/download/v${TILT_VERSION}/checksums.txt
+    curl -sSL https://github.com/tilt-dev/ctlptl/releases/download/v${CTLPTL_VERSION}/checksums.txt
+    curl -sSL https://github.com/jqlang/jq/releases/download/jq-${JQ_VERSION}/sha256sum.txt
+
+Edit the pinned block at the top of this file with what those print.
+EOF
+    exit 1
+}
+
+is_wsl() { grep -qi microsoft /proc/version 2>/dev/null; }
+
+# ── the tools this script itself needs ─────────────────────────────────────
+
+# A fresh minimal image may genuinely have neither curl nor wget. Find out once,
+# up front, rather than half way through the first download.
+DL=""
+pick_downloader() {
+    if command -v curl >/dev/null 2>&1; then DL=curl
+    elif command -v wget >/dev/null 2>&1; then DL=wget
+    else
+        echo "neither curl nor wget is installed, so nothing can be downloaded." >&2
+        echo "Install one first:  $(pkg_install_cmd curl)" >&2
+        exit 1
+    fi
+}
+
+download() {
+    local url="$1" out="$2"
+    case "$DL" in
+        curl) curl -fsSL --retry 3 -o "$out" "$url" ;;
+        wget) wget -q --tries=3 -O "$out" "$url" ;;
+    esac
+}
+
+# sha256sum is coreutils; shasum is the perl one that turns up on stripped
+# images. Verification is not optional, so if neither exists that is fatal.
+SHA=""
+pick_sha() {
+    if command -v sha256sum >/dev/null 2>&1; then SHA=sha256sum
+    elif command -v shasum >/dev/null 2>&1; then SHA="shasum -a 256"
+    else
+        echo "no sha256sum and no shasum — downloads could not be verified." >&2
+        echo "Refusing to install unverified binaries." >&2
+        exit 1
+    fi
+}
+
+# ── package manager, for the instructions only ─────────────────────────────
+# This never runs a package manager. It names one so the reported action is
+# something you can paste, on the distro you are actually on — an apt line on
+# Amazon Linux 2 is a wrong answer dressed up as help.
+
+pkg_install_cmd() {
+    local pkg="$1"
+    if command -v apt-get >/dev/null 2>&1; then echo "sudo apt-get update && sudo apt-get install -y $pkg"
+    elif command -v dnf >/dev/null 2>&1;     then echo "sudo dnf install -y $pkg"
+    elif command -v yum >/dev/null 2>&1;     then echo "sudo yum install -y $pkg"
+    elif command -v zypper >/dev/null 2>&1;  then echo "sudo zypper install -y $pkg"
+    elif command -v apk >/dev/null 2>&1;     then echo "sudo apk add $pkg"
+    else echo "install '$pkg' with this system's package manager"
+    fi
+}
+
+docker_pkg() {
+    # Debian and Ubuntu call it docker.io; the RPM distros call it docker.
+    if command -v apt-get >/dev/null 2>&1; then echo docker.io; else echo docker; fi
+}
+
+# ── detect ─────────────────────────────────────────────────────────────────
+
+detect() {
+    echo "host"
+    echo "  kernel      $(uname -r)"
+    echo "  arch        $(arch)  ($(uname -m))"
+    [ -r /etc/os-release ] && \
+        echo "  distro      $(sed -n 's/^PRETTY_NAME="\(.*\)"/\1/p' /etc/os-release)"
+    if is_wsl; then echo "  platform    WSL"; else echo "  platform    native linux"; fi
+
+    local total_kb avail_kb
+    total_kb=$(awk '/^MemTotal:/{print $2}' /proc/meminfo)
+    avail_kb=$(awk '/^MemAvailable:/{print $2}' /proc/meminfo)
+    printf "  memory      %d GB total, %d GB available\n" \
+        $((total_kb / 1024 / 1024)) $((avail_kb / 1024 / 1024))
+    if [ $((avail_kb / 1024 / 1024)) -lt 4 ]; then
+        echo "  ! under 4 GB available — a cluster will struggle here."
+        echo "    rigmini.sh says how much this box will actually give you."
+    fi
+
+    echo "  install to  $OUT_BIN"
+    detect_libc
+    detect_prereqs
+    detect_docker
+    detect_inotify
+    return 0
+}
+
+# tilt is the one binary here that needs a recent glibc. MEASURED, not guessed:
+# tilt 0.37.6 on Amazon Linux 2 (glibc 2.26) fails with
+#
+#     /lib64/libc.so.6: version `GLIBC_2.34' not found (required by .../tilt)
+#
+# which names a symbol rather than the problem. Amazon Linux 2 is a stock
+# WorkSpaces bundle, so this is the likely case, not an exotic one. Report the
+# version now; `verify` catches the actual failure after installing.
+detect_libc() {
+    local v=""
+    if command -v ldd >/dev/null 2>&1; then
+        v=$(ldd --version 2>/dev/null | head -1 | grep -oE '[0-9]+\.[0-9]+$' || true)
+    fi
+    if [ -z "$v" ]; then
+        echo "  libc        unknown (no ldd) — 'verify' is the real test"
+        return 0
+    fi
+    echo "  libc        glibc $v"
+    if [ "$(printf '%s\n2.34\n' "$v" | sort -V | head -1)" != "2.34" ]; then
+        echo "  ! older than glibc 2.34, which tilt needs. kubectl, kind, jq and"
+        echo "    ctlptl are static or libc-only and work here; tilt will not start."
+        echo "    Install the core tier, or run tilt from a container."
+    fi
+    return 0
+}
+
+# What this script needs to do its own job. Reported here so `detect` answers
+# "will install work?" instead of leaving you to find out one download in.
+# Amazon Linux 2 ships without tar, which is exactly the surprise this catches.
+detect_prereqs() {
+    local missing=""
+    if command -v curl >/dev/null 2>&1; then echo "  download    curl"
+    elif command -v wget >/dev/null 2>&1; then echo "  download    wget"
+    else echo "  ! no curl and no wget — nothing can be downloaded"; missing+=" curl"
+    fi
+
+    if command -v sha256sum >/dev/null 2>&1 || command -v shasum >/dev/null 2>&1; then
+        echo "  checksums   ok"
+    else
+        echo "  ! no sha256sum or shasum — downloads could not be verified"
+        missing+=" coreutils"
+    fi
+
+    if command -v tar >/dev/null 2>&1 && command -v gzip >/dev/null 2>&1; then
+        echo "  archives    tar + gzip"
+    else
+        echo "  ! no tar/gzip — tilt and ctlptl ship as tarballs, so the dev tier"
+        echo "    cannot be unpacked. The core tier is two bare binaries and is fine."
+        missing+=" tar gzip"
+    fi
+
+    if [ -n "$missing" ]; then
+        MANUAL+=("Install what this script needs to run at all:
+      $(pkg_install_cmd "${missing# }")")
+    fi
+    return 0
+}
+
+detect_docker() {
+    # kind builds a cluster out of containers. Without a reachable daemon,
+    # everything here installs perfectly and then does nothing.
+    if ! command -v docker >/dev/null 2>&1; then
+        if [ -S /var/run/docker.sock ]; then
+            echo "  docker      socket present, no cli"
+            return 0
+        fi
+        echo "  ! docker not installed — kind has nothing to build a cluster in"
+        MANUAL+=("Install Docker. It is the one real prerequisite, and the only
+  thing here that needs root:
+      $(pkg_install_cmd "$(docker_pkg)")
+      sudo systemctl enable --now docker
+      sudo usermod -aG docker \"\$USER\"
+  then log out and back in, so the new group applies to your shell.")
+        return 0
+    fi
+    if docker info >/dev/null 2>&1; then
+        echo "  docker      $(docker version --format '{{.Server.Version}}' 2>/dev/null)"
+        # Distro packages routinely omit the compose plugin, so a working
+        # daemon says nothing about whether `docker compose up` will run.
+        if docker compose version >/dev/null 2>&1; then
+            echo "  compose     $(docker compose version --short 2>/dev/null)"
+        else
+            echo "  ! no 'docker compose' plugin — compose files will not start."
+            echo "    The dev tier installs one; no root needed."
+        fi
+        local n
+        n=$(docker ps --filter "label=io.x-k8s.kind.cluster" --format '{{.Names}}' 2>/dev/null | wc -l)
+        # Must be an `if`, not `[ ] && echo`: as the last statement here the
+        # latter returns 1 when the count is zero, and `set -e` kills the
+        # caller. That is the fresh-machine case, where it does most harm.
+        if [ "$n" -gt 0 ]; then
+            echo "  - $n kind node container(s) already running"
+        fi
+    else
+        echo "  ! docker cli present but the daemon is unreachable"
+        MANUAL+=("Start Docker, or add yourself to the docker group:
+      sudo systemctl enable --now docker
+      sudo usermod -aG docker \"\$USER\"   # then log out and back in")
+    fi
+    return 0
+}
+
+# kind and tilt both watch large trees. Distro defaults are far too low and the
+# failure mode is silent: tilt simply stops noticing that files changed.
+detect_inotify() {
+    local w i
+    w=$(cat /proc/sys/fs/inotify/max_user_watches 2>/dev/null || echo 0)
+    i=$(cat /proc/sys/fs/inotify/max_user_instances 2>/dev/null || echo 0)
+    echo "  inotify     watches=$w instances=$i"
+    if [ "$w" -lt 524288 ] || [ "$i" -lt 512 ]; then
+        echo "  ! low — tilt will silently stop seeing file changes"
+        MANUAL+=("Raise the inotify limits (needs root):
+      echo -e 'fs.inotify.max_user_watches=524288\\nfs.inotify.max_user_instances=512' \\
+        | sudo tee /etc/sysctl.d/99-rig.conf
+      sudo sysctl --system")
+    fi
+    return 0
+}
+
+# ── fetch ──────────────────────────────────────────────────────────────────
+
+verify_sha() {
+    local file="$1" want="$2" name="$3" got
+    got=$($SHA "$file" | awk '{print $1}')
+    if [ "$got" != "$want" ]; then
+        echo >&2
+        echo "CHECKSUM MISMATCH for $name — not installing it." >&2
+        echo "  expected $want" >&2
+        echo "  got      $got" >&2
+        echo >&2
+        echo "Either the pin in this script is stale, or what arrived is not what" >&2
+        echo "the publisher released. Neither is worth guessing about." >&2
+        rm -f "$file"
+        exit 1
+    fi
+}
+
+# fetch_bin <name> <url> <sha256> <dest-dir>   — a bare binary
+fetch_bin() {
+    local name="$1" url="$2" sha="$3" dest="$4"
+    local tmp="$dest/.$name.tmp"
+    printf '  %-8s ' "$name"
+    download "$url" "$tmp"
+    verify_sha "$tmp" "$sha" "$name"
+    mv "$tmp" "$dest/$name"
+    chmod +x "$dest/$name"
+    echo "ok"
+}
+
+# fetch_tgz <name> <url> <sha256> <dest-dir> <path-inside> <strip>
+# Archive layouts differ, so the caller says which. tilt and ctlptl both ship
+# the binary at the archive root, hence strip=0.
+fetch_tgz() {
+    local name="$1" url="$2" sha="$3" dest="$4" inner="$5" strip="$6"
+    local tmp="$dest/.$name.tgz"
+    printf '  %-8s ' "$name"
+    download "$url" "$tmp"
+    verify_sha "$tmp" "$sha" "$name"
+    # --no-same-owner: some archives ship as uid 1001, and extracting as root
+    # would otherwise restore an owner that is not you.
+    tar -xzf "$tmp" -C "$dest" --strip-components="$strip" --no-same-owner "$inner"
+    rm -f "$tmp"
+    chmod +x "$dest/$name"
+    echo "ok"
+}
+
+fetch() {
+    local dest="$OUT_BIN" tier="dev"
+    while [ $# -gt 0 ]; do
+        case "$1" in
+            --to)     dest="${2:?--to needs a directory}"; shift 2 ;;
+            core|dev) tier="$1"; shift ;;
+            *) echo "unknown argument: $1" >&2; exit 1 ;;
+        esac
+    done
+    mkdir -p "$dest"
+
+    if ! command -v tar >/dev/null 2>&1 && [ "$tier" = "dev" ]; then
+        echo "tar is missing, and tilt and ctlptl ship as tarballs." >&2
+        echo "  $(pkg_install_cmd tar)" >&2
+        echo "Or install the core tier, which is two bare binaries:  $0 install core" >&2
+        exit 1
+    fi
+
+    echo "fetching '$tier' into $dest  (verifying every checksum)"
+    fetch_bin kubectl "$KUBECTL_URL" "$KUBECTL_SHA256" "$dest"
+    fetch_bin jq      "$JQ_URL"      "$JQ_SHA256"      "$dest"
+    if [ "$tier" = "dev" ]; then
+        fetch_bin kind   "$KIND_URL"   "$KIND_SHA256"   "$dest"
+        fetch_tgz tilt   "$TILT_URL"   "$TILT_SHA256"   "$dest" tilt   0
+        fetch_tgz ctlptl "$CTLPTL_URL" "$CTLPTL_SHA256" "$dest" ctlptl 0
+        fetch_bin docker-compose "$COMPOSE_URL" "$COMPOSE_SHA256" "$dest"
+    fi
+    return 0
+}
+
+# A copy in OUT_BIN only gives you `docker-compose`. The hyphenated form is the
+# retired v1 spelling; every compose file written in the last few years assumes
+# `docker compose`, and that resolves plugins by name from this directory.
+install_compose_plugin() {
+    local src="$OUT_BIN/docker-compose" dir="$HOME/.docker/cli-plugins"
+    [ -x "$src" ] || return 0
+    mkdir -p "$dir"
+    if [ -e "$dir/docker-compose" ] && [ ! -L "$dir/docker-compose" ]; then
+        echo
+        echo "  ! $dir/docker-compose exists and is not a symlink — left alone"
+        MANUAL+=("Something already installs the compose plugin at
+      $dir/docker-compose
+  To use the pinned build instead:
+      ln -sf $src $dir/docker-compose")
+        return 0
+    fi
+    ln -sfn "$src" "$dir/docker-compose"
+    echo
+    echo "  compose plugin linked into $dir"
+    return 0
+}
+
+# ── verify ─────────────────────────────────────────────────────────────────
+
+tier_tools() { [ "$1" = "core" ] && echo "$CORE_TOOLS" || echo "$CORE_TOOLS $DEV_TOOLS"; }
+
+# Downloading a verified binary proves it is the right file, not that this
+# machine can run it. On an old distro tilt fails here, with a linker error
+# about a missing symbol, and finding that out now beats finding out during a
+# first cluster build.
+verify_tools() {
+    local tier="${1:-dev}" b bin out rc broke=0
+    echo "checking that each one actually runs"
+    for b in $(tier_tools "$tier"); do
+        bin="$OUT_BIN/$b"
+        if [ ! -x "$bin" ]; then
+            printf '  %-8s not installed\n' "$b"
+            continue
+        fi
+        rc=0
+        case "$b" in
+            kubectl) out=$("$bin" version --client 2>&1 | head -1) || rc=$? ;;
+            jq)      out=$("$bin" --version 2>&1 | head -1) || rc=$? ;;
+            *)       out=$("$bin" version 2>&1 | head -1) || rc=$? ;;
+        esac
+        if [ "$rc" -eq 0 ]; then
+            printf '  %-8s %s\n' "$b" "$out"
+        else
+            printf '  ! %-6s does not run here: %s\n' "$b" "$out"
+            broke=1
+        fi
+    done
+    if [ "$broke" -eq 1 ]; then
+        echo
+        echo "  A binary that downloads and verifies but will not start is almost"
+        echo "  always this distro's libc being older than the release needs."
+        echo "  'detect' prints the glibc version. The core tier (kubectl + jq)"
+        echo "  has no such dependency and will work regardless."
+    fi
+    return 0
+}
+
+# ── install ────────────────────────────────────────────────────────────────
+
+# Installing into a directory early in PATH silently replaces whatever the
+# machine was already using, which on a shared or corporate machine can break
+# unrelated work — kubectl more than one minor away from its cluster is the
+# common one. Say so; never decide it.
+warn_shadowing() {
+    local b existing shadowed="" tier="${1:-dev}"
+    case ":${PATH}:" in
+        *":$OUT_BIN:"*) ;;
+        *) return 0 ;;   # not on PATH, so nothing is being shadowed yet
+    esac
+    for b in $(tier_tools "$tier"); do
+        [ -x "$OUT_BIN/$b" ] || continue
+        existing=$(PATH=$(echo "$PATH" | tr ':' '\n' | grep -vx "$OUT_BIN" | paste -sd:) \
+                   command -v "$b" 2>/dev/null || true)
+        [ -n "$existing" ] || continue
+        [ "$existing" = "$OUT_BIN/$b" ] && continue
+        shadowed+="      $b  $existing"$'\n'
+    done
+    [ -n "$shadowed" ] || return 0
+
+    echo
+    echo "  ! these were already installed elsewhere and are now shadowed:"
+    printf '%s' "$shadowed"
+    MANUAL+=("Decide which toolchain wins. To keep the previous one:
+      rm -f $(for b in $(tier_tools "$tier"); do printf '%s ' "$OUT_BIN/$b"; done)
+  Or install somewhere private instead:
+      OUT_BIN=\$PWD/bin $0 install")
+    return 0
+}
+
+report_manual() {
+    echo
+    if [ ${#MANUAL[@]} -eq 0 ]; then
+        echo "nothing left to do by hand."
+        return 0
+    fi
+    echo "host actions this cannot perform (${#MANUAL[@]}):"
+    echo
+    local n=1 m
+    for m in "${MANUAL[@]}"; do
+        echo "  $n. $m"
+        echo
+        n=$((n + 1))
+    done
+    return 0
+}
+
+install() {
+    local tier="${1:-dev}"
+    detect
+    echo
+    fetch "$tier"
+    # An `if`, not `[ ] && ...`: on the core tier the test fails, and under
+    # `set -e` a bare failing test here would end the run silently.
+    if [ "$tier" = "dev" ]; then
+        install_compose_plugin
+    fi
+    echo
+    verify_tools "$tier"
+    warn_shadowing "$tier"
+
+    if [ "$tier" = "core" ]; then
+        echo
+        echo "  core tier: no kind, tilt, ctlptl or compose. '$0 install dev' adds them."
+    fi
+
+    case ":${PATH}:" in
+        *":$OUT_BIN:"*) ;;
+        *) MANUAL+=("Put the toolchain on your PATH — add to ~/.bashrc:
+      export PATH=\"${OUT_BIN}:\$PATH\"
+  then:  source ~/.bashrc") ;;
+    esac
+
+    report_manual
+
+    if [ "$tier" = "dev" ]; then
+        echo "Once Docker is reachable and this is on PATH:"
+        echo
+        echo "    kind create cluster --name scratch"
+        echo "    kubectl cluster-info --context kind-scratch"
+        echo "    kind delete cluster --name scratch"
+        echo
+        echo "That round trip is the real test that this machine can host a rig."
+    fi
+    return 0
+}
+
+list() {
+    echo "pinned, linux/amd64 only:"
+    printf '  %-14s %s\n' kubectl "$KUBECTL_VERSION"
+    printf '  %-14s %s\n' jq      "$JQ_VERSION"
+    printf '  %-14s %s\n' kind    "$KIND_VERSION"
+    printf '  %-14s %s\n' tilt    "$TILT_VERSION"
+    printf '  %-14s %s\n' ctlptl  "$CTLPTL_VERSION"
+    printf '  %-14s %s\n' docker-compose "$COMPOSE_VERSION"
+    echo
+    echo "  core = $CORE_TOOLS"
+    echo "  dev  = $CORE_TOOLS $DEV_TOOLS"
+    echo
+    echo "Checksums are pinned in the block at the top of this file. To bump one,"
+    echo "take the new checksum from the publisher's own release list — the header"
+    echo "comment has the exact commands."
+    return 0
+}
+
+# ── main ───────────────────────────────────────────────────────────────────
+
+require_linux
+
+case "${1:-install}" in
+    detect)  detect; report_manual ;;
+    list)    list ;;
+    verify)  verify_tools "${2:-dev}" ;;
+    fetch)   shift; require_amd64; pick_downloader; pick_sha; fetch "$@" ;;
+    install) shift; require_amd64; pick_downloader; pick_sha; install "${1:-dev}" ;;
+    *)       echo "usage: $0 [detect|list|install|fetch|verify]" >&2
+             echo "       install [core|dev]           (default dev)" >&2
+             echo "       fetch   [core|dev] [--to DIR]" >&2
+             echo "       OUT_BIN=<dir> overrides the install directory" >&2
+             exit 1 ;;
+esac

rig/standalone/rigmini.sh

@@ -0,0 +1,635 @@
+#!/usr/bin/env bash
+# How much memory this box will actually give you before something dies.
+#
+# rig answers this for a machine it is installed on. This is the single file
+# version, for a machine rig is not going to: paste it onto a fresh AWS
+# WorkSpace, an EC2 box or a container, run it, and get the same numbers in the
+# same order so two machines can be read side by side.
+#
+# There are two numbers and they are rarely the same. `status` reports what the
+# machine ADVERTISES and what is quietly capping it. `push` finds what it will
+# SURVIVE, by allocating until it stops.
+#
+# The gap between them is the whole reason this exists. Under WSL the cap lives
+# in .wslconfig; in a container or a managed workspace it is a cgroup limit, and
+# there /proc/meminfo reports the HOST's memory while the kernel kills you at a
+# fraction of it. A script that only read MemTotal would confidently report 32 GB
+# on a box that OOMs at 2.
+#
+# Reports and instructs. It never raises a limit, frees anything, writes a
+# config or installs a package — on a machine you are still evaluating, a probe
+# that changes what it is measuring is worse than no probe.
+#
+# Usage:
+#   rigmini.sh status                                  what it has, what caps it
+#   rigmini.sh push [--to GB] [--to-oom]               climb until it stops
+#   rigmini.sh all  [--budget GB]                      both, then the verdict
+set -euo pipefail
+
+# ── defaults ───────────────────────────────────────────────────────────────
+
+STEP_MB=0             # per allocation; 0 means scale it to the ceiling. See push().
+STEP_EXPLICIT=no      # whether --step was given, which turns the scaling off.
+TO_MB=""              # --to: stop here regardless. Empty means no hard cap.
+TO_OOM=no             # --to-oom: opt in to running until the kernel intervenes.
+BUDGET_GB=6           # what the rig data profile is assumed to want; see all().
+BUDGET_EXPLICIT=no    # whether --budget was given, which retires the guess below.
+
+# ── platform ───────────────────────────────────────────────────────────────
+
+# Windows outside WSL — Git Bash, MSYS, Cygwin — looks close enough to work and
+# then fails in a pile of confusing ways: no /proc, no docker socket, none of
+# the tooling. Detectable, so name it instead.
+require_linux() {
+    case "$(uname -s)" in
+        MINGW*|MSYS*|CYGWIN*)
+            cat >&2 <<'EOF'
+This has to run inside WSL, not Git Bash / MSYS / Cygwin.
+
+If WSL is not installed yet, from an elevated PowerShell or Command Prompt:
+
+    wsl --install
+
+That enables Windows features and needs a reboot, so it is not something this
+script will do for you. Afterwards, open the Linux shell it installs and run
+this from there.
+EOF
+            exit 1 ;;
+    esac
+
+    # Everything below reads /proc. Without it there is nothing to measure, and
+    # failing here beats printing a page of empty fields.
+    if [ ! -r /proc/meminfo ]; then
+        echo "no readable /proc/meminfo — this needs a Linux kernel." >&2
+        echo "On macOS or a BSD none of the numbers below exist." >&2
+        exit 1
+    fi
+}
+
+is_wsl() { grep -qi microsoft /proc/version 2>/dev/null; }
+
+is_container() {
+    [ -f /.dockerenv ] && return 0
+    grep -qE '(docker|containerd|kubepods|lxc|podman)' /proc/1/cgroup 2>/dev/null
+}
+
+platform() {
+    if is_wsl; then echo WSL
+    elif is_container; then echo container
+    else echo "native linux"
+    fi
+}
+
+# ── reading memory ─────────────────────────────────────────────────────────
+
+mb() { echo $(( $(awk "/^$1:/{print \$2}" /proc/meminfo) / 1024 )); }
+
+# MemAvailable arrived in kernel 3.14. Older kernels — and they turn up on
+# corporate images — need the estimate it replaced, which is worse but not wrong.
+avail_meminfo_mb() {
+    if grep -q '^MemAvailable:' /proc/meminfo; then
+        mb MemAvailable
+    else
+        awk '/^(MemFree|Buffers|Cached):/{t+=$2} END{print int(t/1024)}' /proc/meminfo
+    fi
+}
+
+# Where a cgroup records this cgroup's own limit and usage. Set once by
+# find_cgroup, because every later reading needs both and hunting for the files
+# on each call would be the slow part of the poll loop.
+CG_MAX_FILE=""
+CG_CUR_FILE=""
+CG_VERSION=""
+
+find_cgroup() {
+    local rel
+
+    # Inside a container the cgroup namespace makes the top of the tree BE the
+    # container's own cgroup, so the unqualified path is already the right one.
+    # On a host it is the root cgroup, which is never limited — hence the second
+    # attempt via /proc/self/cgroup, which names the slice this shell is in.
+    if [ -r /sys/fs/cgroup/memory.max ]; then
+        CG_VERSION=v2
+        CG_MAX_FILE=/sys/fs/cgroup/memory.max
+        CG_CUR_FILE=/sys/fs/cgroup/memory.current
+    elif [ -r /sys/fs/cgroup/memory/memory.limit_in_bytes ]; then
+        CG_VERSION=v1
+        CG_MAX_FILE=/sys/fs/cgroup/memory/memory.limit_in_bytes
+        CG_CUR_FILE=/sys/fs/cgroup/memory/memory.usage_in_bytes
+    fi
+
+    rel=$(awk -F: '$1=="0"{print $3; exit}' /proc/self/cgroup 2>/dev/null || true)
+    if [ -n "$rel" ] && [ "$rel" != "/" ] && [ -r "/sys/fs/cgroup${rel}/memory.max" ]; then
+        CG_VERSION=v2
+        CG_MAX_FILE="/sys/fs/cgroup${rel}/memory.max"
+        CG_CUR_FILE="/sys/fs/cgroup${rel}/memory.current"
+        return 0
+    fi
+
+    rel=$(awk -F: '$2 ~ /(^|,)memory(,|$)/{print $3; exit}' /proc/self/cgroup 2>/dev/null || true)
+    if [ -n "$rel" ] && [ "$rel" != "/" ] \
+       && [ -r "/sys/fs/cgroup/memory${rel}/memory.limit_in_bytes" ]; then
+        CG_VERSION=v1
+        CG_MAX_FILE="/sys/fs/cgroup/memory${rel}/memory.limit_in_bytes"
+        CG_CUR_FILE="/sys/fs/cgroup/memory${rel}/memory.usage_in_bytes"
+    fi
+    return 0
+}
+
+# The cap in MB, or "" when there is none worth reporting. v2 spells unlimited
+# "max"; v1 spells it as a number near 2^63, which is why this compares against
+# MemTotal rather than testing for a magic value — a "limit" above the machine's
+# own memory is not a limit, however it is written.
+cgroup_cap_mb() {
+    local raw cap
+    [ -n "$CG_MAX_FILE" ] && [ -r "$CG_MAX_FILE" ] || { echo ""; return 0; }
+    raw=$(cat "$CG_MAX_FILE" 2>/dev/null || echo max)
+    [ "$raw" = "max" ] && { echo ""; return 0; }
+    case "$raw" in ''|*[!0-9]*) echo ""; return 0 ;; esac
+    cap=$((raw / 1024 / 1024))
+    [ "$cap" -ge "$(mb MemTotal)" ] && { echo ""; return 0; }
+    echo "$cap"
+}
+
+cgroup_used_mb() {
+    local raw
+    [ -n "$CG_CUR_FILE" ] && [ -r "$CG_CUR_FILE" ] || { echo ""; return 0; }
+    raw=$(cat "$CG_CUR_FILE" 2>/dev/null || echo "")
+    case "$raw" in ''|*[!0-9]*) echo ""; return 0 ;; esac
+    echo $((raw / 1024 / 1024))
+}
+
+# ulimit -v is a per-process address-space cap. It stops YOU long before the box
+# does, and because it is inherited from a login shell it is easy to hit without
+# knowing it is set.
+ulimit_v_mb() {
+    local v; v=$(ulimit -v 2>/dev/null || echo unlimited)
+    [ "$v" = "unlimited" ] && { echo ""; return 0; }
+    case "$v" in ''|*[!0-9]*) echo ""; return 0 ;; esac
+    echo $((v / 1024))
+}
+
+# The number everything else is about: the lowest of the things that can stop
+# you. Printed at the end of `status` and used as the sanity bound in `push`.
+effective_ceiling_mb() {
+    local c; c=$(mb MemTotal)
+    local cap; cap=$(cgroup_cap_mb)
+    local ul;  ul=$(ulimit_v_mb)
+    [ -n "$cap" ] && [ "$cap" -lt "$c" ] && c="$cap"
+    [ -n "$ul" ]  && [ "$ul"  -lt "$c" ] && c="$ul"
+    echo "$c"
+}
+
+# How much room is left RIGHT NOW, from whichever accounting actually governs.
+# In a capped container /proc/meminfo describes the host and is worse than
+# useless for this — it would report tens of gigabytes free on a box that is one
+# allocation from being killed.
+headroom_mb() {
+    local cap used
+    cap=$(cgroup_cap_mb)
+    used=$(cgroup_used_mb)
+    if [ -n "$cap" ] && [ -n "$used" ]; then
+        echo $(( cap - used ))
+    else
+        avail_meminfo_mb
+    fi
+}
+
+# ── status ─────────────────────────────────────────────────────────────────
+
+# /mnt/c/Users can hold several real accounts — a renamed login leaves the old
+# directory behind — so picking the first alphabetically is a coin toss. Ask
+# Windows, then fall back to whichever profile actually owns a config.
+wslconfig_path() {
+    local profile winpath found
+    profile=$(cmd.exe /c "echo %USERPROFILE%" 2>/dev/null | tr -d "\r\n" || true)
+    case "$profile" in
+        ""|*%*) ;;
+        *) winpath=$(wslpath -u "$profile" 2>/dev/null || true)
+           if [ -n "$winpath" ] && [ -d "$winpath" ]; then
+               echo "$winpath/.wslconfig"; return 0
+           fi ;;
+    esac
+    found=$(ls -d /mnt/c/Users/*/.wslconfig 2>/dev/null | head -1 || true)
+    [ -n "$found" ] && echo "$found"
+    return 0
+}
+
+hogs() {
+    echo "  holding the most:"
+    ps -eo rss,comm --sort=-rss 2>/dev/null \
+      | awk 'NR>1 && NR<=6 {printf "    %6.0f MB  %s\n", $1/1024, $2}'
+    return 0
+}
+
+status() {
+    local total avail swap_total swap_free cap ul cur
+
+    echo "host"
+    echo "  platform    $(platform)"
+    echo "  kernel      $(uname -r)"
+    [ -r /etc/os-release ] && \
+        echo "  distro      $(sed -n 's/^PRETTY_NAME="\(.*\)"/\1/p' /etc/os-release)"
+    echo "  cpu         $(getconf _NPROCESSORS_ONLN 2>/dev/null || echo '?') online, load $(cut -d' ' -f1-3 /proc/loadavg)"
+
+    # ── the caps first, because they decide what the totals below are worth ──
+    echo
+    echo "caps"
+    cap=$(cgroup_cap_mb)
+    if [ -n "$cap" ]; then
+        cur=$(cgroup_used_mb)
+        echo "  cgroup      ${cap} MB  (${CG_VERSION}, ${CG_CUR_FILE##*/} says ${cur:-?} MB used)"
+        echo "  ! /proc/meminfo below describes the HOST, not this cgroup."
+        echo "    $(mb MemTotal) MB total is not yours; ${cap} MB is."
+    elif [ -n "$CG_VERSION" ]; then
+        echo "  cgroup      none  (${CG_VERSION} present, no memory limit set)"
+    else
+        echo "  cgroup      no memory controller found"
+    fi
+
+    ul=$(ulimit_v_mb)
+    if [ -n "$ul" ]; then
+        echo "  ! ulimit -v  ${ul} MB — a per-process cap, inherited from your shell"
+        echo "    it stops this process long before the machine runs out"
+    else
+        echo "  ulimit -v   unlimited"
+    fi
+
+    # overcommit_memory=0 is the default heuristic: a large allocation is
+    # granted on a guess, and the reckoning arrives later as an OOM kill rather
+    # than as a failed malloc. It is why `push` touches every page it asks for.
+    local om or_
+    om=$(cat /proc/sys/vm/overcommit_memory 2>/dev/null || echo '?')
+    or_=$(cat /proc/sys/vm/overcommit_ratio 2>/dev/null || echo '?')
+    case "$om" in
+        0) echo "  overcommit  0 heuristic — allocations are granted on a guess," ;;
+        1) echo "  overcommit  1 always — every allocation succeeds; the OOM killer is the only limit," ;;
+        2) echo "  overcommit  2 strict (ratio ${or_}%) — allocation fails honestly instead of killing later," ;;
+        *) echo "  overcommit  ${om}" ;;
+    esac
+    [ "$om" != "?" ] && echo "              so RSS is the number to trust, not what a process asked for"
+
+    # ── what it says it has ──
+    total=$(mb MemTotal); avail=$(avail_meminfo_mb)
+    swap_total=$(mb SwapTotal); swap_free=$(mb SwapFree)
+    echo
+    echo "memory"
+    echo "  total       ${total} MB"
+    echo "  available   ${avail} MB"
+    echo "  swap        ${swap_total} MB  ($(( swap_total - swap_free )) MB used)"
+    if [ "$swap_total" -eq 0 ]; then
+        echo "  - no swap: this box has no cushion. It goes from fine to OOM-killed"
+        echo "    with nothing in between, which is the abrupt failure you get in a VM."
+    fi
+
+    # postgres puts its shared buffers in /dev/shm. Docker's default is 64 MB,
+    # and the resulting failure names neither shm nor the size.
+    if [ -d /dev/shm ]; then
+        local shm; shm=$(df -Pm /dev/shm 2>/dev/null | awk 'NR==2{print $2}')
+        if [ -n "$shm" ]; then
+            if [ "$shm" -le 64 ]; then
+                echo "  ! /dev/shm   ${shm} MB — postgres puts shared memory here and 64 MB"
+                echo "    is docker's default. Raise it with --shm-size when the cabinet fails."
+            else
+                echo "  /dev/shm    ${shm} MB"
+            fi
+        fi
+    fi
+
+    echo
+    echo "disk"
+    local d
+    for d in / /tmp /var/lib/docker; do
+        [ -d "$d" ] || continue
+        df -Pm "$d" 2>/dev/null | awk -v p="$d" 'NR==2{printf "  %-12s %s MB free of %s MB\n", p, $4, $2}'
+    done
+
+    # kind and Tilt both watch large trees, and the failure mode is silent:
+    # they simply stop noticing file changes. Cheap to report while we are here.
+    local w i
+    w=$(cat /proc/sys/fs/inotify/max_user_watches 2>/dev/null || echo 0)
+    i=$(cat /proc/sys/fs/inotify/max_user_instances 2>/dev/null || echo 0)
+    echo
+    echo "tooling"
+    echo "  inotify     watches=$w instances=$i"
+    if [ "$w" -lt 524288 ] || [ "$i" -lt 512 ]; then
+        echo "  ! low — anything watching files will silently stop seeing changes"
+    fi
+
+    if ! command -v docker >/dev/null 2>&1; then
+        if [ -S /var/run/docker.sock ]; then
+            echo "  docker      socket present, no cli"
+        else
+            echo "  docker      not installed"
+        fi
+    elif docker info >/dev/null 2>&1; then
+        local n
+        n=$(docker ps -q 2>/dev/null | wc -l)
+        echo "  docker      $(docker version --format '{{.Server.Version}}' 2>/dev/null), ${n} container(s) running"
+    else
+        echo "  ! docker     cli present but the daemon is unreachable"
+    fi
+
+    # WSL keeps its cap on the Windows side, in a file this shell can read but
+    # not usefully apply — the change costs a full VM restart. Report it, and
+    # report the commonest mistake, which is editing it and not restarting.
+    if is_wsl; then
+        local cfg conf
+        cfg=$(wslconfig_path)
+        echo
+        echo "wsl"
+        if [ -z "$cfg" ]; then
+            echo "  ! cannot tell which Windows profile owns .wslconfig"
+        else
+            echo "  config      $cfg"
+            conf=$(sed -n 's/^[[:space:]]*memory[[:space:]]*=[[:space:]]*//p' "$cfg" 2>/dev/null \
+                   | tail -1 | tr -d '[:space:]')
+            if [ -n "$conf" ]; then
+                echo "  configured  $conf   (booted ${total} MB)"
+                echo "  - if those disagree the edit has not been applied."
+                echo "    From a WINDOWS terminal:  wsl --shutdown"
+            else
+                echo "  configured  no memory= set (WSL defaults to half the host RAM, or 8 GB,"
+                echo "              whichever is less — which is where your Airflow ceiling comes from)"
+            fi
+        fi
+    fi
+
+    echo
+    echo "effective ceiling  $(effective_ceiling_mb) MB"
+    echo "  the lowest of MemTotal, the cgroup cap and ulimit -v. What the box"
+    echo "  claims. 'push' measures what it will actually hand over."
+
+    [ "$avail" -lt $(( total / 5 )) ] && { echo; hogs; }
+    return 0
+}
+
+# ── push ───────────────────────────────────────────────────────────────────
+
+STATE=""
+CHILD=""
+
+cleanup() {
+    if [ -n "$CHILD" ] && kill -0 "$CHILD" 2>/dev/null; then
+        kill -KILL "$CHILD" 2>/dev/null || true
+        wait "$CHILD" 2>/dev/null || true
+    fi
+    [ -n "$STATE" ] && rm -f "$STATE"
+    return 0
+}
+
+# The child allocates and stops itself; the parent only watches. That split is
+# the point: under --to-oom the allocating process is expected to be killed, and
+# something has to survive to say how far it got.
+allocator() {
+    # Raise our own OOM score to the maximum so the kernel picks THIS process
+    # first. Raising needs no privilege (only lowering does). Without it, the
+    # kernel is free to choose your shell, your ssh session or dockerd — on a
+    # box you are still using, that is not an acceptable coin toss.
+    echo 1000 > "/proc/$BASHPID/oom_score_adj" 2>/dev/null || true
+
+    local arr=() held=0 i=0 rss swapped avail first_swap=0
+    local bytes=$((STEP_MB * 1024 * 1024))
+    local swap_used_start
+    swap_used_start=$(( $(mb SwapTotal) - $(mb SwapFree) ))
+
+    while :; do
+        # Written STRAIGHT INTO the array element. The obvious spelling —
+        # build one chunk and `arr+=("$chunk")` — costs three copies per step,
+        # not one: the template stays resident, expanding "$chunk" makes a
+        # temporary word, and the append makes the element. A 128 MB step then
+        # needs 384 MB transiently, and on a small box it is killed on the
+        # first append while reporting a third of the true ceiling.
+        #
+        # printf -v into a subscript also means every page is written, so it is
+        # resident rather than merely promised — the only kind of allocation
+        # that measures anything under heuristic overcommit.
+        printf -v "arr[$i]" '%*s' "$bytes" ''
+        i=$((i + 1)); held=$((held + STEP_MB))
+
+        rss=$(awk '/^VmRSS:/{print int($2/1024)}' "/proc/$BASHPID/status" 2>/dev/null || echo 0)
+        avail=$(headroom_mb)
+        swapped=$(( $(mb SwapTotal) - $(mb SwapFree) - swap_used_start ))
+        [ "$swapped" -lt 0 ] && swapped=0
+
+        printf '%8s MB held   rss %7s MB   headroom %7s MB   swap +%s MB\n' \
+            "$held" "$rss" "$avail" "$swapped"
+        printf '%s %s %s %s\n' "$held" "$rss" "$avail" "$swapped" >> "$STATE"
+
+        # Worth calling out separately from the ceiling: this is where the box
+        # stops being fast and starts being unusable, which for a scheduler is
+        # a different and earlier problem than being killed.
+        if [ "$swapped" -gt 0 ] && [ "$first_swap" -eq 0 ]; then
+            first_swap=$held
+            echo "  - first swap page at ${held} MB — past here it works but crawls"
+            echo "swapat $held" >> "$STATE"
+        fi
+
+        if [ -n "$TO_MB" ] && [ "$held" -ge "$TO_MB" ]; then
+            echo "stop reached-the-cap" >> "$STATE"; return 0
+        fi
+        if [ "$TO_OOM" = no ] && [ "$avail" -lt "$FLOOR_MB" ]; then
+            echo "stop floor" >> "$STATE"; return 0
+        fi
+    done
+}
+
+push() {
+    local total ceiling rc=0 last held rss swapat stop
+    total=$(mb MemTotal)
+    ceiling=$(effective_ceiling_mb)
+
+    # A step is worth about a sixty-fourth of the ceiling: enough resolution to
+    # find the edge, few enough lines to read, and small enough that the
+    # transient cost of one allocation never dominates a small box. A fixed
+    # size cannot do all three — 128 MB is fine on 16 GB and absurd on 512 MB.
+    if [ "$STEP_EXPLICIT" = no ]; then
+        STEP_MB=$(( ceiling / 64 ))
+        [ "$STEP_MB" -lt 4 ] && STEP_MB=4
+        [ "$STEP_MB" -gt 256 ] && STEP_MB=256
+    fi
+
+    # Stop with a cushion rather than riding it to the kill. How big a cushion
+    # depends on what it is protecting. Under a cgroup cap, running out kills
+    # only this container's own processes, so it need cover no more than the
+    # shell that prints the result — and a 512 MB cushion on a 1 GB box would
+    # halve the answer. On a host there is everything else to protect, and the
+    # OOM killer does not promise to pick the process that caused the problem.
+    if [ -n "$(cgroup_cap_mb)" ]; then FLOOR_MB=64; else FLOOR_MB=512; fi
+    [ $(( ceiling / 20 )) -gt "$FLOOR_MB" ] && FLOOR_MB=$(( ceiling / 20 ))
+
+    STATE=$(mktemp "${TMPDIR:-/tmp}/rigmini.XXXXXX")
+    trap cleanup EXIT
+    # INT kills the child and lets the summary below print anyway, so an
+    # impatient Ctrl-C still tells you how far it got — and, more importantly,
+    # still gives the memory back.
+    trap 'echo; echo "  interrupted"; echo "stop interrupted" >> "$STATE"; [ -n "$CHILD" ] && kill -KILL "$CHILD" 2>/dev/null || true' INT
+
+    echo "push"
+    echo "  step        ${STEP_MB} MB per allocation, every page touched"
+    echo "  ceiling     ${ceiling} MB claimed"
+    if [ -n "$TO_MB" ]; then
+        echo "  stopping    at ${TO_MB} MB (--to)"
+    elif [ "$TO_OOM" = yes ]; then
+        echo "  ! stopping   only when the kernel stops it (--to-oom)"
+        echo "    the allocating child is marked as the preferred OOM victim,"
+        echo "    but nothing about an OOM kill is entirely polite. Not on a box"
+        echo "    running anything you mind losing."
+    else
+        echo "  stopping    when headroom drops below ${FLOOR_MB} MB"
+    fi
+    echo
+
+    allocator &
+    CHILD=$!
+    wait "$CHILD" || rc=$?
+    CHILD=""
+    trap - INT
+
+    last=$(grep -E '^[0-9]' "$STATE" 2>/dev/null | tail -1 || true)
+    held=$(echo "$last" | awk '{print $1}')
+    rss=$(echo "$last"  | awk '{print $2}')
+    swapat=$(awk '/^swapat/{print $2}' "$STATE" 2>/dev/null | head -1 || true)
+    stop=$(awk '/^stop/{print $2}' "$STATE" 2>/dev/null | head -1 || true)
+
+    echo
+    if [ -z "$held" ]; then
+        echo "  ! nothing was allocated. Even one ${STEP_MB} MB chunk failed —"
+        echo "    try a smaller --step, or check ulimit -v in 'status'."
+        return 1
+    fi
+
+    echo "  reached     ${rss:-$held} MB resident"
+    [ -n "$swapat" ] && echo "  swapping    from ${swapat} MB"
+
+    case "$stop" in
+        reached-the-cap)
+            echo "  outcome     stopped at the --to cap, not at a limit."
+            echo "              The box held ${TO_MB} MB without complaint; there is more." ;;
+        floor)
+            echo "  outcome     stopped with a cushion intact, by choice."
+            echo "              The real ceiling is higher — --to-oom finds it, at the"
+            echo "              cost of an actual OOM kill." ;;
+        interrupted)
+            echo "  outcome     interrupted at ${rss:-$held} MB — where you stopped it,"
+            echo "              not where the box did." ;;
+        *)
+            # No stop line means the child did not decide to stop: it was ended.
+            if [ "$rc" -ge 128 ]; then
+                echo "  outcome     the child was killed (signal $((rc - 128))) at ${rss:-$held} MB."
+            elif [ "$rc" -ne 0 ]; then
+                echo "  outcome     the allocation failed at ${rss:-$held} MB (exit ${rc})."
+                echo "              bash could not get the next chunk — an honest malloc"
+                echo "              failure rather than a kill. That is the strict-overcommit"
+                echo "              or ulimit path."
+            else
+                echo "  outcome     ended at ${rss:-$held} MB."
+            fi
+            local ev
+            ev=$(dmesg 2>/dev/null | tail -80 | grep -iE 'oom-kill|killed process' | tail -1 || true)
+            if [ -n "$ev" ]; then
+                echo "  kernel      ${ev#*] }"
+            else
+                echo "  - dmesg is unreadable here (dmesg_restrict, or no privilege),"
+                echo "    so the kill cannot be confirmed from this side. The number stands."
+            fi ;;
+    esac
+
+    # The gap between the claim and the measurement is the finding — but only
+    # when the BOX chose where to stop. An empty $stop means the child was ended
+    # rather than deciding to end; anything else (--to, the floor) is a stop we
+    # asked for, and flagging those as short of the ceiling would put a warning
+    # on every deliberately small run.
+    local got="${rss:-$held}"
+    echo
+    if [ -z "$stop" ] && [ "$got" -lt $(( ceiling * 70 / 100 )) ]; then
+        echo "  ! claimed ${ceiling} MB, gave up ${got} MB — under 70% of it."
+        echo "    Something is taking the difference. 'status' names the candidates:"
+        echo "    a cgroup cap, ulimit -v, or memory already resident."
+    fi
+    return 0
+}
+
+# ── all ────────────────────────────────────────────────────────────────────
+
+all() {
+    status
+    echo
+    echo "────────────────────────────────────────────────────────────"
+    echo
+    push
+
+    local got budget_mb ceiling
+    budget_mb=$(( BUDGET_GB * 1024 ))
+    ceiling=$(effective_ceiling_mb)
+    got=$(grep -E '^[0-9]' "$STATE" 2>/dev/null | tail -1 | awk '{print $2}' || true)
+    [ -n "$got" ] || got=0
+
+    echo
+    echo "verdict"
+    echo "  budget      ${BUDGET_GB} GB for kind + postgres + redis + airflow"
+    # Only worth explaining while it is still a guess. Once --budget is given
+    # the number came from somewhere better than this reasoning, and repeating
+    # the derivation would describe a figure that is no longer in use.
+    if [ "$BUDGET_EXPLICIT" = no ]; then
+        echo "  - that is 2 GB per kind node, which is rig's own figure, plus about"
+        echo "    4 GB for the three cabinets. THE 4 GB IS AN ESTIMATE, not something"
+        echo "    measured. Re-run with --budget once you have watched the real thing."
+    fi
+    echo "  measured    ${got} MB handed over"
+
+    if [ "$got" -ge "$budget_mb" ]; then
+        echo "  fits, with $(( got - budget_mb )) MB spare."
+        if [ "$got" -lt $(( budget_mb * 130 / 100 )) ]; then
+            echo "  - under 30% spare is thin for a scheduler. Airflow's memory use"
+            echo "    is spiky, and the spikes are what get killed."
+        fi
+    else
+        echo "  ! short by $(( budget_mb - got )) MB."
+        if [ "$ceiling" -ge "$budget_mb" ]; then
+            echo "    The box CLAIMS enough (${ceiling} MB) but did not deliver it."
+            echo "    Free something, or read the caps section again."
+        else
+            echo "    The box does not have it to give. A bigger bundle, or a smaller"
+            echo "    profile: PROFILE=minimal drops the cabinets entirely."
+        fi
+    fi
+    return 0
+}
+
+# ── main ───────────────────────────────────────────────────────────────────
+
+parse_flags() {
+    while [ $# -gt 0 ]; do
+        case "$1" in
+            --to)      TO_MB=$(( ${2:?--to needs a value in GB} * 1024 )); shift 2 ;;
+            --to-mb)   TO_MB="${2:?--to-mb needs a value in MB}"; shift 2 ;;
+            --step)    STEP_MB="${2:?--step needs a value in MB}"; STEP_EXPLICIT=yes; shift 2 ;;
+            --to-oom)  TO_OOM=yes; shift ;;
+            --budget)  BUDGET_GB="${2:?--budget needs a value in GB}"; BUDGET_EXPLICIT=yes; shift 2 ;;
+            *) echo "unknown argument: $1" >&2; exit 1 ;;
+        esac
+    done
+    if [ "$TO_OOM" = yes ] && [ -n "$TO_MB" ]; then
+        echo "--to and --to-oom contradict each other: one stops early, the other" >&2
+        echo "refuses to stop at all. Pick one." >&2
+        exit 1
+    fi
+    return 0
+}
+
+require_linux
+find_cgroup
+
+cmd="${1:-status}"
+[ $# -gt 0 ] && shift
+
+case "$cmd" in
+    status) parse_flags "$@"; status ;;
+    push)   parse_flags "$@"; push ;;
+    all)    parse_flags "$@"; all ;;
+    *)      echo "usage: $0 [status|push|all]" >&2
+            echo "       push [--to GB] [--to-mb MB] [--step MB] [--to-oom]" >&2
+            echo "       all  [--budget GB]" >&2
+            exit 1 ;;
+esac

soleprint/artery/shunts/mercadopago/templates/index.html

@@ -1,296 +1,411 @@
 <!DOCTYPE html>
-<html lang="en">
+<!--
+  MercadoPago shunt — config UI.
+
+  Brought onto the theme: it used to carry ~30 hardcoded hexes, its own reset,
+  its own button and input rules, and zero var(--…), so it was the one page in
+  the tree that could not follow a theme at all. Everything structural now comes
+  from the baked parts; what is left below is this shunt's own.
+
+  A shunt serves this on its own port and cannot fetch soleprint's /theme.css,
+  so the theme and the parts are baked in between markers by
+  common/theme/bake.py. Run `make theme bake` after changing a class.
+
+  Only `base` and `panel` are baked here — this page has no split pane, so it
+  carries no split.css and no split.js. That is the mechanism doing its job.
+-->
+<html lang="en" data-theme="soleprint">
 <head>
     <meta charset="UTF-8">
     <meta name="viewport" content="width=device-width, initial-scale=1.0">
     <title>MercadoPago API (MOCK) - Configuration</title>
+    <!-- theme:here -->
+    <!-- theme:baked-defaults — generated by common/theme/bake.py; do not edit -->
     <style>
-        * { box-sizing: border-box; margin: 0; padding: 0; }
-        body {
-            font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif;
-            background: #111827;
-            color: #e5e7eb;
-            padding: 20px;
-        }
-        .container { max-width: 1200px; margin: 0 auto; }
-        header {
-            background: #0071f2;
-            color: white;
-            padding: 20px;
-            border-radius: 8px;
-            margin-bottom: 24px;
-        }
-        h1 { font-size: 1.5rem; font-weight: 600; margin-bottom: 8px; }
-        .subtitle { opacity: 0.9; font-size: 0.875rem; }
+      :root {
+        --accent: #d4a574;
+        --font-mono: "JetBrains Mono", "Cascadia Mono", Consolas, monospace;
+        --font-size-base: 13px;
+        --font-size-sm: 11px;
+        --font-ui: Inter, "Segoe UI", system-ui, -apple-system, Arial, sans-serif;
+        --label-spacing: 0.04em;
+        --muted: #8888a0;
+        --panel-border: 1px solid #2e2e38;
+        --panel-header-height: 36px;
+        --panel-radius: 6px;
+        --space-1: 4px;
+        --space-2: 8px;
+        --space-3: 12px;
+        --space-4: 16px;
+        --space-6: 24px;
+        --status-error: #f06565;
+        --status-escalating: #f5a623;
+        --status-idle: #555568;
+        --status-live: #3ecf8e;
+        --status-processing: #4f9cf9;
+        --surface-0: #0d0d0f;
+        --surface-1: #16161a;
+        --surface-2: #1e1e24;
+        --surface-3: #2e2e38;
+        --text-dim: #555568;
+        --text-primary: #e8e8f0;
+        --text-secondary: #8888a0;
+      }
+    </style>
+    <!-- /theme:baked-defaults -->
+    <!-- theme:parts — generated by common/theme/bake.py; do not edit -->
+    <style>
+    /* part: base — common/theme/parts/base.css */
+    * {
+      box-sizing: border-box;
+    }
+
+    html,
+    body {
+      margin: 0;
+    }
+
+    body {
+      background: var(--surface-0);
+      color: var(--text-primary);
+      font-family: var(--font-ui);
+      font-size: var(--font-size-base);
+    }
+
+    /* Opt in on <html>, <body> and the top element when the page is an app shell
+     * that should fill the viewport. Left off, the page scrolls like a document —
+     * which is what most ad-hoc vein pages actually want. */
+    .fills {
+      height: 100%;
+      width: 100%;
+    }
+
+    button {
+      font-family: var(--font-ui);
+      font-size: var(--font-size-base);
+      color: var(--text-primary);
+      background: var(--surface-2);
+      border: var(--panel-border);
+      border-radius: var(--panel-radius);
+      padding: var(--space-1) var(--space-3);
+      cursor: pointer;
+    }
+
+    button:hover:not(:disabled) {
+      background: var(--surface-3);
+    }
+
+    button:disabled {
+      opacity: 0.5;
+      cursor: default;
+    }
+
+    input,
+    select,
+    textarea {
+      font-family: var(--font-ui);
+      font-size: var(--font-size-base);
+      color: var(--text-primary);
+      background: var(--surface-0);
+      border: var(--panel-border);
+      border-radius: 4px;
+      padding: var(--space-1) var(--space-2);
+    }
+
+    ::-webkit-scrollbar {
+      width: 10px;
+      height: 10px;
+    }
+    ::-webkit-scrollbar-thumb {
+      background: var(--surface-3);
+      border-radius: 5px;
+    }
+    ::-webkit-scrollbar-track {
+      background: transparent;
+    }
+    /* part: panel — common/theme/parts/panel.css */
+    .panel {
+      position: relative;
+      background: var(--surface-1);
+      border: var(--panel-border);
+      border-radius: var(--panel-radius);
+      overflow: hidden;
+      display: flex;
+      flex-direction: column;
+    }
+
+    .panel-header {
+      display: flex;
+      align-items: center;
+      gap: var(--space-2);
+      height: var(--panel-header-height);
+      padding: 0 var(--space-3);
+      background: var(--surface-2);
+      border-bottom: var(--panel-border);
+      flex-shrink: 0;
+    }
+
+    .panel-title {
+      font-family: var(--font-ui);
+      font-size: var(--font-size-sm);
+      font-weight: 600;
+      color: var(--text-secondary);
+      text-transform: uppercase;
+      letter-spacing: 0.04em;
+    }
+
+    .panel-actions {
+      margin-left: auto;
+      display: flex;
+      align-items: center;
+      gap: var(--space-2);
+    }
+
+    .panel-status {
+      width: 8px;
+      height: 8px;
+      border-radius: 50%;
+      flex-shrink: 0;
+      background: var(--status-idle);
+    }
+    .panel-status.idle { background: var(--status-idle); }
+    .panel-status.live { background: var(--status-live); }
+    .panel-status.processing { background: var(--status-processing); }
+    .panel-status.error { background: var(--status-error); }
+
+    .panel-body {
+      flex: 1;
+      overflow: auto;
+      padding: var(--space-2);
+      min-height: 0;
+    }
+    </style>
+    <!-- /theme:parts -->
+
+    <style>
+        /* This page's own, and only its own. */
+        body { padding: var(--space-6); }
+        .container { max-width: 1100px; margin: 0 auto; }
+
+        header { margin-bottom: var(--space-6); }
+        h1 { margin: 0; font-size: 20px; font-family: var(--font-ui); }
+        .subtitle { color: var(--muted); font-size: 12px; margin-top: 4px; }
+
         .mock-badge {
-            display: inline-block;
-            background: white;
-            color: #0071f2;
-            padding: 4px 12px;
-            border-radius: 4px;
-            font-size: 0.75rem;
-            font-weight: 600;
-            text-transform: uppercase;
-            margin-left: 12px;
-        }
-        .section {
-            background: #1f2937;
-            border-radius: 8px;
-            padding: 20px;
-            margin-bottom: 20px;
-        }
-        .section-header {
-            font-size: 1.1rem;
-            font-weight: 600;
-            margin-bottom: 16px;
-            color: #f9fafb;
-        }
-        .endpoint-list { display: flex; flex-direction: column; gap: 12px; }
-        .endpoint-card {
-            background: #374151;
-            border: 2px solid transparent;
-            border-radius: 6px;
-            padding: 16px;
-            cursor: pointer;
-            transition: all 0.2s;
-        }
-        .endpoint-card:hover { border-color: #0071f2; background: #4b5563; }
-        .endpoint-card.active { border-color: #0071f2; background: #4b5563; }
-        .endpoint-method {
-            display: inline-block;
-            padding: 2px 8px;
-            border-radius: 4px;
-            font-size: 0.75rem;
-            font-weight: 600;
-            margin-right: 8px;
-        }
-        .method-post { background: #10b981; color: white; }
-        .method-get { background: #3b82f6; color: white; }
-        .endpoint-path { font-family: monospace; font-size: 0.875rem; }
-        .endpoint-desc { font-size: 0.75rem; color: #9ca3af; margin-top: 6px; }
-        .form-group { margin-bottom: 16px; }
-        .form-label {
-            display: block;
-            font-size: 0.875rem;
-            font-weight: 500;
-            margin-bottom: 6px;
-            color: #f9fafb;
+            display: inline-block; margin-left: var(--space-2);
+            padding: 2px 8px; border-radius: 3px;
+            background: var(--status-escalating); color: var(--surface-0);
+            font-size: 10px; font-weight: 600; text-transform: uppercase;
+            letter-spacing: var(--label-spacing, .08em); vertical-align: middle;
         }
-        .form-input, .form-textarea, .form-select {
-            width: 100%;
-            padding: 10px 12px;
-            background: #374151;
-            border: 1px solid #4b5563;
-            border-radius: 6px;
-            color: #e5e7eb;
-            font-size: 0.875rem;
-        }
-        .form-textarea { min-height: 200px; font-family: monospace; }
-        .form-input:focus, .form-textarea:focus, .form-select:focus {
-            outline: none;
-            border-color: #0071f2;
-        }
-        .btn {
-            padding: 10px 20px;
-            border: none;
-            border-radius: 6px;
-            font-size: 0.875rem;
-            font-weight: 600;
-            cursor: pointer;
-            transition: all 0.2s;
+
+        .panel { margin-bottom: var(--space-4); }
+        .lede { color: var(--muted); font-size: 12px; margin: 0 0 var(--space-3); }
+
+        .grid {
+            display: grid; gap: var(--space-2);
+            grid-template-columns: repeat(auto-fit, minmax(190px, 1fr));
         }
-        .btn-primary {
-            background: #0071f2;
-            color: white;
+
+        /* Selectable cards — one treatment, used by both lists. */
+        .card {
+            padding: var(--space-3); cursor: pointer; text-align: left;
+            background: var(--surface-2); border: var(--panel-border);
+            border-radius: var(--panel-radius);
         }
-        .btn-primary:hover { background: #005ac1; }
-        .btn-secondary {
-            background: #4b5563;
-            color: #e5e7eb;
-            margin-left: 8px;
+        .card:hover:not(:disabled) { border-color: var(--accent); background: var(--surface-3); }
+        .card.on { border-color: var(--accent); background: var(--surface-3); }
+        .card-name { font-weight: 600; color: var(--text-primary); margin-bottom: 4px; }
+        .card-desc { font-size: 11px; color: var(--muted); }
+
+        .verb {
+            font-size: 10px; font-weight: 600; padding: 1px 6px;
+            border-radius: 3px; border: 1px solid currentColor; margin-right: var(--space-2);
         }
-        .btn-secondary:hover { background: #6b7280; }
-        .status-grid {
-            display: grid;
-            grid-template-columns: repeat(auto-fit, minmax(150px, 1fr));
-            gap: 12px;
+        .verb.POST { color: var(--status-live); }
+        .verb.GET { color: var(--status-processing); }
+        .path { font-family: var(--font-mono); font-size: 12px; }
+
+        .field { margin-bottom: var(--space-3); }
+        .field label {
+            display: block; margin-bottom: 4px; font-size: 11px;
+            color: var(--muted); text-transform: uppercase; letter-spacing: .04em;
         }
-        .status-option {
-            background: #374151;
-            padding: 12px;
-            border-radius: 6px;
-            cursor: pointer;
-            border: 2px solid transparent;
+        .field input, .field textarea { width: 100%; font-family: var(--font-mono); font-size: 12px; }
+        .field textarea { min-height: 180px; resize: vertical; }
+
+        .actions { display: flex; gap: var(--space-2); }
+        .primary { background: var(--accent); color: var(--surface-0); border-color: transparent; }
+        .primary:hover:not(:disabled) { opacity: .9; background: var(--accent); }
+
+        .url {
+            font-family: var(--font-mono); font-size: 12px; user-select: all;
+            padding: var(--space-2) var(--space-3);
+            background: var(--surface-0); border: var(--panel-border);
+            border-radius: var(--panel-radius);
         }
-        .status-option:hover { border-color: #0071f2; }
-        .status-option.selected { border-color: #0071f2; background: #4b5563; }
-        .status-name { font-weight: 600; color: #f9fafb; margin-bottom: 4px; }
-        .status-desc { font-size: 0.75rem; color: #9ca3af; }
+        .note { color: var(--text-dim); font-size: 11px; margin: var(--space-2) 0 0; }
     </style>
 </head>
 <body>
-    <div class="container">
-        <header>
-            <h1>MercadoPago <span class="mock-badge">MOCK</span></h1>
-            <div class="subtitle">Configure mock payment responses and behavior</div>
-        </header>
-
-        <!-- Payment Status Configuration -->
-        <div class="section">
-            <div class="section-header">Default Payment Status</div>
-            <p style="color: #9ca3af; margin-bottom: 16px;">Choose what status new payments should return:</p>
-            <div class="status-grid">
-                <div class="status-option selected" onclick="selectStatus('approved')">
-                    <div class="status-name">Approved</div>
-                    <div class="status-desc">Payment successful</div>
-                </div>
-                <div class="status-option" onclick="selectStatus('rejected')">
-                    <div class="status-name">Rejected</div>
-                    <div class="status-desc">Payment failed</div>
-                </div>
-                <div class="status-option" onclick="selectStatus('pending')">
-                    <div class="status-name">Pending</div>
-                    <div class="status-desc">Awaiting confirmation</div>
-                </div>
-                <div class="status-option" onclick="selectStatus('in_process')">
-                    <div class="status-name">In Process</div>
-                    <div class="status-desc">Being processed</div>
-                </div>
-            </div>
+<div class="container">
+    <header>
+        <h1>MercadoPago <span class="mock-badge">mock</span></h1>
+        <div class="subtitle">Configure mock payment responses and behaviour</div>
+    </header>
+
+    <div class="panel">
+        <div class="panel-header">
+            <span class="panel-title">Default payment status</span>
+            <span class="panel-status live"></span>
+        </div>
+        <div class="panel-body">
+            <p class="lede">What status new payments should return.</p>
+            <div class="grid" id="statuses"></div>
         </div>
+    </div>
 
-        <!-- Endpoint Configuration -->
-        <div class="section">
-            <div class="section-header">Configure Endpoint Responses</div>
-            <div class="endpoint-list">
-                <div class="endpoint-card" onclick="selectEndpoint('POST', '/checkout/preferences', 'preference')">
-                    <div>
-                        <span class="endpoint-method method-post">POST</span>
-                        <span class="endpoint-path">/checkout/preferences</span>
-                    </div>
-                    <div class="endpoint-desc">Create payment preference (Checkout Pro)</div>
-                </div>
-                <div class="endpoint-card" onclick="selectEndpoint('POST', '/v1/payments', 'payment')">
-                    <div>
-                        <span class="endpoint-method method-post">POST</span>
-                        <span class="endpoint-path">/v1/payments</span>
-                    </div>
-                    <div class="endpoint-desc">Create payment (Checkout API)</div>
-                </div>
-                <div class="endpoint-card" onclick="selectEndpoint('GET', '/v1/payments/{id}', 'payment_get')">
-                    <div>
-                        <span class="endpoint-method method-get">GET</span>
-                        <span class="endpoint-path">/v1/payments/{id}</span>
-                    </div>
-                    <div class="endpoint-desc">Get payment details</div>
-                </div>
-                <div class="endpoint-card" onclick="selectEndpoint('POST', '/oauth/token', 'oauth')">
-                    <div>
-                        <span class="endpoint-method method-post">POST</span>
-                        <span class="endpoint-path">/oauth/token</span>
-                    </div>
-                    <div class="endpoint-desc">OAuth token exchange/refresh</div>
-                </div>
-            </div>
+    <div class="panel">
+        <div class="panel-header">
+            <span class="panel-title">Endpoint responses</span>
+            <span class="panel-actions"><span class="card-desc" id="count"></span></span>
+        </div>
+        <div class="panel-body">
+            <div class="grid" id="endpoints"></div>
         </div>
+    </div>
 
-        <!-- Response Editor -->
-        <div class="section" id="responseEditor" style="display: none;">
-            <div class="section-header">Edit Response</div>
-            <div class="form-group">
-                <label class="form-label">Endpoint</label>
-                <input class="form-input" id="endpointDisplay" readonly>
+    <div class="panel" id="editor" hidden>
+        <div class="panel-header">
+            <span class="panel-title">Edit response</span>
+            <span class="panel-status processing"></span>
+        </div>
+        <div class="panel-body">
+            <div class="field">
+                <label for="endpointDisplay">Endpoint</label>
+                <input id="endpointDisplay" readonly>
             </div>
-            <div class="form-group">
-                <label class="form-label">Mock Response (JSON)</label>
-                <textarea class="form-textarea" id="responseJson" placeholder='{"id": "123456", "status": "approved", "_mock": "MercadoPago"}'></textarea>
+            <div class="field">
+                <label for="responseJson">Mock response (JSON)</label>
+                <textarea id="responseJson"></textarea>
             </div>
-            <div class="form-group">
-                <label class="form-label">HTTP Status Code</label>
-                <input type="number" class="form-input" id="statusCode" value="200">
+            <div class="field">
+                <label for="statusCode">HTTP status code</label>
+                <input type="number" id="statusCode" value="200">
             </div>
-            <div class="form-group">
-                <label class="form-label">Delay (ms)</label>
-                <input type="number" class="form-input" id="delay" value="0">
+            <div class="field">
+                <label for="delay">Delay (ms)</label>
+                <input type="number" id="delay" value="0">
             </div>
-            <div>
-                <button class="btn btn-primary" onclick="saveResponse()">Save Response</button>
-                <button class="btn btn-secondary" onclick="closeEditor()">Cancel</button>
+            <div class="actions">
+                <button class="primary" onclick="saveResponse()">Save response</button>
+                <button onclick="closeEditor()">Cancel</button>
             </div>
+            <p class="note">Saving is not implemented yet — it was not implemented before this
+               page was rebrought onto the theme either, and pretending otherwise would be worse.</p>
         </div>
+    </div>
 
-        <!-- Quick Test -->
-        <div class="section">
-            <div class="section-header">Quick Test</div>
-            <p style="color: #9ca3af; margin-bottom: 12px;">Test endpoint URL to hit for configured responses:</p>
-            <div class="form-input" style="background: #374151; user-select: all;">
-                http://localhost:8006/v1/payments
-            </div>
+    <div class="panel">
+        <div class="panel-header"><span class="panel-title">Quick test</span></div>
+        <div class="panel-body">
+            <p class="lede">Hit this URL to get the configured responses.</p>
+            <div class="url">http://localhost:8006/v1/payments</div>
         </div>
     </div>
+</div>
 
-    <script>
-        let selectedEndpoint = null;
-        let selectedPaymentStatus = 'approved';
+<script>
+const STATUSES = [
+    { key: 'approved',   name: 'Approved',   desc: 'Payment successful' },
+    { key: 'rejected',   name: 'Rejected',   desc: 'Payment failed' },
+    { key: 'pending',    name: 'Pending',    desc: 'Awaiting confirmation' },
+    { key: 'in_process', name: 'In Process', desc: 'Being processed' },
+];
 
-        function selectStatus(status) {
-            selectedPaymentStatus = status;
-            document.querySelectorAll('.status-option').forEach(opt => opt.classList.remove('selected'));
-            event.currentTarget.classList.add('selected');
-        }
+const ENDPOINTS = [
+    { verb: 'POST', path: '/checkout/preferences', type: 'preference',  desc: 'Create payment preference (Checkout Pro)' },
+    { verb: 'POST', path: '/v1/payments',          type: 'payment',     desc: 'Create payment (Checkout API)' },
+    { verb: 'GET',  path: '/v1/payments/{id}',     type: 'payment_get', desc: 'Get payment details' },
+    { verb: 'POST', path: '/oauth/token',          type: 'oauth',       desc: 'OAuth token exchange/refresh' },
+];
 
-        function selectEndpoint(method, path, type) {
-            selectedEndpoint = {method, path, type};
-            document.querySelectorAll('.endpoint-card').forEach(c => c.classList.remove('active'));
-            event.currentTarget.classList.add('active');
-            document.getElementById('responseEditor').style.display = 'block';
-            document.getElementById('endpointDisplay').value = `${method} ${path}`;
-            document.getElementById('responseJson').value = getDefaultResponse(type);
-        }
+let paymentStatus = 'approved';
+let selected = null;
 
-        function getDefaultResponse(type) {
-            const defaults = {
-                preference: JSON.stringify({
-                    "id": "123456-pref-id",
-                    "init_point": "https://www.mercadopago.com.ar/checkout/v1/redirect?pref_id=123456",
-                    "sandbox_init_point": "https://sandbox.mercadopago.com.ar/checkout/v1/redirect?pref_id=123456",
-                    "_mock": "MercadoPago"
-                }, null, 2),
-                payment: JSON.stringify({
-                    "id": 123456,
-                    "status": selectedPaymentStatus,
-                    "status_detail": selectedPaymentStatus === 'approved' ? 'accredited' : 'cc_rejected_other_reason',
-                    "transaction_amount": 1500,
-                    "currency_id": "ARS",
-                    "_mock": "MercadoPago"
-                }, null, 2),
-                payment_get: JSON.stringify({
-                    "id": 123456,
-                    "status": "approved",
-                    "status_detail": "accredited",
-                    "transaction_amount": 1500,
-                    "_mock": "MercadoPago"
-                }, null, 2),
-                oauth: JSON.stringify({
-                    "access_token": "APP_USR-123456-mock-token",
-                    "token_type": "Bearer",
-                    "expires_in": 15552000,
-                    "refresh_token": "TG-123456-mock-refresh",
-                    "_mock": "MercadoPago"
-                }, null, 2)
-            };
-            return defaults[type] || '{}';
-        }
+const $ = (id) => document.getElementById(id);
 
-        function saveResponse() {
-            alert('Mock response saved (feature pending implementation)');
-        }
+function pick(container, el) {
+    container.querySelectorAll('.card').forEach((c) => c.classList.remove('on'));
+    el.classList.add('on');
+}
 
-        function closeEditor() {
-            document.getElementById('responseEditor').style.display = 'none';
-            selectedEndpoint = null;
-            document.querySelectorAll('.endpoint-card').forEach(c => c.classList.remove('active'));
-        }
-    </script>
+STATUSES.forEach((s, i) => {
+    const b = document.createElement('button');
+    b.className = 'card' + (i === 0 ? ' on' : '');
+    b.innerHTML = '<div class="card-name"></div><div class="card-desc"></div>';
+    b.firstChild.textContent = s.name;
+    b.lastChild.textContent = s.desc;
+    b.onclick = () => { paymentStatus = s.key; pick($('statuses'), b); };
+    $('statuses').appendChild(b);
+});
+
+ENDPOINTS.forEach((e) => {
+    const b = document.createElement('button');
+    b.className = 'card';
+    b.innerHTML = '<div><span class="verb ' + e.verb + '">' + e.verb + '</span>' +
+                  '<span class="path"></span></div><div class="card-desc"></div>';
+    b.querySelector('.path').textContent = e.path;
+    b.lastChild.textContent = e.desc;
+    b.onclick = () => {
+        selected = e;
+        pick($('endpoints'), b);
+        $('editor').hidden = false;
+        $('endpointDisplay').value = e.verb + ' ' + e.path;
+        $('responseJson').value = defaultResponse(e.type);
+    };
+    $('endpoints').appendChild(b);
+});
+
+$('count').textContent = ENDPOINTS.length + ' endpoints';
+
+function defaultResponse(type) {
+    const bodies = {
+        preference: {
+            id: '123456-pref-id',
+            init_point: 'https://www.mercadopago.com.ar/checkout/v1/redirect?pref_id=123456',
+            sandbox_init_point: 'https://sandbox.mercadopago.com.ar/checkout/v1/redirect?pref_id=123456',
+            _mock: 'MercadoPago',
+        },
+        payment: {
+            id: 123456,
+            status: paymentStatus,
+            status_detail: paymentStatus === 'approved' ? 'accredited' : 'cc_rejected_other_reason',
+            transaction_amount: 1500,
+            currency_id: 'ARS',
+            _mock: 'MercadoPago',
+        },
+        payment_get: {
+            id: 123456, status: 'approved', status_detail: 'accredited',
+            transaction_amount: 1500, _mock: 'MercadoPago',
+        },
+        oauth: {
+            access_token: 'APP_USR-123456-mock-token', token_type: 'Bearer',
+            expires_in: 15552000, refresh_token: 'TG-123456-mock-refresh', _mock: 'MercadoPago',
+        },
+    };
+    return JSON.stringify(bodies[type] || {}, null, 2);
+}
+
+function saveResponse() {
+    alert('Mock response saved (feature pending implementation)');
+}
+
+function closeEditor() {
+    $('editor').hidden = true;
+    selected = null;
+    document.querySelectorAll('#endpoints .card').forEach((c) => c.classList.remove('on'));
+}
+</script>
 </body>
 </html>

soleprint/artery/veins/jira/main.py

@@ -2,13 +2,35 @@
 Jira Vein - FastAPI app.
 """
 
+from pathlib import Path
+
 from fastapi import FastAPI
+from fastapi.responses import FileResponse, JSONResponse
 from .api.routes import router
 from .core.config import settings
 
 app = FastAPI(title="Jira Vein", version="0.1.0")
 app.include_router(router)
 
+UI = Path(__file__).parent / "ui" / "index.html"
+
+
+@app.get("/ui", include_in_schema=False)
+def ui():
+    """The vein's ad-hoc interface.
+
+    One file, served as-is. Its theme and its parts are baked in by
+    common/theme/bake.py, so this is a plain FileResponse rather than a
+    template: there is nothing left to fill in at request time, and the same
+    bytes open from a double-click when nothing is serving them.
+    """
+    if not UI.is_file():
+        return JSONResponse(
+            {"error": "no ui/index.html", "hint": "run `make theme bake` from the spr root"},
+            status_code=404,
+        )
+    return FileResponse(UI, media_type="text/html")
+
 
 if __name__ == "__main__":
     import uvicorn

soleprint/artery/veins/jira/ui/index.html

@@ -0,0 +1,496 @@
+<!DOCTYPE html>
+<!--
+  The jira vein's ad-hoc interface — the `ui/` slot veins/__init__.py has
+  declared since the beginning and no vein had filled.
+
+  It does the vein page job, which rig-ui states twice in its own comments after
+  rebuilding this look by hand: name what the thing exposes, and show what comes
+  back. Tool chrome and output are styled apart on purpose — that separation is
+  what tells you whether you are reading the tool or its result.
+
+  THREE RULES, all consequences of "a vein serves this on its own port, and it
+  must also open from a double-clicked file":
+
+    no /theme.css       an absolute path assumes a server at the root
+    no build step       no npm, no bundler, no Vue
+    no webfont          a blocked stylesheet is a stall, not a fallback
+
+  So the theme and the parts are BAKED IN, between markers, by
+  common/theme/bake.py. Everything outside those markers is this page's, to edit
+  freely. Run `make theme bake` after changing a class; `make theme check` says
+  when a copy has gone stale.
+
+  The parts are chosen by what the markup uses — panel and split here, nothing
+  else. That is the whole point: this page carries no table, no log view, no
+  uplot, no vue-flow, because it uses none of them.
+-->
+<html lang="en" data-theme="soleprint">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<title>jira — vein</title>
+<!-- theme:here -->
+<!-- theme:baked-defaults — generated by common/theme/bake.py; do not edit -->
+<style>
+  :root {
+    --font-mono: "JetBrains Mono", "Cascadia Mono", Consolas, monospace;
+    --font-size-base: 13px;
+    --font-size-sm: 11px;
+    --font-ui: Inter, "Segoe UI", system-ui, -apple-system, Arial, sans-serif;
+    --muted: #8888a0;
+    --panel-border: 1px solid #2e2e38;
+    --panel-header-height: 36px;
+    --panel-radius: 6px;
+    --space-1: 4px;
+    --space-2: 8px;
+    --space-3: 12px;
+    --space-4: 16px;
+    --status-error: #f06565;
+    --status-idle: #555568;
+    --status-live: #3ecf8e;
+    --status-processing: #4f9cf9;
+    --surface-0: #0d0d0f;
+    --surface-1: #16161a;
+    --surface-2: #1e1e24;
+    --surface-3: #2e2e38;
+    --text-dim: #555568;
+    --text-primary: #e8e8f0;
+    --text-secondary: #8888a0;
+  }
+</style>
+<!-- /theme:baked-defaults -->
+<!-- theme:parts — generated by common/theme/bake.py; do not edit -->
+<style>
+/* part: base — common/theme/parts/base.css */
+* {
+  box-sizing: border-box;
+}
+
+html,
+body {
+  margin: 0;
+}
+
+body {
+  background: var(--surface-0);
+  color: var(--text-primary);
+  font-family: var(--font-ui);
+  font-size: var(--font-size-base);
+}
+
+/* Opt in on <html>, <body> and the top element when the page is an app shell
+ * that should fill the viewport. Left off, the page scrolls like a document —
+ * which is what most ad-hoc vein pages actually want. */
+.fills {
+  height: 100%;
+  width: 100%;
+}
+
+button {
+  font-family: var(--font-ui);
+  font-size: var(--font-size-base);
+  color: var(--text-primary);
+  background: var(--surface-2);
+  border: var(--panel-border);
+  border-radius: var(--panel-radius);
+  padding: var(--space-1) var(--space-3);
+  cursor: pointer;
+}
+
+button:hover:not(:disabled) {
+  background: var(--surface-3);
+}
+
+button:disabled {
+  opacity: 0.5;
+  cursor: default;
+}
+
+input,
+select,
+textarea {
+  font-family: var(--font-ui);
+  font-size: var(--font-size-base);
+  color: var(--text-primary);
+  background: var(--surface-0);
+  border: var(--panel-border);
+  border-radius: 4px;
+  padding: var(--space-1) var(--space-2);
+}
+
+::-webkit-scrollbar {
+  width: 10px;
+  height: 10px;
+}
+::-webkit-scrollbar-thumb {
+  background: var(--surface-3);
+  border-radius: 5px;
+}
+::-webkit-scrollbar-track {
+  background: transparent;
+}
+/* part: panel — common/theme/parts/panel.css */
+.panel {
+  position: relative;
+  background: var(--surface-1);
+  border: var(--panel-border);
+  border-radius: var(--panel-radius);
+  overflow: hidden;
+  display: flex;
+  flex-direction: column;
+}
+
+.panel-header {
+  display: flex;
+  align-items: center;
+  gap: var(--space-2);
+  height: var(--panel-header-height);
+  padding: 0 var(--space-3);
+  background: var(--surface-2);
+  border-bottom: var(--panel-border);
+  flex-shrink: 0;
+}
+
+.panel-title {
+  font-family: var(--font-ui);
+  font-size: var(--font-size-sm);
+  font-weight: 600;
+  color: var(--text-secondary);
+  text-transform: uppercase;
+  letter-spacing: 0.04em;
+}
+
+.panel-actions {
+  margin-left: auto;
+  display: flex;
+  align-items: center;
+  gap: var(--space-2);
+}
+
+.panel-status {
+  width: 8px;
+  height: 8px;
+  border-radius: 50%;
+  flex-shrink: 0;
+  background: var(--status-idle);
+}
+.panel-status.idle { background: var(--status-idle); }
+.panel-status.live { background: var(--status-live); }
+.panel-status.processing { background: var(--status-processing); }
+.panel-status.error { background: var(--status-error); }
+
+.panel-body {
+  flex: 1;
+  overflow: auto;
+  padding: var(--space-2);
+  min-height: 0;
+}
+/* part: split — common/theme/parts/split.css */
+.split-pane {
+  display: flex;
+  width: 100%;
+  height: 100%;
+  min-height: 0;
+  min-width: 0;
+  overflow: hidden;
+}
+
+.split-pane.horizontal {
+  flex-direction: row;
+}
+
+.split-pane.vertical {
+  flex-direction: column;
+}
+
+.split-first,
+.split-second {
+  min-height: 0;
+  min-width: 0;
+  overflow: hidden;
+  flex: 1;
+}
+
+/* Children fill their pane. */
+.split-first > *,
+.split-second > * {
+  width: 100%;
+  height: 100%;
+}
+
+.split-divider {
+  flex-shrink: 0;
+  background: transparent;
+  transition: background 0.15s;
+  touch-action: none;
+  z-index: 10;
+}
+
+.split-divider:hover,
+.split-divider.dragging {
+  background: var(--text-dim);
+}
+
+.split-pane.horizontal > .split-divider {
+  width: 4px;
+  cursor: col-resize;
+  margin: 0 -2px;
+}
+
+.split-pane.vertical > .split-divider {
+  height: 4px;
+  cursor: row-resize;
+  margin: -2px 0;
+}
+</style>
+<script>
+/* part: split — common/theme/parts/split.js */
+(function () {
+  'use strict'
+
+  function setup(root) {
+    var divider = root.querySelector(':scope > .split-divider')
+    if (!divider) return // no divider: a fixed split, deliberately
+
+    var first = root.querySelector(':scope > .split-first')
+    var second = root.querySelector(':scope > .split-second')
+    if (!first || !second) return
+
+    var horizontal = !root.classList.contains('vertical')
+    var mode = root.dataset.mode === 'px' ? 'px' : 'ratio'
+    var anchor = root.dataset.anchor === 'second' ? 'second' : 'first'
+    var size = parseFloat(root.dataset.size)
+    if (isNaN(size)) size = 1
+    var min = parseFloat(root.dataset.min)
+    if (isNaN(min)) min = mode === 'px' ? 0 : 0.1
+    var max = parseFloat(root.dataset.max)
+    if (isNaN(max)) max = mode === 'px' ? Infinity : 10
+
+    var sized = anchor === 'second' ? second : first
+    var flexed = anchor === 'second' ? first : second
+    var dragging = false
+    var startPos = 0
+
+    function apply() {
+      flexed.style.flex = '1'
+      if (mode === 'px') {
+        sized.style.flex = '0 0 auto'
+        sized.style[horizontal ? 'width' : 'height'] = size + 'px'
+      } else {
+        sized.style.flex = String(size)
+      }
+    }
+
+    divider.addEventListener('pointerdown', function (e) {
+      dragging = true
+      startPos = horizontal ? e.clientX : e.clientY
+      divider.classList.add('dragging')
+      divider.setPointerCapture(e.pointerId)
+    })
+
+    divider.addEventListener('pointermove', function (e) {
+      if (!dragging) return
+      var pos = horizontal ? e.clientX : e.clientY
+      var delta = pos - startPos
+      startPos = pos
+
+      // Dragging right/down grows the first pane. When the SECOND pane is the
+      // anchored one, that same gesture must shrink it, so invert.
+      if (anchor === 'second') delta = -delta
+
+      // Ratio mode is unitless, so pixels are scaled into it. The two constants
+      // are the SFC's, kept rather than re-derived: they are what the existing
+      // panes were tuned against, and vertical drags cover less travel.
+      var step = mode === 'px' ? delta : delta * (horizontal ? 0.01 : 0.02)
+      size = Math.max(min, Math.min(max, size + step))
+      apply()
+    })
+
+    function end(e) {
+      if (!dragging) return
+      dragging = false
+      divider.classList.remove('dragging')
+      if (e && e.pointerId !== undefined && divider.hasPointerCapture(e.pointerId)) {
+        divider.releasePointerCapture(e.pointerId)
+      }
+    }
+    divider.addEventListener('pointerup', end)
+    divider.addEventListener('pointercancel', end)
+
+    apply()
+  }
+
+  function start() {
+    var panes = document.querySelectorAll('[data-split]')
+    for (var i = 0; i < panes.length; i++) setup(panes[i])
+  }
+
+  if (document.readyState === 'loading') {
+    document.addEventListener('DOMContentLoaded', start)
+  } else {
+    start()
+  }
+})()
+</script>
+<!-- /theme:parts -->
+
+<style>
+  /* This page's own, and only its own. */
+  body { padding: var(--space-4); }
+  .page { display: flex; flex-direction: column; gap: var(--space-3); height: calc(100vh - 2 * var(--space-4)); }
+
+  header { display: flex; align-items: baseline; gap: var(--space-3); flex-wrap: wrap; }
+  h1 { margin: 0; font-size: 20px; font-family: var(--font-ui); }
+  .sub { color: var(--muted); font-family: var(--font-mono); font-size: 12px; }
+
+  .split-pane { border: var(--panel-border); border-radius: var(--panel-radius); }
+
+  .routes { display: flex; flex-direction: column; gap: 2px; }
+  .route {
+    display: flex; align-items: center; gap: var(--space-2);
+    padding: 4px var(--space-2); border-radius: var(--panel-radius);
+    font-family: var(--font-mono); font-size: 12px;
+    background: none; border: 0; width: 100%; text-align: left;
+  }
+  .route:hover:not(:disabled) { background: var(--surface-2); }
+  .route.on { background: var(--surface-3); }
+  .verb {
+    font-size: 10px; font-weight: 600; padding: 1px 6px; border-radius: 3px;
+    border: 1px solid currentColor; flex-shrink: 0;
+  }
+  .verb.GET { color: var(--status-processing); }
+  .verb.POST { color: var(--status-live); }
+
+  .controls { display: flex; gap: var(--space-2); margin-bottom: var(--space-2); }
+  .controls input { flex: 1; font-family: var(--font-mono); font-size: 12px; }
+
+  /* OUTPUT — a raised block, monospace, selectable. It is a payload, not
+     furniture, and should not read like the tool that produced it. */
+  .out {
+    margin: 0; padding: var(--space-3);
+    background: var(--surface-0); border: var(--panel-border);
+    border-radius: var(--panel-radius);
+    font-family: var(--font-mono); font-size: 12px; line-height: 1.5;
+    white-space: pre-wrap; word-break: break-word;
+    user-select: text; min-height: 100%;
+  }
+  .out.err { color: var(--status-error); }
+  .hint { color: var(--text-dim); }
+</style>
+</head>
+
+<body>
+<div class="page">
+  <header>
+    <h1>jira</h1>
+    <span class="sub">vein &middot; stateless API connector</span>
+    <span class="sub hint" id="base"></span>
+  </header>
+
+  <div class="split-pane horizontal" data-split data-size="1" data-min="0.4" data-max="3" style="flex:1; min-height:0;">
+    <div class="split-first">
+      <div class="panel">
+        <div class="panel-header">
+          <span class="panel-title">Exposes</span>
+          <span class="panel-actions"><span class="sub" id="count"></span></span>
+        </div>
+        <div class="panel-body">
+          <div class="routes" id="routes"></div>
+        </div>
+      </div>
+    </div>
+
+    <div class="split-divider"></div>
+
+    <div class="split-second">
+      <div class="panel">
+        <div class="panel-header">
+          <span class="panel-title">Returns</span>
+          <span class="panel-actions"><button id="run" disabled>send</button></span>
+          <span class="panel-status idle" id="dot"></span>
+        </div>
+        <div class="panel-body">
+          <div class="controls">
+            <input id="arg" placeholder="pick a route" disabled>
+          </div>
+          <pre class="out hint" id="out">Pick a route on the left, then send.
+
+Opened from a file rather than served? Nothing will answer — the routes are
+still the contract, which is half of what this page is for.</pre>
+        </div>
+      </div>
+    </div>
+  </div>
+</div>
+
+<script>
+/* The vein's surface, as the page understands it. Kept here rather than fetched
+   so the page still says what the vein exposes when nothing is serving it. */
+const ROUTES = [
+  { verb: 'GET',  path: '/health',            arg: null,  hint: 'connection check' },
+  { verb: 'GET',  path: '/mine',              arg: null,  hint: 'tickets assigned to you' },
+  { verb: 'GET',  path: '/backlog',           arg: null,  hint: 'the backlog' },
+  { verb: 'GET',  path: '/sprint',            arg: null,  hint: 'the current sprint' },
+  { verb: 'GET',  path: '/ticket/{key}',      arg: 'key', hint: 'one ticket, e.g. PROJ-123' },
+  { verb: 'POST', path: '/search',            arg: 'jql', hint: 'a JQL query' },
+  { verb: 'GET',  path: '/epic/{key}/status', arg: 'key', hint: 'epic processing status' },
+];
+
+const $ = (id) => document.getElementById(id);
+let active = null;
+
+$('base').textContent = location.protocol === 'file:' ? 'not served — file://' : location.origin;
+$('count').textContent = ROUTES.length + ' routes';
+
+ROUTES.forEach((r, i) => {
+  const b = document.createElement('button');
+  b.className = 'route';
+  b.innerHTML = '<span class="verb ' + r.verb + '">' + r.verb + '</span>' +
+                '<span>' + r.path + '</span>';
+  b.title = r.hint;
+  b.onclick = () => select(i, b);
+  $('routes').appendChild(b);
+});
+
+function select(i, el) {
+  active = ROUTES[i];
+  document.querySelectorAll('.route').forEach((n) => n.classList.remove('on'));
+  el.classList.add('on');
+  $('arg').disabled = !active.arg;
+  $('arg').placeholder = active.arg ? active.hint : 'no argument';
+  $('arg').value = '';
+  $('run').disabled = false;
+}
+
+function status(state) { $('dot').className = 'panel-status ' + state; }
+
+$('run').onclick = async () => {
+  if (!active) return;
+  status('processing');
+  $('out').className = 'out';
+  $('out').textContent = '…';
+  let path = active.path, init = { method: active.verb };
+  const v = $('arg').value.trim();
+  if (active.arg === 'key') path = path.replace('{key}', encodeURIComponent(v));
+  if (active.arg === 'jql') {
+    init.headers = { 'Content-Type': 'application/json' };
+    init.body = JSON.stringify({ jql: v });
+  }
+  try {
+    const res = await fetch(path, init);
+    const text = await res.text();
+    let body = text;
+    try { body = JSON.stringify(JSON.parse(text), null, 2); } catch (_) {}
+    $('out').textContent = res.status + ' ' + res.statusText + '\n\n' + body;
+    status(res.ok ? 'live' : 'error');
+    if (!res.ok) $('out').className = 'out err';
+  } catch (e) {
+    $('out').className = 'out err';
+    $('out').textContent = String(e) +
+      (location.protocol === 'file:' ? '\n\nThis page is not being served.' : '');
+    status('error');
+  }
+};
+</script>
+</body>
+</html>

soleprint/atlas2/docgen/.gitignore

@@ -0,0 +1,5 @@
+# Everything this makes.
+out/
+__pycache__/
+*.pyc
+.venv/

soleprint/atlas2/docgen/Makefile

@@ -0,0 +1,179 @@
+# docgen — code to diagram, and to everything else the IR can feed.
+#
+# Derived from where this file sits, so the folder can be copied anywhere and
+# renamed and still work. The logic lives in the Python, never here: every target
+# is one line calling `python3 -m docgen <command>`.
+#
+#   make sync                      create .venv with every optional group (uv)
+#   make book SRC=../station       the whole operation, measured at both ends
+#   make run CONFIG=docgen.toml    every book a run file lists
+#   make check                     prove docgen, on a tree it builds itself
+#   make check BOOK=out/book/x     prove one book — its own level
+#   make ir SRC=../station         extract -> out/ir.json  (one step, on its own)
+#   make self                      docgen's book of itself, then check it
+#   make doctor                    what this machine has
+#
+# Every step target still works alone — that is the property the book spine
+# exists to preserve, not to replace. The steps compose by hand too:
+#
+#   python3 -m docgen extract python --root SRC -o ir.json
+#   python3 -m docgen view ir.json --overview -o view.json
+#   python3 -m docgen emit dot view.json -o graph.svg --theme dark
+
+HERE   := $(patsubst %/,%,$(dir $(abspath $(lastword $(MAKEFILE_LIST)))))
+PKG    := $(notdir $(HERE))
+PARENT := $(patsubst %/,%,$(dir $(HERE)))
+VENV_PY := $(HERE)/.venv/bin/python
+# The synced environment when there is one (`make sync`), the system Python when
+# there is not — so `make check` still works with nothing installed at all.
+PY     ?= $(if $(wildcard $(VENV_PY)),$(VENV_PY),python3)
+CLI    := PYTHONPATH=$(PARENT) $(PY) -m $(PKG)
+
+OUT    ?= $(HERE)/out
+SRC    ?=
+SCHEMA ?=
+OPENAPI ?=
+HAR    ?=
+STYLE  ?= lucid
+THEME  ?=
+SCALE  ?= 0.55
+BOOK   ?=
+SLUG   ?=
+# NOT `LANG`: that is the shell's locale variable, so `?=` inherits
+# en_US.UTF-8 from the environment and --reader rejects it.
+READER ?= python
+OVERLAY ?=
+CONFIG ?= docgen.toml
+ONLY   ?=
+CHECK  ?=
+
+comma  := ,
+THEME_ARG := $(if $(THEME),--theme $(THEME))
+STYLE_ARGS := --style $(STYLE) $(THEME_ARG)
+SLUG_ARG  := $(if $(SLUG),--slug $(SLUG))
+OVER_ARG  := $(if $(OVERLAY),--overlay $(OVERLAY))
+ONLY_ARGS := $(foreach n,$(subst $(comma), ,$(ONLY)),--only $(n))
+
+.PHONY: help sync lock book run check ir db code view graph index site minimap explore docs self doctor clean
+
+help:  ## List every target
+	@echo "docgen — static analysis of a tree, and the artifacts that fall out of it"
+	@echo
+	@grep -E '^[a-z-]+:.*?## .*$$' $(MAKEFILE_LIST) \
+	  | awk 'BEGIN{FS=":.*?## "}{printf "  \033[1m%-10s\033[0m %s\n", $$1, $$2}'
+	@echo
+	@echo "  SRC=/path/to/tree     what to read           OUT=/path   where output goes"
+	@echo "  SCHEMA=schema.json    a database instead     STYLE=lucid THEME=dark|lucid"
+	@echo "  OPENAPI=spec.yaml     an API document        HAR=session.har  a recording"
+	@echo "  BOOK=/path            where a book goes, and which book to check"
+	@echo "  READER=python|code    ast, or tree-sitter    SLUG=name   what to call the book"
+	@echo "  OVERLAY=overlay.json  hand-written notebook additions, re-applied every build"
+	@echo "  CONFIG=docgen.toml    a run file             ONLY=a,b    just these books"
+	@echo "  CHECK=1               with run: each book's own level after building it"
+	@echo
+	@echo "  Three levels of test, by what they assert about:"
+	@echo "    make doctor              the machine. Never fails."
+	@echo "    make check               docgen. Exits 1."
+	@echo "    make check BOOK=<dir>    that book. Exits 1."
+
+sync:  ## Create .venv with every optional group, from uv.lock
+	@command -v uv >/dev/null || { echo "Error: uv is not installed — docgen still runs on the system python3" >&2; exit 1; }
+	@cd $(HERE) && uv sync --all-groups
+
+lock:  ## Re-resolve uv.lock after editing pyproject.toml
+	@cd $(HERE) && uv lock
+
+book:  ## SRC (or SCHEMA/OPENAPI/HAR) -> one operation, measured at both ends
+	@test -n "$(SRC)$(SCHEMA)$(OPENAPI)$(HAR)" \
+	  || { echo "Error: set SRC=/path/to/tree (or SCHEMA=, OPENAPI=, HAR=)" >&2; exit 1; }
+	@$(CLI) book \
+	  $(if $(SRC),--root "$(SRC)" --reader $(READER)) \
+	  $(if $(SCHEMA),--schema "$(SCHEMA)") \
+	  $(if $(OPENAPI),--openapi "$(OPENAPI)") \
+	  $(if $(HAR),--har "$(HAR)") \
+	  -o "$(if $(BOOK),$(BOOK),$(OUT)/book)" \
+	  $(STYLE_ARGS) $(SLUG_ARG) $(OVER_ARG)
+
+run:  ## CONFIG (a run file) -> every book it lists; ONLY=a,b for some
+	@$(CLI) run "$(CONFIG)" $(ONLY_ARGS) $(if $(CHECK),--check)
+
+check:  ## Prove docgen (or one book, with BOOK=<dir>)
+	@$(CLI) check $(if $(BOOK),"$(BOOK)")
+
+ir:  ## Extract SRC into OUT/ir.json
+	@test -n "$(SRC)" || { echo "Error: set SRC=/path/to/tree" >&2; exit 1; }
+	@$(CLI) extract python --root "$(SRC)" -o $(OUT)/ir.json
+	@$(CLI) validate $(OUT)/ir.json
+
+code:  ## Extract C#/TypeScript from SRC (needs the `code` group)
+	@test -n "$(SRC)" || { echo "Error: set SRC=/path/to/tree" >&2; exit 1; }
+	@$(CLI) extract code --root "$(SRC)" -o $(OUT)/ir.json
+	@$(CLI) validate $(OUT)/ir.json
+
+db:  ## Extract a graphgen-compatible SCHEMA into OUT/ir.json
+	@test -n "$(SCHEMA)" || { echo "Error: set SCHEMA=/path/to/schema.json" >&2; exit 1; }
+	@$(CLI) extract db --schema "$(SCHEMA)" -o $(OUT)/ir.json
+	@$(CLI) validate $(OUT)/ir.json
+
+view:  ## OUT/ir.json -> OUT/view.json, the default view for its source type
+	@$(CLI) view $(OUT)/ir.json --overview -o $(OUT)/view.json
+
+graph: view  ## OUT/view.json -> whatever its structure asks for
+	@$(CLI) emit auto $(OUT)/view.json -o $(OUT) $(STYLE_ARGS)
+
+index:  ## OUT/ir.json -> OUT/index.md and OUT/sidebar.json
+	@$(CLI) emit index $(OUT)/ir.json -o $(OUT)/index.md
+	@$(CLI) emit index $(OUT)/ir.json -o $(OUT)/sidebar.json
+
+site: view  ## OUT/view.json -> a self-contained docs site in OUT/site
+	@$(CLI) emit site $(OUT)/view.json -o $(OUT)/site $(STYLE_ARGS)
+	@echo "  open $(OUT)/site/index.html"
+
+minimap:  ## OUT/ir.json -> OUT/minimap.svg — what is where, read from the colours
+	@$(CLI) emit minimap $(OUT)/ir.json -o $(OUT)/minimap.svg $(STYLE_ARGS) --scale $(SCALE)
+
+explore: ## OUT/ir.json -> OUT/explore/ — navigate on one side, explore on the other
+	@$(CLI) emit explore $(OUT)/ir.json -o $(OUT)/explore $(STYLE_ARGS) --scale $(SCALE)
+	@echo "  open $(OUT)/explore/explore.html"
+
+docs:  ## Regenerate the figures in docs/ — docgen documented by docgen
+	@mkdir -p $(HERE)/docs/img
+	@$(CLI) extract python --root $(HERE) -o /tmp/$(PKG)-docs.json >/dev/null
+	@$(CLI) view /tmp/$(PKG)-docs.json --overview -o /tmp/$(PKG)-docs-view.json >/dev/null
+	@$(CLI) emit dot /tmp/$(PKG)-docs-view.json -o $(HERE)/docs/img/architecture.svg -q
+	@$(CLI) emit minimap /tmp/$(PKG)-docs.json -o $(HERE)/docs/img/minimap.svg --scale 0.5 --width 860
+	@$(CLI) emit erd $(OUT)/ir.json -o $(HERE)/docs/img/erd.svg 2>/dev/null \
+	  || echo "  (erd figure kept — needs a schema IR at $(OUT)/ir.json to refresh)"
+	@PYTHONPATH=$(PARENT) $(PY) -c "from $(PKG).emitters.site import VIEWER, _slots, _fill; \
+from $(PKG).style import Style; import pathlib; \
+pathlib.Path('$(HERE)/docs/viewer.html').write_text( \
+  _fill(VIEWER.replace('__TITLE__', 'docgen docs'), _slots(Style.load('lucid'))))"
+	@echo "  open $(HERE)/docs/index.html"
+
+self:  ## docgen's book of the widest tree it can see, then check it
+	@$(eval SELF_SRC := $(shell PYTHONPATH=$(PARENT) $(PY) -c "from $(PKG) import reference; \
+r = reference.root(); print(r if r else '$(HERE)')"))
+	@echo "  self-hosting on $(SELF_SRC)"
+	@$(MAKE) --no-print-directory book SRC=$(SELF_SRC) SLUG=self \
+	  BOOK=$(OUT)/book/self OUT=$(OUT)
+	@echo
+	@$(MAKE) --no-print-directory check BOOK=$(OUT)/book/self
+
+doctor:  ## Report whether this machine can run it
+	@printf 'python   : %s ' '$(PY)'; $(PY) --version 2>&1 || echo MISSING
+	@printf 'uv       : '; if ! command -v uv >/dev/null; then echo 'absent — fine; docgen runs on the system python3'; \
+	  elif [ -x $(VENV_PY) ]; then echo "$$(uv --version), .venv synced"; \
+	  else echo "$$(uv --version), .venv not synced — make sync for the optional groups"; fi
+	@printf 'dot      : '; (dot -V 2>&1) || echo 'MISSING — sudo apt install graphviz (only to render)'
+	@printf 'tree-sit : '; $(PY) -c 'import tree_sitter, tree_sitter_c_sharp, tree_sitter_typescript; print("ok — C# and TypeScript available")' 2>/dev/null || echo 'absent — Python only. make sync, or the `code` group'
+	@printf 'lxml     : '; $(PY) -c 'import lxml; print("ok — theme harvesting available")' 2>/dev/null || echo 'absent — only used to harvest a theme'
+	@printf 'yaml     : '; $(PY) -c 'import yaml; print("ok — needed only to read OpenAPI")' 2>/dev/null || echo 'absent — only used by the OpenAPI reader'
+	@printf 'networkx : '; $(PY) -c 'import networkx; print(networkx.__version__ + " — for lab/ experiments")' 2>/dev/null || echo 'absent — only used in lab/'
+	@printf 'reference: '; PYTHONPATH=$(PARENT) $(PY) -c "from $(PKG) import reference; print(reference.describe())"
+	@printf 'package  : %s (from %s)\n' '$(PKG)' '$(PARENT)'
+	@printf 'styles   : '; PYTHONPATH=$(PARENT) $(PY) -c "from $(PKG).style import Style; print(', '.join(Style.available()))"
+	@PYTHONPATH=$(PARENT) $(PY) -c "import $(PKG).ir, $(PKG).emitters.dot, $(PKG).ops, $(PKG).cli" >/dev/null 2>&1 \
+	  && echo 'import   : ok' || echo 'import   : FAILED — is the folder intact?'
+
+clean:  ## Delete OUT. Nothing else is ever written to
+	@rm -rf "$(OUT)" && echo "Removed $(OUT)"

soleprint/atlas2/docgen/README.md

@@ -0,0 +1,276 @@
+# docgen
+
+Static analysis of a tree, and the artifacts that fall out of it.
+
+The point is not the diagram. The point is the format in the middle — diagrams
+are one consumer of it, and not the one that reaches the most people.
+
+```
+extractors/        →   graph IR (JSON)   →   emitters/
+(per source type)      (one schema)          (per output target)
+                              ↑
+                         style/*.json
+                    (consumed by emitters only)
+```
+
+```bash
+make sync                       # optional: .venv with every group, from uv.lock
+make check                      # prove it, on a tree it builds itself
+make book SRC=/path/to/repo     # one whole operation, measured at both ends
+make run CONFIG=docgen.toml     # every book a run file lists
+make help
+```
+
+The full documentation is `docs/index.html` — open it in a browser.
+
+## Layout
+
+```
+cli/            the command line — every command, and nothing else
+book/           an operation: larder measure, steps, web output, run files
+extractors/     source -> IR, one per source type
+ir/             the contract: schema.json, the dataclasses, the validator
+ops/            IR -> a smaller IR
+emitters/       IR -> an artifact
+notebook/       the notebook spec, before it is an .ipynb
+style/          slots, themes, and harvesting a theme from real diagrams
+fixtures/       docgen's own test inputs
+lab/            sanctioned experiments; nothing imports it
+reference.py    the one seam to the repo above, for reading OpenAPI
+```
+
+Library packages hold **no command-line code** — no argparse, no `__main__.py`,
+no `cli_dot.py` beside `dot.py`. It all lives in `cli/`, behind one entry point,
+and the selftest asserts it stays there.
+
+`pyproject.toml` declares no required dependency: the structural path is the
+stdlib, Python 3.11+. The optional groups (`code`, `openapi`, `harvest`, `lab`)
+are pinned in `uv.lock`; `[tool.uv] package = false`, as dataconvert does, since
+this is a folder run in place rather than something to install.
+
+## Run files
+
+```toml
+# docgen.toml — beside the project it describes; paths relative to this file
+[defaults]
+out = "out/book"
+
+[[book]]
+name = "station"
+root = "../soleprint/station"
+
+[[book]]
+name   = "shop"
+schema = "schemas/shop.json"
+```
+
+`make run CONFIG=docgen.toml [ONLY=station] [CHECK=1]`. One failing book never
+stops the others; unknown keys are refused; a rebuild replaces the previous
+build's outputs and never touches a hand-written `checks.py`.
+`docgen.example.toml` runs against the shipped fixtures.
+
+Or as three composable commands, which is what the Makefile is wrapping — run
+from the directory above this one:
+
+```bash
+python3 -m docgen extract python --root SRC -o ir.json
+python3 -m docgen view ir.json --drop-stdlib -o view.json
+python3 -m docgen emit dot view.json -o graph.svg --theme dark
+```
+
+## DOT collapses three concerns; this separates them
+
+| concern | question | owner |
+|---|---|---|
+| **structure** | what the graph *is* | `ir/schema.json` — versioned, golden-tested |
+| **meaning** | what things *mean visually* | `style/*.json`, keyed on `kind` |
+| **placement** | where things *go* | Graphviz defaults. Phase two |
+
+An extractor has never heard of SVG, colours or layout. An emitter has never
+heard of Python, `ast` or SQL. **The IR carries no visual information** — if a
+field would change between light and dark theme, it does not belong in it.
+`shape="cylinder"` is not a field; it is `kind="datastore"` plus a style rule,
+which is what lets the same IR render in a theme that has no cylinders.
+
+The selftest asserts all three of those, because they are the design rather than
+a nicety and they are exactly what erodes first.
+
+## The IR
+
+```json
+{
+  "meta":  { "source": "python", "root": "app/", "schema_version": "1" },
+  "nodes": [ { "id": "app.models.User", "kind": "class", "label": "User",
+               "parent": "app.models",
+               "attrs": { "file": "app/models.py", "line": 12 } } ],
+  "edges": [ { "source": "app.models.User", "target": "app.db.Base",
+               "kind": "inherits", "attrs": {} } ]
+}
+```
+
+- **`id`** is fully qualified and **stable across runs**. That is what makes two
+  graphs from two commits diffable.
+- **`kind`** is the hinge, and the only field style and layout may key on.
+- **`parent`** is containment. Relationships are edges.
+- **`attrs`** is an open bag; `file`/`line` let a UI link a box to a line.
+
+Stdlib dataclasses, not Pydantic. A format that needs a library installed to be
+opened is not a format, it is an API. `ir/validate.py` is the check at the
+boundary, and it reads the field lists out of `schema.json` so the two cannot
+drift.
+
+```bash
+python3 -m docgen validate ir.json
+```
+
+It catches what a schema cannot: an edge naming a node that does not exist, a
+containment cycle, duplicate ids, and a visual field smuggled into `attrs`.
+
+## Extraction is deterministic
+
+**No LLM in the structural path.** A diagram from an AST cannot be out of date
+with the code; one from a model's reading of the code is wrong the moment the
+model has a bad day, which is the problem this exists to fix.
+
+`ast` resolves nothing on its own — `class User(Base)` yields the literal string
+`"Base"`. So there are two passes: one collects each module's definitions and
+imports, the other resolves names against those tables.
+
+```
+from .db import Base ;  class User(Base)
+    →  app.models.User --inherits--> app.db.Base        not "Base"
+```
+
+**Unresolved names become `kind: "external"` nodes and keep their edges.**
+Dropping them is the worse failure: the diagram looks complete and has quietly
+lost a dependency. Gathered by the index emitter, they *are* the project's
+dependency surface.
+
+An unparseable file is recorded as a node with an `error` attr, not a crash —
+one bad file must not cost you the other four hundred.
+
+`calls` edges are deliberately **not** attempted. Resolving `self.foo()` needs
+type inference, and a call graph that is quietly 60% right is worse than none
+because it reads as authoritative.
+
+### A second source
+
+`extractors/db.py` reads the published `{models, relationships, source}`
+contract that `modelgen` already emits and `graphgen` already consumes. Tables
+become nodes, columns become contained nodes, foreign keys become edges — with
+no new top-level field, which was the checkpoint on whether the schema was right.
+
+Connecting to a live database is not here. `modelgen from-db --url ...` does
+that and writes the schema this reads; the two-step also keeps credentials out
+of this pipeline entirely.
+
+## Views are not an emitter concern
+
+The first real diagram out of this pipeline was a 3000px strip: four modules of
+content and sixty `sys`/`json`/`typing` boxes, all peers. The emitter was
+correct and the picture was useless. That is a **missing view**, and the fix
+belongs to every consumer at once — the index, the diagram and the diff all want
+"just this subsystem, two hops out, without the stdlib".
+
+```bash
+python3 -m docgen view ir.json --drop-stdlib --around docgen.ir --hops 2 -o view.json
+```
+
+`drop_stdlib`, `drop_external`, `only_kinds`, `drop_kinds`, `subtree`,
+`neighbourhood`, `collapse_to_depth`. All IR→IR, all composable, each producing
+a document that still validates.
+
+Graph *algorithms* are not here. Transitive reduction, cycle detection and
+dominators are `networkx`'s, and reimplementing them is the classic way to
+acquire a quiet bug. `lab/` is where that dependency gets tried against real IRs
+before anything depends on it — the aim being to learn which part of it is
+actually attractive, rather than adopting all of it on faith.
+
+## One colour language
+
+A style rule names a **slot**, never a colour. `"border": "atlas"` is the rule;
+the theme binds `atlas` to `#43A047` in print and `#15803d` on the docs site.
+
+That indirection is the whole point. `common/theme/tokens.css`,
+`docs/graphs/themes/*.gvpr` and `style/lucid.json` use the same slot names, so a
+diagram and the page around it match by construction — which is the rule
+`docs/graphs/README.md` already states. The `dark` theme's `artery`, `atlas` and
+`station` slots are exactly the `--system-accent` values set in
+`artery/index.html:30`, `atlas/index.html:25` and `station/index.html:29`, and
+the selftest fails if they drift apart.
+
+An unknown `kind` falls back to `default` rather than crashing, so a new
+extractor renders plainly and legibly on day one instead of needing a style file
+written first.
+
+**How a container picks its colour without the IR naming one:** it does not. The
+IR says which spr model a group belongs to (`attrs.domain` — semantic), and
+`domain_slots` maps that to a slot. Same mechanism as `--system-accent`. With no
+domain, the emitter assigns by sorted id, so two runs agree.
+
+## Use DOT until it hits its limits
+
+The emitter writes what DOT expresses natively and stops at the boundary rather
+than growing machinery. The limits are recorded in `style/lucid.json` under
+`limits` and reachable as `Style.limits()`:
+
+| | |
+|---|---|
+| header bars | a cluster has a label and a fill, not a 100%-width header rectangle |
+| `stroke-dasharray` | not parameterised — `4,4` and `5,5` collapse to one dash |
+| corner radius | `rounded` is binary, so 4px and 6px are identical |
+| icon above label | needs an HTML-like label table |
+| sequence badges | `xlabel` carries the number; the circle does not exist |
+
+Those mark where a richer emitter would begin. The style file carries the full
+spec regardless, so that emitter needs no re-authoring.
+
+One limit that *was* worth solving: DOT cannot use a cluster as an edge
+endpoint, so every module-to-module import silently vanished. The native answer
+is `compound=true` with `lhead`/`ltail` — draw between a representative leaf and
+clip at the cluster border.
+
+## The output is addressable
+
+`id` and `kind` pass through to the SVG as the element's `id` and `class`, and
+`attrs.file`/`attrs.line` become an `href`. A front end can bind behaviour to a
+box and a box can link to the line it came from, without the emitter knowing
+about either.
+
+## Testing
+
+```bash
+make check                  # docgen's own suite: 287 after make sync, 272 with nothing
+make check BOOK=out/book/x  # one book's own level — generated and hand-written checks
+make doctor                 # the machine; never fails
+```
+
+**Golden tests go on the IR, never on the SVG.** Graphviz measures label text
+with the host's fonts to size nodes, so identical input gives different geometry
+on a machine with different fontconfig. The IR is deterministic; the SVG is not.
+
+Self-hosting is the honest end-to-end check, and it is where the real bugs came
+from — two name-resolution faults that no fixture had reached:
+
+```bash
+make self      # docgen's book of the widest tree it can see, then its checks
+```
+
+## Where this sits
+
+`docgen` belongs to Atlas — documentation is whose concern it is. It is **not** a
+station tool and is not under `station/tools/`; it *may depend on* station tools,
+which is the permitted direction.
+
+Atlas 2 is a successor, not a replacement. `soleprint/atlas/` is untouched: it
+carries client information and an idea still worth extracting — deriving frontend
+and backend tests from one source, which is the same shape as this pointed the
+other way.
+
+## Not here
+
+No layout system, no positioning, no ELK. No HTML-like labels, no SVG post-pass.
+No LLM in the structural path — annotation (summarising a module, naming a
+cluster) is a later layer, cached to its own file keyed by node `id`, merged into
+`attrs` at emit time, and extraction must work with it absent. No configuration
+knobs until two real consumers disagree.

soleprint/atlas2/docgen/__init__.py

@@ -0,0 +1 @@
+"""Docgen — code to diagram. The IR is the product; diagrams are one consumer."""

soleprint/atlas2/docgen/__main__.py

@@ -0,0 +1,8 @@
+"""    python3 -m docgen <command> [args]   — see cli/__init__.py for the commands."""
+
+import sys
+
+from .cli import main
+
+if __name__ == "__main__":
+    sys.exit(main())

soleprint/atlas2/docgen/book/__init__.py

@@ -0,0 +1,315 @@
+"""
+A **book** — one docgen operation, measured at both ends.
+
+    larder ──► step ──► step ──► step ──► book
+      what        each one usable          what
+      came in     by itself                came out
+
+The first step says what went into the operation. The last step says what came
+out, and renders it for the web. Everything between is an ordinary artifact that
+stands on its own — `ir.json` is still an `ir.json`, and `make ir` still works
+without knowing this file exists.
+
+## Why both ends, and why they are steps rather than a wrapper
+
+Because the two numbers are only worth having **together**. "102 nodes" is not a
+fact about anything; "49 files in, 102 nodes out, nothing lost" is. A book that
+read 45 of 47 files and drew a clean diagram is lying by omission, and before
+this there was no place for the other 2 to be mentioned.
+
+They are steps, not a wrapper, because a wrapper is something you can forget to
+apply. A step is in the sequence, and the sequence is the thing the notebook
+emits — so the measure is in the document whether or not anyone remembered.
+
+## The two things it is not
+
+**Not a gate.** Running one step alone is still a book, just a short one. An
+operation that cannot measure something says what it could not measure and
+carries on. Gating would break the property that makes the intermediate
+artifacts useful, and that property is the whole reason the sequence is worth
+having.
+
+**Not a new pipeline.** Everything here composes functions that already existed
+— `ops.overview`, `emitters.dot`, `emitters.site`, `notebook.spec`. The spine
+adds a ledger and two measurements. If it ever starts doing the work itself,
+something has gone wrong.
+
+## The web end is deliberately the loose one
+
+It is last, so nothing depends on it, so it can be replaced wholesale without
+touching a single thing upstream. That is what lets it rule what gets generated
+without being a stable contract: the book measure is the promise, and the page
+that displays it is free to change drastically and often.
+"""
+
+import json
+from dataclasses import dataclass, field
+from pathlib import Path
+
+from .larder import Larder
+
+# The two spine steps. Named here because the notebook, the site and the checks
+# all have to agree on what they are called, and three string literals in three
+# files is how they stop agreeing.
+FIRST = "larder"
+LAST = "book"
+
+
+@dataclass
+class Step:
+    """One thing the operation did, and what it left behind."""
+
+    id: str
+    label: str
+    artifact: str | None = None      # relative to the book directory
+    bytes: int = 0
+    note: str = ""
+    skipped: str = ""                # why, when it did not run
+
+    def to_dict(self) -> dict:
+        out = {"id": self.id, "label": self.label}
+        if self.artifact:
+            out["artifact"] = self.artifact
+            out["bytes"] = self.bytes
+        if self.note:
+            out["note"] = self.note
+        if self.skipped:
+            out["skipped"] = self.skipped
+        return out
+
+
+# How a larder's unit shows up on the output side. Per-kind because the relation
+# genuinely differs, and because getting it wrong produces a check that passes
+# for the wrong reason — which is what the first version of this did.
+#
+#   relation   "exact"     one unit in, one node out. Fewer means input was dropped.
+#              "collapse"  many units in, fewer nodes out, by design.
+#   noun       what to call the output-side thing, in the claim
+#   represents whether a unit that FAILED still appears as a node. Where it does,
+#              that is checkable and is the whole-input form of the rule that an
+#              unresolved name becomes an `external` node rather than vanishing.
+RELATION = {
+    "python": ("exact", "module", True),
+    "code": ("exact", "module", True),
+    "db": ("exact", "table", False),
+    "openapi": ("exact", "endpoint", False),
+    # A HAR entry with no URL has nothing to represent, so there is no node to
+    # look for. Its absence is the correct outcome and is recorded in `failed`.
+    "usage": ("collapse", "call", False),
+}
+
+
+def _unit_counts(ir: dict, larder: Larder) -> tuple[int, int]:
+    """(nodes from units that were read, nodes from units that failed).
+
+    Split because a failed file still gets a module node — carrying
+    `attrs.error`, so the gap is visible in the graph rather than only in a log.
+    Counting them together made "2 files read produced 4 modules" pass a check
+    that was supposed to prove nothing had been dropped.
+    """
+    nodes = ir.get("nodes") or []
+
+    if larder.kind in ("python", "code"):
+        mods = [n for n in nodes
+                if n["kind"] == "module" and (n.get("attrs") or {}).get("file")]
+        broken = {n["attrs"]["file"] for n in mods if (n.get("attrs") or {}).get("error")}
+        whole = {n["attrs"]["file"] for n in mods} - broken
+        return len(whole), len(broken)
+
+    if larder.kind == "db":
+        return sum(1 for n in nodes if n["kind"] == "table"), 0
+
+    if larder.kind == "openapi":
+        return len({
+            (n.get("attrs") or {}).get("path") for n in nodes
+            if n["kind"] == "endpoint" and (n.get("attrs") or {}).get("path")
+        }), 0
+
+    if larder.kind == "usage":
+        return sum(1 for n in nodes if n["kind"] in ("endpoint", "operation")), 0
+
+    return 0, 0
+
+
+class Book:
+    """A named operation, its ledger, and the two measures that bracket it."""
+
+    def __init__(self, slug: str, larder: Larder, out):
+        self.slug = slug
+        self.larder = larder
+        self.out = Path(out)
+        self.steps: list[Step] = []
+        self.ir: dict | None = None
+        self.notebook: str | None = None
+
+        # The first step, recorded before any work happens. Doing it here rather
+        # than at the end is the difference between a measure and a summary: it
+        # says what the operation *set out* to read, so a crash halfway leaves a
+        # book that still says what went in.
+        self.steps.append(Step(
+            id=FIRST,
+            label="what came in",
+            note=larder.line(),
+        ))
+
+    # -- the ledger -------------------------------------------------------
+
+    def step(self, id: str, label: str, path=None, note: str = "",
+             skipped: str = "") -> Step:
+        """Record a step. `path` is written already; this measures it."""
+        artifact, size = None, 0
+        if path is not None:
+            path = Path(path)
+            if path.exists():
+                artifact = str(path.relative_to(self.out)) if self.out in path.parents \
+                    or path.parent == self.out else str(path)
+                size = path.stat().st_size if path.is_file() else _tree_bytes(path)
+        s = Step(id=id, label=label, artifact=artifact, bytes=size,
+                 note=note, skipped=skipped)
+        self.steps.append(s)
+        return s
+
+    # -- the last step ----------------------------------------------------
+
+    def measure(self) -> dict:
+        """What came out. Counted off the final IR and the ledger."""
+        ir = self.ir or {"nodes": [], "edges": []}
+        by_kind: dict[str, int] = {}
+        for n in ir.get("nodes") or []:
+            by_kind[n["kind"]] = by_kind.get(n["kind"], 0) + 1
+        edge_kinds: dict[str, int] = {}
+        for e in ir.get("edges") or []:
+            edge_kinds[e["kind"]] = edge_kinds.get(e["kind"], 0) + 1
+
+        artifacts = [s for s in self.steps if s.artifact]
+        return {
+            "nodes": sum(by_kind.values()),
+            "by_kind": {k: by_kind[k] for k in sorted(by_kind)},
+            "edges": sum(edge_kinds.values()),
+            "edges_by_kind": {k: edge_kinds[k] for k in sorted(edge_kinds)},
+            "external": by_kind.get("external", 0),
+            "steps": len(self.steps),
+            "artifacts": [{"path": s.artifact, "bytes": s.bytes} for s in artifacts],
+            "bytes": sum(s.bytes for s in artifacts),
+        }
+
+    def compare(self) -> list[dict]:
+        """Reconcile the two ends. This is what having both is *for*.
+
+        Returns observations, each with a verdict, rather than raising: the book
+        is already built by the time anyone can compare, and refusing to write it
+        would destroy the evidence. `book/checks.py` turns these into pass/fail
+        at the book test level, and the CLI exits 1 when one of them is not ok.
+        """
+        out = []
+        ir = self.ir or {}
+        produced, represented = _unit_counts(ir, self.larder)
+        relation, noun, represents = RELATION.get(
+            self.larder.kind, ("collapse", "node", False))
+        read, unit = self.larder.read, self.larder.unit
+        n_failed = len(self.larder.failed)
+
+        if relation == "exact":
+            ok = produced >= read
+            out.append({
+                "id": "units-accounted-for",
+                "ok": ok,
+                "claim": f"{read} {unit}(s) read produced {produced} {noun}(s)",
+                "why": (
+                    "one unit in, one node out" if ok else
+                    f"{read - produced} {unit}(s) were read and produced nothing — "
+                    "the input was dropped between the extractor and the document, "
+                    "which is the failure this measure exists to catch"
+                ),
+            })
+        else:
+            out.append({
+                "id": "collapse-is-intended",
+                "ok": produced >= 1 or read == 0,
+                "claim": f"{read} {unit}(s) collapsed to {produced} {noun}(s)",
+                "why": "many recorded requests describe few endpoints — that is the point",
+            })
+
+        if n_failed:
+            out.append({
+                "id": "failures-surfaced",
+                "ok": True,
+                "claim": f"{n_failed} {unit}(s) could not be read",
+                "why": "named in the larder measure and on the landing page, not only in a log",
+            })
+
+        # The whole-input form of "an unresolved name becomes an `external` node".
+        # A file that failed to parse must still be in the graph, or the diagram
+        # shows a tree that is smaller than the tree on disk and says nothing.
+        if represents and n_failed:
+            out.append({
+                "id": "failures-still-in-the-graph",
+                "ok": represented >= n_failed,
+                "claim": f"{n_failed} unreadable {unit}(s) appear as {represented} "
+                         f"marked {noun}(s)",
+                "why": (
+                    "a gap that is drawn can be seen" if represented >= n_failed else
+                    f"{n_failed - represented} unreadable {unit}(s) are missing from the "
+                    "graph entirely — the picture is smaller than the source and does "
+                    "not say so"
+                ),
+            })
+        return out
+
+    # -- writing ----------------------------------------------------------
+
+    def close(self, *, site=None) -> Step:
+        """Append the last step. Call once, after the web output is written.
+
+        Must be called BEFORE the notebook is built, because the notebook quotes
+        the book measure and the measure is not complete until this step exists.
+
+        The site is this step's *artifact*. The notebook is not: it is the whole
+        sequence's rendering rather than an item in it, and it is set on
+        `self.notebook` afterwards, by name only — a notebook cannot report its
+        own byte count without changing it.
+        """
+        m = self.measure()
+        summary = " · ".join(f"{v} {k}" for k, v in
+                             sorted(m["by_kind"].items(), key=lambda kv: -kv[1]))
+        artifact, size = None, 0
+        if site is not None:
+            site = Path(site)
+            if site.exists():
+                artifact, size = str(site.relative_to(self.out)), _tree_bytes(site)
+        self.steps.append(Step(
+            id=LAST,
+            label="what came out",
+            artifact=artifact,
+            bytes=size,
+            note=f"{summary} · {m['edges']} edges" if summary else "nothing was produced",
+        ))
+        return self.steps[-1]
+
+    def to_dict(self) -> dict:
+        """The book, as the ledger someone else can read.
+
+        `generated_at` is absent for the same reason the IR's is: an unchanged
+        larder must serialise to the same bytes, or nothing downstream can tell
+        a real change from a rebuild.
+        """
+        out = {
+            "slug": self.slug,
+            "larder": self.larder.to_dict(),
+            "steps": [s.to_dict() for s in self.steps],
+            "book": self.measure(),
+            "reconciled": self.compare(),
+        }
+        if self.notebook:
+            out["notebook"] = self.notebook
+        return out
+
+    def write(self) -> Path:
+        self.out.mkdir(parents=True, exist_ok=True)
+        path = self.out / "book.json"
+        path.write_text(json.dumps(self.to_dict(), indent=2, sort_keys=False) + "\n")
+        return path
+
+
+def _tree_bytes(root: Path) -> int:
+    return sum(p.stat().st_size for p in root.rglob("*") if p.is_file())

soleprint/atlas2/docgen/book/build.py

@@ -0,0 +1,380 @@
+"""
+Run a book: larder, the steps, the web output, the book measure.
+
+    python3 -m docgen book --root ../station -o out/book/station
+
+Composes functions that already exist. Nothing here parses, lays out or styles
+anything — `ops.overview`, `emitters.auto.draw`, `emitters.site` and `notebook.spec`
+do all of it, and this decides the order and writes the ledger. If this file
+ever starts doing the work, the seam has moved to the wrong place.
+
+## The notebook is the sequence, not an item in it
+
+So it is not a step. Its **first cell is the larder measure and its last cell is
+the book measure**, which is what makes the two ends part of the document rather
+than part of the tooling. Between them, one pair of cells per step: what the
+step did, and a code cell that loads that step's artifact on its own.
+
+That last part is the "usable by themselves" property made executable. Each cell
+reads one artifact and prints one fact, so a reader can start anywhere in the
+sequence, and `selftest.py` runs them.
+"""
+
+import json
+from pathlib import Path
+
+from . import Book
+from .larder import Larder
+
+# Loader for the notebook's step cells. Guarded on purpose: a notebook is
+# opened from wherever somebody happens to open it, and a traceback on cell 2
+# is a worse answer than a sentence saying which directory to run it from.
+PRELUDE = '''from pathlib import Path
+import json
+
+BOOK = Path.cwd()          # the book directory — change if you opened this elsewhere
+
+
+def load(rel):
+    """One step's artifact, on its own. Returns None when it is not here.
+
+    A step's artifact can be a directory — the explorer is one — so this returns
+    a file list for those rather than trying to read a directory as text.
+    """
+    p = BOOK / rel
+    if not p.exists():
+        print(f"{rel} is not here — run: make book OUT={BOOK}")
+        return None
+    if p.is_dir():
+        return sorted(f.relative_to(p).as_posix() for f in p.rglob("*") if f.is_file())
+    return json.loads(p.read_text()) if p.suffix == ".json" else p.read_text()
+'''
+
+
+# Every extractor a book can run, as data rather than as an if/elif chain.
+#
+# This is the shape that gives the extension contract teeth. `selftest.py` loops
+# over this dict and asserts each entry reports a larder measure — so adding an
+# extractor makes the test start asking about it without anyone remembering to
+# go and add a case. rig's `CONFIG_OVERRIDABLE` loop is the precedent, and it
+# found two real bugs that way.
+#
+#   kind -> (module, function, what the source is)
+EXTRACTORS = {
+    "python": (".extractors.python", "extract", "tree"),
+    "code": (".extractors.code", "extract", "tree"),
+    "db": (".extractors.db", "extract", "file"),
+    "openapi": (".extractors.openapi", "extract", "file"),
+    "usage": (".extractors.usage", "extract", "file"),
+}
+
+# Extractors that walk a directory take an exclude list; the ones that read a
+# single document have nothing to exclude.
+TAKES_EXCLUDE = {"python", "code"}
+
+
+def extract(kind: str, source, *, exclude=(), identity=None):
+    """(IR dict, Larder) for one source type, dispatched through EXTRACTORS."""
+    from importlib import import_module
+
+    if kind not in EXTRACTORS:
+        raise ValueError(
+            f"no extractor named {kind!r} — have {', '.join(sorted(EXTRACTORS))}"
+        )
+    module_name, fn_name, _ = EXTRACTORS[kind]
+    # The package name is derived, not written: the Makefile takes PKG from the
+    # directory name so the folder can be copied anywhere and renamed, and a
+    # literal "docgen" here would quietly undo that.
+    root_pkg = __package__.rsplit(".", 1)[0]
+    fn = getattr(import_module(module_name, package=root_pkg), fn_name)
+
+    kwargs = {"identity": identity or str(source)}
+    if kind in TAKES_EXCLUDE:
+        kwargs["exclude"] = exclude
+    g = fn(source, **kwargs)
+
+    ir = g.to_dict()
+    measured = ir["meta"].get("larder")
+    if measured is None:
+        # Not fatal. An extractor that cannot count its input still produces a
+        # book; what it does not get to do is pretend it measured one.
+        return ir, None
+    return ir, Larder.from_dict(measured)
+
+
+def spec_from(book: Book, ir: dict) -> dict:
+    """The book's ledger -> a notebook spec, with the two measures at the ends.
+
+    Reuses `notebook.spec`'s step vocabulary rather than inventing one, so
+    `merge()` keeps working and a hand-written overlay can annotate a spine step
+    the same way it annotates any other.
+    """
+    from ..notebook import spec as spec_mod
+
+    steps = [
+        spec_mod._step(
+            "larder", "md", title=f"{book.slug} — what came in",
+            text=(
+                f"`{book.larder.identity}`\n\n**{book.larder.line()}**\n\n"
+                "This is the first step of the book and the only measure of the "
+                "input. Everything below is derived from it, so a number here "
+                "that looks wrong makes everything below it suspect."
+                + (
+                    "\n\nCould not be read:\n\n"
+                    + "\n".join(f"- `{f['name']}` — {f['error']}"
+                                for f in book.larder.failed)
+                    if book.larder.failed else ""
+                )
+            ),
+        ),
+        spec_mod._step("prelude", "code", title="Reading a step on its own",
+                       code=PRELUDE),
+    ]
+
+    for s in book.steps:
+        if s.id in ("larder", "book") or not s.artifact:
+            continue
+        steps.append(spec_mod._step(
+            f"step-{s.id}", "md", title=s.label,
+            text=f"`{s.artifact}` — {s.bytes:,} bytes" + (f"\n\n{s.note}" if s.note else ""),
+        ))
+        steps.append(spec_mod._step(
+            f"load-{s.id}", "code", code=_load_cell(s.id, s.artifact),
+        ))
+
+    # Where the larder is API-shaped, the generated endpoint walkthrough slots in
+    # as further steps. For a tree of source there are no endpoints and this adds
+    # nothing, which is the correct amount for it to add.
+    if any(n["kind"] in ("endpoint", "operation") for n in ir.get("nodes") or []):
+        generated = spec_mod.from_ir(ir)
+        steps.extend(s for s in generated["steps"] if s["id"] not in ("intro",))
+
+    m = book.measure()
+    # `external` is reported on its own below, so it is dropped here rather than
+    # appearing twice in one line — which is how it read before.
+    summary = " · ".join(f"{v} {k}" for k, v in
+                         sorted(m["by_kind"].items(), key=lambda kv: -kv[1])
+                         if k != "external")
+    reconciled = "\n".join(
+        f"- {'✓' if r['ok'] else '✗'} {r['claim']} — {r['why']}"
+        for r in book.compare()
+    )
+    steps.append(spec_mod._step(
+        "book", "md", title=f"{book.slug} — what came out",
+        text=(
+            f"**{summary} · {m['edges']} edges · {m['external']} external**\n\n"
+            f"{len(m['artifacts'])} artifact(s), {m['bytes']:,} bytes.\n\n"
+            f"Reconciled against what came in:\n\n{reconciled}\n\n"
+            "The web output is `site/index.html`."
+        ),
+    ))
+
+    return {"version": spec_mod.SPEC_VERSION, "steps": steps}
+
+
+def _load_cell(step_id: str, artifact: str) -> str:
+    """A cell that reads one artifact and prints one fact about it.
+
+    The fact has to suit the artifact. An earlier version printed "nodes, edges"
+    for every JSON file, so `sidebar.json` — which has neither — reported
+    "0 nodes, 0 edges", which is a true sentence about the wrong thing and worse
+    than saying nothing.
+    """
+    if not artifact.endswith((".json", ".svg", ".md")):
+        # A directory, e.g. the explorer.
+        return (
+            f'files = load("{artifact}")\n'
+            'if files is not None:\n'
+            '    print(f"{len(files)} file(s)")\n'
+            '    print("\\n".join(files[:5]))'
+        )
+    if artifact.endswith(".json"):
+        return (
+            f'data = load("{artifact}")\n'
+            'if data:\n'
+            '    if "nodes" in data:\n'
+            '        print(f\'{len(data["nodes"])} nodes, {len(data.get("edges", []))} edges\')\n'
+            '    else:\n'
+            '        print(", ".join(f"{k}: {len(v) if isinstance(v, (list, dict)) else v}"\n'
+            '                        for k, v in data.items()))'
+        )
+    if artifact.endswith(".svg"):
+        # No IPython import, guarded or otherwise. A generated notebook is a
+        # build artifact and has to run wherever it is opened; requiring a
+        # kernel package in order to *load a file* would make the cell fail on
+        # the machine that produced it, which is where this was found.
+        return (
+            f'svg = load("{artifact}")\n'
+            'if svg:\n'
+            '    print(f"{len(svg):,} bytes of SVG")\n'
+            '    # In Jupyter: from IPython.display import SVG; SVG(svg)'
+        )
+    if artifact.endswith(".md"):
+        return (
+            f'text = load("{artifact}")\n'
+            'if text:\n'
+            '    print(text[:400])'
+        )
+    return f'print(load("{artifact}") is not None)'
+
+
+# What a build writes into a book directory, and therefore what a rebuild may
+# remove first. Named, never globbed: a book directory also holds the hand-written
+# `checks.py` and `overlay.json`, and a misconfigured output path could point at
+# somebody's source tree — so nothing outside this list is ever deleted.
+GENERATED = ("book.json", "notebook.ipynb", "steps", "site", "explore")
+
+
+def clear(out) -> list[str]:
+    """Remove a previous build's outputs from `out`. Returns what was removed.
+
+    Without this, re-running a book leaves the last run's artifacts beside the new
+    ledger: a graph that is now `graph.md` keeps its old `graph.svg`, and the site
+    goes on showing it. The ledger would not list the stale file, so nothing would
+    say it was stale — which is the failure the book measure exists to prevent.
+
+    Only acts on a directory that already holds a `book.json`, i.e. one this code
+    wrote. Anything else is left exactly as found.
+    """
+    import shutil
+
+    out = Path(out)
+    if not (out / "book.json").is_file():
+        return []
+    removed = []
+    for name in GENERATED:
+        target = out / name
+        if target.is_dir():
+            shutil.rmtree(target)
+            removed.append(name)
+        elif target.is_file():
+            target.unlink()
+            removed.append(name)
+    return removed
+
+
+def run(kind: str, source, out, *, slug: str | None = None, style: str = "lucid",
+        theme: str | None = None, exclude=(), overlay=None, quiet: bool = False) -> Book:
+    """The whole book, in one process. Returns it; `book.json` is written."""
+    from ..ir import check
+    from ..ops import classify, overview
+    from ..style import Style
+
+    out = Path(out)
+    slug = slug or Path(str(source)).name or "book"
+    clear(out)
+    steps_dir = out / "steps"
+    steps_dir.mkdir(parents=True, exist_ok=True)
+
+    def say(text):
+        if not quiet:
+            print(text)
+
+    ir, larder = extract(kind, source, exclude=exclude)
+    if larder is None:
+        larder = Larder(kind=kind, identity=str(source), unit="document", seen=0)
+    book = Book(slug=slug, larder=larder, out=out)
+    book.ir = ir
+    say(f"  larder   {larder.line()}")
+
+    problems = check(ir)
+    if problems:
+        # Recorded as a step rather than raised: a book that cannot be trusted
+        # should exist and say so, because the alternative is that nobody can
+        # see what went wrong.
+        book.step("validate", "the IR did not validate", note="; ".join(problems[:3]))
+        say(f"  WARNING  IR has {len(problems)} problem(s)")
+
+    p = steps_dir / "ir.json"
+    p.write_text(json.dumps(ir, indent=2) + "\n")
+    book.step("ir", "the graph, extracted", p, note=f"{len(ir['nodes'])} nodes")
+
+    view = overview(ir)
+    p = steps_dir / "view.json"
+    p.write_text(json.dumps(view, indent=2) + "\n")
+    verdict = classify(view)
+    book.step("view", "the default view for this source", p,
+              note=f"{verdict['kind']} — {verdict['why']}")
+
+    style_obj = Style.load(style, theme=theme)
+
+    # The drawing, chosen by structure rather than by the caller — the same
+    # `draw()` the `emit auto` and `emit site` commands use, so the three can
+    # never disagree about what a graph wants.
+    from ..emitters.auto import draw
+
+    drawn = draw(view, style_obj)
+    graph_rel = None
+    labels = {"erd": "drawn as an ERD", "dot": f"drawn as a {verdict['kind']}",
+              "index": "not a diagram — written as a list"}
+    if drawn["content"] is None:
+        book.step("graph", "not drawn", skipped=drawn["why"])
+    else:
+        p = steps_dir / f"graph{drawn['suffix']}"
+        if isinstance(drawn["content"], bytes):
+            p.write_bytes(drawn["content"])
+        else:
+            p.write_text(drawn["content"])
+        if drawn["suffix"] == ".svg":
+            graph_rel = f"steps/{p.name}"
+        book.step("graph", labels[drawn["emitter"]], p, note=drawn["why"])
+
+    from ..emitters.index import to_markdown, to_sidebar
+    p = steps_dir / "index.md"
+    p.write_text(to_markdown(ir))
+    book.step("index", "readable without a diagram", p)
+    p = steps_dir / "sidebar.json"
+    p.write_text(json.dumps(to_sidebar(ir), indent=2) + "\n")
+    book.step("sidebar", "navigation, for whatever renders it", p)
+
+    try:
+        from ..emitters.minimap import emit as mm_emit
+        p = steps_dir / "minimap.svg"
+        p.write_text(mm_emit(ir, style_obj))
+        book.step("minimap", "what is where, read from the colours", p)
+    except Exception as e:  # noqa: BLE001
+        book.step("minimap", "minimap not drawn", skipped=f"{type(e).__name__}: {e}")
+
+    try:
+        from ..emitters.explore import write as exp_write
+        exp_write(ir, style_obj, out / "explore")
+        book.step("explore", "navigate on one side, explore on the other",
+                  out / "explore")
+    except Exception as e:  # noqa: BLE001
+        book.step("explore", "explorer not built", skipped=f"{type(e).__name__}: {e}")
+
+    # -- the last step ----------------------------------------------------
+    from ..emitters.site import write as site_write
+    site_dir = out / "site"
+    site_write(view, style_obj, site_dir,
+               graph=Path(graph_rel).name if graph_rel else None,
+               title=slug, book=book.to_dict())
+    if graph_rel and (out / graph_rel).exists():
+        (site_dir / Path(graph_rel).name).write_bytes((out / graph_rel).read_bytes())
+
+    # close() BEFORE the notebook spec is built, not after. The spec quotes the
+    # book measure, and the book measure only includes the site once the last
+    # step exists — build it the other way round and the notebook says 7
+    # artifacts while book.json says 8, which is exactly what it did.
+    book.close(site=site_dir)
+
+    from ..emitters.notebook import write as nb_write
+    from ..notebook import spec as spec_mod
+    spec, spec_problems = spec_mod.merge(spec_from(book, ir), overlay)
+    nb_write(spec, out / "notebook.ipynb")
+    # Recorded by name and not by size: a notebook cannot report its own byte
+    # count without changing it.
+    book.notebook = "notebook.ipynb"
+
+    for pr in spec_problems:
+        say(f"  overlay  {pr}")
+    path = book.write()
+
+    m = book.measure()
+    say(f"  book     {m['nodes']} nodes · {m['edges']} edges · "
+        f"{m['external']} external · {len(m['artifacts'])} artifacts")
+    for r in book.compare():
+        say(f"  {'ok  ' if r['ok'] else 'LOST'}     {r['claim']}")
+    say(f"  open     {site_dir / 'index.html'}")
+    say(f"           {path}")
+    return book

soleprint/atlas2/docgen/book/checks.py

@@ -0,0 +1,274 @@
+"""
+Does *this* book still hold — the third test level.
+
+    python3 -m docgen check out/book/station
+
+docgen has three levels of test, and they differ by what they assert *about*.
+The distinction is rig's, from `rig/ctrl/selftest.sh`, and it is worth keeping
+because it decides what a failure means:
+
+    make doctor    the MACHINE.   Never fails; it reports.
+    make check     DOCGEN.        Exits 1. "docgen no longer does what it says."
+    make check BOOK=<dir>
+                   THIS BOOK.     Exits 1. "this book no longer holds."
+
+The third is the one that reaches a project docgen has never seen. It is also
+where **framework and custom checks live together**, at different levels:
+
+    generated   the spine's own assertions, identical for every book. Both
+                measures present, the reconciliation holding, every artifact
+                where the ledger says it is, the notebook still executing.
+    custom      hand-written, in the book's own `checks.py`, using the same
+                check/note/skip helpers — so a project's line and a framework
+                line read identically and fail identically.
+
+Same split as the notebook's base and overlay, for the same reason: generation
+alone cannot know what *this* project cares about, and hand-authoring alone rots.
+
+## The discipline these are written in
+
+Carried from rig verbatim, because it is the point of the whole level:
+
+> Each check is ONE decision that has already been made, with the reason above
+> it — not coverage, and deliberately not an exhaustive sweep of use cases. A
+> rule without its reason gets overridden the first time it is inconvenient.
+> Failing one should read as **"you are about to undo this"** rather than
+> "something broke".
+
+## Writing a book's own checks
+
+Put a `checks.py` next to `book.json`:
+
+    def checks(book, check, note, skip):
+        note("what this project will not give up")
+
+        # Payments moved once already and the move broke three dashboards.
+        # If it is not here, something renamed it again.
+        check("payments is still a module", True,
+              any(n["id"] == "app.payments" for n in book.ir["nodes"]))
+
+`book` carries `.data` (the parsed `book.json`), `.ir` (the extracted graph),
+and `.dir`. `check` takes (name, expected, actual) — expected first, because a
+failure report is only useful if it says what was wanted.
+"""
+
+import json
+import sys
+from pathlib import Path
+
+from . import FIRST, LAST
+
+
+class Loaded:
+    """A book read back off disk, for checking rather than building."""
+
+    def __init__(self, directory):
+        self.dir = Path(directory)
+        path = self.dir / "book.json"
+        if not path.exists():
+            raise FileNotFoundError(
+                f"{path} does not exist — is {self.dir} a book directory?\n"
+                "Build one with: python3 -m docgen book --root <src> -o <dir>"
+            )
+        self.data = json.loads(path.read_text())
+
+        ir_path = self.dir / "steps" / "ir.json"
+        self.ir = json.loads(ir_path.read_text()) if ir_path.exists() else None
+
+        nb_path = self.dir / (self.data.get("notebook") or "notebook.ipynb")
+        self.notebook = json.loads(nb_path.read_text()) if nb_path.exists() else None
+
+    @property
+    def larder(self) -> dict:
+        return self.data.get("larder") or {}
+
+    @property
+    def measure(self) -> dict:
+        return self.data.get("book") or {}
+
+
+class Report:
+    """rig's reporting shape: sections, one line per decision, nothing aborts."""
+
+    def __init__(self):
+        self.passed, self.failed, self.skipped = 0, [], 0
+
+    def note(self, text: str) -> None:
+        print(f"\n{text}")
+
+    def check(self, name: str, expected, actual) -> bool:
+        if expected == actual:
+            print(f"  ok    {name}")
+            self.passed += 1
+            return True
+        print(f"  FAIL  {name}\n        expected: {expected!r}\n        got:      {actual!r}")
+        self.failed.append(name)
+        return False
+
+    def skip(self, name: str, why: str) -> None:
+        print(f"  --    {name}  ({why})")
+        self.skipped += 1
+
+    def total(self) -> int:
+        print()
+        # stdout is block-buffered when redirected and stderr is not, so the
+        # failure summary printed below would otherwise arrive BEFORE the checks
+        # it summarises — which is how this was found, piping to `tail`.
+        sys.stdout.flush()
+        if not self.failed:
+            print(f"{self.passed} checks passed — this book still holds"
+                  + (f", {self.skipped} skipped" if self.skipped else ""))
+            return 0
+        print(f"FAILED — {len(self.failed)} of {self.passed + len(self.failed)}: "
+              f"{', '.join(self.failed)}", file=sys.stderr)
+        print("Read the comment next to the check. A failure here means a decision "
+              "has drifted, not that a tool is broken.", file=sys.stderr)
+        return 1
+
+
+def generated(book: Loaded, r: Report) -> None:
+    """The spine's own assertions. Identical for every book, by design."""
+
+    r.note("the book is bracketed — both ends present")
+    # The one structural promise the whole design makes: an operation is
+    # measured at both ends. A book missing either end is not a short book, it
+    # is a book whose numbers cannot be reconciled against anything.
+    ids = [s["id"] for s in book.data.get("steps") or []]
+    r.check("the first step is the larder measure", FIRST, ids[0] if ids else None)
+    r.check("the last step is the book measure", LAST, ids[-1] if ids else None)
+    r.check("nothing is bracketed twice", [1, 1],
+            [ids.count(FIRST), ids.count(LAST)])
+
+    r.note("what came in was measured, not guessed")
+    larder = book.larder
+    for key in ("kind", "identity", "unit", "seen", "read", "failed"):
+        r.check(f"larder records {key}", True, key in larder)
+    # read is derived, so a stored value that disagrees means somebody wrote it
+    # by hand. This is the same arithmetic ir/validate.py enforces; asserted
+    # again here because a book can be assembled without going through the IR.
+    r.check("read == seen - failed", larder.get("read"),
+            max(0, larder.get("seen", 0) - len(larder.get("failed") or [])))
+    # The one field that could carry a secret out of a database URL.
+    r.check("the identity carries no password", True,
+            ":***@" in larder.get("identity", "") or "@" not in larder.get("identity", ""))
+
+    r.note("the two ends reconcile")
+    reconciled = book.data.get("reconciled") or []
+    r.check("something was reconciled", True, len(reconciled) > 0)
+    for item in reconciled:
+        # Each of these is a claim the book makes about itself. A false one
+        # means the document below is smaller than the source and does not say
+        # so, which is the single failure this level exists to catch.
+        r.check(item["claim"], True, item["ok"])
+
+    r.note("the ledger describes files that exist")
+    for step in book.data.get("steps") or []:
+        artifact = step.get("artifact")
+        if not artifact:
+            continue
+        path = book.dir / artifact
+        # A ledger naming a file that is not there is worse than no ledger: it
+        # is a manifest somebody will build tooling against.
+        r.check(f"{artifact} is where the ledger says", True, path.exists())
+        if path.is_file():
+            r.check(f"{artifact} is {step['bytes']} bytes", step["bytes"],
+                    path.stat().st_size)
+
+    r.note("the notebook carries the sequence, with the measures at its ends")
+    if book.notebook is None:
+        r.skip("notebook", "no notebook.ipynb in this book")
+    else:
+        cells = book.notebook.get("cells") or []
+        first = "".join(cells[0]["source"]) if cells else ""
+        last = "".join(cells[-1]["source"]) if cells else ""
+        r.check("its first cell is what came in", True, "what came in" in first)
+        r.check("its last cell is what came out", True, "what came out" in last)
+        # Compiling is not enough — see selftest.py. `json.dumps` writes
+        # `false`/`true`/`null`, which are valid Python *identifiers*, so a
+        # generated body full of them compiles and then raises NameError.
+        ran, failure = _run_cells(cells, book.dir)
+        r.check(f"its {ran} code cells run, not merely compile", None, failure)
+
+    r.note("the web output exists and shows both ends")
+    index = book.dir / "site" / "index.html"
+    if not index.exists():
+        r.skip("site", "no site/index.html in this book")
+    else:
+        html = index.read_text()
+        # The site is the last step precisely because it is the artifact somebody
+        # definitely opens. A page that shows the result without showing what
+        # went in is the thing this whole change corrects.
+        r.check("the page says what came in", True, "what came in" in html)
+        r.check("the page says what came out", True, "what came out" in html)
+        if larder.get("failed"):
+            r.check("the page admits the book is incomplete", True,
+                    "This book is incomplete" in html)
+
+    r.note("the book is reproducible")
+    # A timestamp would make two builds of an unchanged larder differ, which
+    # destroys the only useful property a ledger has: that a diff means a real
+    # change. Same rule as the IR's `generated_at`.
+    r.check("no timestamp in the ledger", True, "generated_at" not in json.dumps(book.data))
+
+
+def custom(book: Loaded, r: Report) -> None:
+    """This book's own assertions, if it has any."""
+    path = book.dir / "checks.py"
+    if not path.exists():
+        r.note("this book's own checks")
+        r.skip("custom checks", f"no {path.name} — write one to assert what this project cares about")
+        return
+
+    namespace: dict = {"__file__": str(path), "__name__": "book_checks"}
+    try:
+        exec(compile(path.read_text(), str(path), "exec"), namespace)
+    except Exception as e:  # noqa: BLE001 - report, do not traceback
+        r.note("this book's own checks")
+        r.check(f"{path.name} loads", None, f"{type(e).__name__}: {e}")
+        return
+
+    fn = namespace.get("checks")
+    if not callable(fn):
+        r.note("this book's own checks")
+        r.check(f"{path.name} defines checks(book, check, note, skip)", True, False)
+        return
+
+    try:
+        fn(book, r.check, r.note, r.skip)
+    except Exception as e:  # noqa: BLE001
+        r.check(f"{path.name} ran to completion", None, f"{type(e).__name__}: {e}")
+
+
+def _run_cells(cells: list, cwd: Path) -> tuple[int, str | None]:
+    """Execute the notebook's code cells from the book directory."""
+    import contextlib
+    import io
+    import os
+
+    ns, ran, failure = {}, 0, None
+    previous = os.getcwd()
+    try:
+        os.chdir(cwd)
+        with contextlib.redirect_stdout(io.StringIO()):
+            for c in cells:
+                if c.get("cell_type") != "code":
+                    continue
+                src = "".join(c["source"])
+                if "urlopen" in src or ("call(" in src and "def call" not in src):
+                    # Anything that would reach the network is compiled, not run.
+                    try:
+                        compile(src, c["id"], "exec")
+                        ran += 1
+                    except SyntaxError as e:
+                        failure = f"{c['id']}: {e}"
+                        break
+                    continue
+                try:
+                    exec(compile(src, c["id"], "exec"), ns)
+                    ran += 1
+                except Exception as e:  # noqa: BLE001 - any failure is the finding
+                    failure = f"{c['id']}: {type(e).__name__}: {e}"
+                    break
+    finally:
+        os.chdir(previous)
+    return ran, failure

soleprint/atlas2/docgen/book/config.py

@@ -0,0 +1,327 @@
+"""
+A run file: every book a project builds, so a rebuild is one command.
+
+    python3 -m docgen run docgen.toml
+    python3 -m docgen run docgen.toml --only station --only shop
+    python3 -m docgen run docgen.toml --list          # what would run, resolved
+
+Books are rebuilt many times — after a merge, before a release, whenever the
+source moves — and each one is a source, an output directory and a handful of
+options. Typed out every time, those drift: one run excludes `migrations`, the
+next forgets. Written down once, the rebuild is exact.
+
+## The shape
+
+    # docgen.toml
+    reference = "../soleprint"          # optional; $DOCGEN_REFERENCE wins if set
+
+    [defaults]
+    out     = "out/book"                # each book goes to <out>/<name>
+    style   = "lucid"
+    theme   = "dark"
+    exclude = ["migrations", "tests"]
+
+    [[book]]
+    name = "station"
+    root = "../soleprint/station"       # a tree; reader = "python" by default
+
+    [[book]]
+    name    = "orders-api"
+    openapi = "specs/orders.yaml"
+    overlay = "overlays/orders.json"
+    out     = "/srv/docs/orders"        # overrides <defaults.out>/<name>
+
+Each book names **exactly one** source — `root`, `schema`, `openapi` or `har` —
+the same choice the `book` command makes you take.
+
+## Three rules, each one a thing that went wrong somewhere else
+
+**Paths are relative to the run file, not to wherever you ran the command.** A
+run file is kept beside the project it describes; if its paths meant different
+things from different directories, the same file would build different books.
+
+**A book's value replaces the default, for every key.** Including `exclude`,
+which is the one where merging looks tempting. One rule nobody has to remember
+beats a clever one somebody has to look up.
+
+**Unknown keys are refused, not ignored.** `exlude = [...]` silently doing
+nothing is how a rebuild quietly starts reading `node_modules`. The style loader
+takes the same line (requirements R30).
+
+## Why TOML
+
+It is read by the stdlib (`tomllib`), it allows comments — and a run file is
+hand-written, so the reason a book excludes something belongs next to the
+exclusion — and it is already the format of the `pyproject.toml` beside it. The
+IR and style files stay JSON, because those are data that tools write; this is
+configuration that people write.
+"""
+
+import os
+import re
+from dataclasses import dataclass, field
+from pathlib import Path
+
+try:
+    import tomllib
+except ModuleNotFoundError:  # pragma: no cover - Python < 3.11
+    tomllib = None
+
+SOURCES = {"root": "python", "schema": "db", "openapi": "openapi", "har": "usage"}
+READERS = ("python", "code")
+
+TOP_KEYS = {"reference", "defaults", "book"}
+DEFAULT_KEYS = {"out", "style", "theme", "exclude", "reader"}
+BOOK_KEYS = {"name", "out", "style", "theme", "exclude", "reader", "overlay", *SOURCES}
+
+# A book name becomes a directory name and a slug, so it is held to what is safe
+# as both — no separators, nothing that means something to a shell.
+NAME = re.compile(r"^[A-Za-z0-9][A-Za-z0-9._-]*$")
+
+
+class ConfigError(ValueError):
+    """A run file that cannot be run. Carries every problem, not the first."""
+
+    def __init__(self, path, problems: list[str]):
+        self.path, self.problems = path, problems
+        super().__init__(f"{path}: {len(problems)} problem(s)\n  " + "\n  ".join(problems))
+
+
+@dataclass
+class Entry:
+    """One book, fully resolved — nothing relative, nothing defaulted later."""
+
+    name: str
+    kind: str                         # python | code | db | openapi | usage
+    source: Path
+    out: Path
+    style: str = "lucid"
+    theme: str | None = None
+    overlay: Path | None = None
+    exclude: tuple = ()
+
+    def line(self) -> str:
+        return f"{self.name:<18} {self.kind:<8} {self.source}  ->  {self.out}"
+
+
+@dataclass
+class RunFile:
+    path: Path
+    books: list[Entry] = field(default_factory=list)
+    reference: Path | None = None
+
+    def select(self, names) -> list[Entry]:
+        """The named books, in run-file order. An unknown name is an error."""
+        if not names:
+            return list(self.books)
+        known = {b.name for b in self.books}
+        unknown = [n for n in names if n not in known]
+        if unknown:
+            raise ConfigError(self.path, [
+                f"no book named {n!r} — have: {', '.join(sorted(known))}" for n in unknown
+            ])
+        wanted = set(names)
+        return [b for b in self.books if b.name in wanted]
+
+
+def _path(value, base: Path) -> Path:
+    # Normalised lexically, so `--list` shows `atlas2/out` rather than
+    # `atlas2/docgen/../out`. Not resolved: a symlinked project should be named
+    # the way its owner names it.
+    p = Path(str(value)).expanduser()
+    return Path(os.path.normpath(p if p.is_absolute() else base / p))
+
+
+def _within(inner: Path, outer: Path) -> bool:
+    inner, outer = inner.resolve(), outer.resolve()
+    return inner == outer or outer in inner.parents
+
+
+def load(path) -> RunFile:
+    """Read and resolve a run file. Raises ConfigError listing every problem."""
+    path = Path(path)
+    if tomllib is None:
+        raise ConfigError(path, ["run files need Python 3.11+ (tomllib)"])
+    try:
+        data = tomllib.loads(path.read_text())
+    except OSError as e:
+        raise ConfigError(path, [f"cannot read: {e}"]) from None
+    except tomllib.TOMLDecodeError as e:
+        raise ConfigError(path, [f"not valid TOML: {e}"]) from None
+
+    base = path.resolve().parent
+    problems: list[str] = []
+
+    for key in sorted(set(data) - TOP_KEYS):
+        problems.append(f"unknown top-level key {key!r} — have: {', '.join(sorted(TOP_KEYS))}")
+
+    defaults = data.get("defaults") or {}
+    if not isinstance(defaults, dict):
+        problems.append("[defaults] must be a table")
+        defaults = {}
+    for key in sorted(set(defaults) - DEFAULT_KEYS):
+        problems.append(f"[defaults] has unknown key {key!r} — have: "
+                        f"{', '.join(sorted(DEFAULT_KEYS))}")
+
+    reference = None
+    if "reference" in data:
+        reference = _path(data["reference"], base)
+        if not reference.is_dir():
+            problems.append(f"reference {reference} is not a directory")
+
+    raw_books = data.get("book") or []
+    if not isinstance(raw_books, list) or not raw_books:
+        problems.append("no books — add at least one [[book]] table")
+        raw_books = []
+
+    books: list[Entry] = []
+    for i, raw in enumerate(raw_books):
+        where = f"book[{i}]"
+        if not isinstance(raw, dict):
+            problems.append(f"{where} must be a table")
+            continue
+        name = raw.get("name")
+        if isinstance(name, str):
+            where = f"book {name!r}"
+        if not isinstance(name, str) or not NAME.match(name):
+            problems.append(f"{where} needs a name of letters, digits, '.', '_' or '-'")
+            continue
+
+        for key in sorted(set(raw) - BOOK_KEYS):
+            problems.append(f"{where} has unknown key {key!r} — have: "
+                            f"{', '.join(sorted(BOOK_KEYS))}")
+
+        named = [k for k in SOURCES if k in raw]
+        if len(named) != 1:
+            problems.append(
+                f"{where} must name exactly one source ({', '.join(SOURCES)}); "
+                f"it names {', '.join(named) if named else 'none'}"
+            )
+            continue
+        source_key = named[0]
+        source = _path(raw[source_key], base)
+
+        def pick(key, default=None):
+            # A book's value replaces the default, for every key.
+            return raw[key] if key in raw else defaults.get(key, default)
+
+        reader = pick("reader", "python")
+        if source_key == "root":
+            if reader not in READERS:
+                problems.append(f"{where}: reader must be one of {READERS}, got {reader!r}")
+                continue
+            kind = reader
+            if not source.is_dir():
+                problems.append(f"{where}: root {source} is not a directory")
+        else:
+            if "reader" in raw:
+                problems.append(f"{where}: reader only applies to root, not {source_key}")
+            kind = SOURCES[source_key]
+            if not source.is_file():
+                problems.append(f"{where}: {source_key} {source} is not a file")
+
+        if "out" in raw:
+            out = _path(raw["out"], base)
+        else:
+            out = _path(defaults.get("out", "out"), base) / name
+
+        overlay = _path(raw["overlay"], base) if "overlay" in raw else None
+        if overlay is not None and not overlay.is_file():
+            problems.append(f"{where}: overlay {overlay} is not a file")
+
+        exclude = pick("exclude", [])
+        if not isinstance(exclude, list) or not all(isinstance(x, str) for x in exclude):
+            problems.append(f"{where}: exclude must be a list of directory names")
+            exclude = []
+        elif source_key != "root":
+            # Only an exclude the book sets itself is a mistake. One inherited
+            # from [defaults] is meant for the trees and simply does not apply —
+            # refusing it would make a shared default impossible to write.
+            if "exclude" in raw:
+                problems.append(f"{where}: exclude only applies to a root, not {source_key}")
+            exclude = []
+
+        # Writing a book inside the tree it reads means the next run reads the
+        # last run's output. That is a rebuild that changes on every rebuild.
+        if source_key == "root" and source.is_dir() and _within(out, source):
+            problems.append(f"{where}: out {out} is inside the tree it reads ({source})")
+
+        books.append(Entry(
+            name=name, kind=kind, source=source, out=out,
+            style=pick("style", "lucid"), theme=pick("theme"),
+            overlay=overlay, exclude=tuple(exclude),
+        ))
+
+    seen_names, seen_outs = {}, {}
+    for b in books:
+        if b.name in seen_names:
+            problems.append(f"book {b.name!r} is listed twice")
+        seen_names[b.name] = b
+        key = b.out.resolve()
+        if key in seen_outs:
+            # Two books into one directory: the second clears the first on every
+            # run, and nothing says so.
+            problems.append(f"books {seen_outs[key]!r} and {b.name!r} both write to {b.out}")
+        seen_outs[key] = b.name
+
+    if problems:
+        raise ConfigError(path, problems)
+    return RunFile(path=path, books=books, reference=reference)
+
+
+def apply_reference(runfile: RunFile) -> str | None:
+    """Point the one seam at the run file's reference — unless the caller's
+    environment already does. The caller's env beats the file, which is the
+    precedence rig uses for every setting it has.
+    """
+    from .. import reference as ref
+
+    if runfile.reference is None or os.environ.get(ref.ENV_VAR):
+        return None
+    os.environ[ref.ENV_VAR] = str(runfile.reference)
+    return str(runfile.reference)
+
+
+def run(runfile: RunFile, names=None, *, check: bool = False, quiet: bool = True):
+    """Build every selected book, in order. One failure never stops the rest.
+
+    Returns one result per book:
+        {"name", "out", "ok", "error", "lost": [claims], "checks_failed": [names]}
+    """
+    from . import checks as checks_mod
+    from .build import run as build
+
+    apply_reference(runfile)
+    results = []
+    for entry in runfile.select(names):
+        result = {"name": entry.name, "out": str(entry.out), "ok": False,
+                  "error": None, "lost": [], "checks_failed": []}
+        try:
+            overlay = None
+            if entry.overlay is not None:
+                from ..notebook import spec as spec_mod
+                overlay = spec_mod.load(entry.overlay)
+            book = build(entry.kind, entry.source, entry.out, slug=entry.name,
+                         style=entry.style, theme=entry.theme, exclude=entry.exclude,
+                         overlay=overlay, quiet=quiet)
+            result["lost"] = [r["claim"] for r in book.compare() if not r["ok"]]
+        except Exception as e:  # noqa: BLE001 - one bad book must not cost the run
+            result["error"] = f"{type(e).__name__}: {e}"
+            results.append(result)
+            continue
+
+        if check:
+            import contextlib
+            import io
+
+            report = checks_mod.Report()
+            loaded = checks_mod.Loaded(entry.out)
+            sink = io.StringIO() if quiet else None
+            with contextlib.redirect_stdout(sink) if sink else contextlib.nullcontext():
+                checks_mod.generated(loaded, report)
+                checks_mod.custom(loaded, report)
+            result["checks_failed"] = list(report.failed)
+
+        result["ok"] = not result["lost"] and not result["checks_failed"]
+        results.append(result)
+    return results