add heavy loggin

.gitignore

@@ -17,10 +17,8 @@ env/
 *.pot
 *.pyc
 db.sqlite3
-media/in/*
+media/*
-!media/in/.gitkeep
+!media/.gitkeep
-media/out/*
-!media/out/.gitkeep
 
 # Node
 node_modules/
@@ -39,3 +37,4 @@ Thumbs.db
 
 # Project specific
 def/
+ctrl/k8s/overlays/dev/local-config.yaml

core/api/detect_sources.py

@@ -168,7 +168,7 @@ def run_pipeline(req: RunRequest):
     from detect.state import DetectState
 
     local_path = _resolve_video_path(req.video_path)
-    job_id = str(uuid.uuid4())[:8]
+    job_id = str(uuid.uuid4())
 
     if req.skip_vlm:
         os.environ["SKIP_VLM"] = "1"
@@ -200,8 +200,6 @@ def run_pipeline(req: RunRequest):
         source_asset_id=req.source_asset_id,
     )
 
-    import traceback
-
     from detect.graph import PipelineCancelled, set_cancel_check, clear_cancel_check
 
     set_cancel_check(job_id, lambda: job_id in _cancelled_jobs)
@@ -218,9 +216,17 @@ def run_pipeline(req: RunRequest):
             emit.job_complete(job_id, {"status": "cancelled"})
         except Exception as e:
             logger.exception("Pipeline run %s failed: %s", job_id, e)
-            tb = traceback.format_exc()
+            # Mark the current/last stage as error in the graph
+            from detect.graph import _node_states, NODES
+            if job_id in _node_states:
+                states = _node_states[job_id]
+                for node in reversed(NODES):
+                    if states.get(node) in ("running", "done"):
+                        states[node] = "error"
+                        break
+                nodes = [{"id": n, "status": states[n]} for n in NODES]
+                emit.graph_update(job_id, nodes)
             emit.log(job_id, "Pipeline", "ERROR", str(e))
-            emit.log(job_id, "Pipeline", "DEBUG", tb)
             emit.job_complete(job_id, {"status": "failed", "error": str(e)})
         finally:
             _running_jobs.pop(job_id, None)

core/api/detect_sse.py

@@ -26,16 +26,16 @@ router = APIRouter(prefix="/detect", tags=["detect"])
 
 async def _event_generator(job_id: str) -> AsyncGenerator[str, None]:
     cursor = 0
-    timeout = time.monotonic() + 3600  # 1 hour max (detection jobs are long)
+    timeout = time.monotonic() + 3600  # 1 hour max
 
     while time.monotonic() < timeout:
         events, cursor = poll_events(job_id, cursor, prefix=DETECT_EVENTS_PREFIX)
 
         if not events:
-            yield f"event: waiting\ndata: {json.dumps({'job_id': job_id})}\n\n"
+            await asyncio.sleep(0.2)
-            await asyncio.sleep(0.1)
             continue
 
+        is_terminal = False
         for data in events:
             event_type = data.pop("event", "update")
             payload = {**data, "job_id": job_id}
@@ -43,7 +43,14 @@ async def _event_generator(job_id: str) -> AsyncGenerator[str, None]:
             yield f"event: {event_type}\ndata: {json.dumps(payload)}\n\n"
 
             if event_type in TERMINAL_EVENTS:
+                is_terminal = True
+
+        if is_terminal:
             yield f"event: done\ndata: {json.dumps({'job_id': job_id})}\n\n"
+            # Don't return — keep connection alive so EventSource doesn't reconnect.
+            # Just idle until the client disconnects or timeout.
+            while time.monotonic() < timeout:
+                await asyncio.sleep(5)
             return
 
         await asyncio.sleep(0.05)

ctrl/Tiltfile

@@ -46,6 +46,7 @@ k8s_resource('gateway', resource_deps=['fastapi', 'detection-ui'],
 # Group uncategorized resources (configmaps, namespace) under infra
 k8s_resource(
     objects=['mpr:namespace', 'mpr-config:configmap', 'minio-config:configmap',
-             'postgres-config:configmap', 'envoy-gateway-config:configmap'],
+             'postgres-config:configmap', 'envoy-gateway-config:configmap',
+             'minio-data:persistentvolumeclaim'],
     new_name='infra',
 )

ctrl/k8s/base/configmap.yaml

@@ -8,4 +8,5 @@ data:
   DEBUG: "1"
   FASTAPI_PORT: "8702"
   DETECTION_UI_PORT: "5175"
+  INFERENCE_URL: ""
   GATEWAY_PORT: "8080"

ctrl/k8s/kind-config.yaml.tpl

@@ -7,6 +7,14 @@ nodes:
       - containerPort: 30080
         hostPort: 80
         protocol: TCP
+      - containerPort: 30379
+        hostPort: 6379
+        listenAddress: "0.0.0.0"
+        protocol: TCP
+      - containerPort: 30432
+        hostPort: 5432
+        listenAddress: "0.0.0.0"
+        protocol: TCP
     extraMounts:
       - hostPath: ${MEDIA_HOST_PATH}
         containerPath: /mnt/media

ctrl/k8s/overlays/dev/kustomization.yaml

@@ -5,6 +5,9 @@ resources:
   - ../../base
   - minio-pvc.yaml
 
+patchesStrategicMerge:
+  - local-config.yaml
+
 patches:
   # Gateway as NodePort for local access
   - target:
@@ -30,6 +33,18 @@ patches:
         path: /spec/ports/0/nodePort
         value: 30379
 
+  # Postgres as NodePort for external access
+  - target:
+      kind: Service
+      name: postgres
+    patch: |
+      - op: replace
+        path: /spec/type
+        value: NodePort
+      - op: add
+        path: /spec/ports/0/nodePort
+        value: 30432
+
   # MinIO with persistent storage + host media mount for seeding.
   # PV survives pod restarts. Host mount is read-only for mc mirror seeding.
   # Requires kind cluster created with MEDIA_HOST_PATH extraMount (see kind-create.sh).

detect/emit.py

@@ -6,6 +6,9 @@ Stages call these instead of constructing dicts or dataclasses directly.
 
 Run context (run_id, parent_job_id) is set once at pipeline start via
 set_run_context() and automatically injected into all events.
+
+Log level is set per-run with optional per-stage overrides.
+DEBUG events are only pushed when the run (or stage) log level allows it.
 """
 
 from __future__ import annotations
@@ -16,23 +19,53 @@ from datetime import datetime, timezone
 from detect.events import push_detect_event
 from detect.models import PipelineStats
 
+# Log level ordering for comparison
+_LEVEL_ORDER = {"DEBUG": 0, "INFO": 1, "WARN": 2, "ERROR": 3}
+
 # Module-level run context — set once per pipeline invocation
 _run_context: dict = {}
+_run_log_level: str = "INFO"
+_stage_log_levels: dict[str, str] = {}  # stage_name → level override
 
 
-def set_run_context(run_id: str = "", parent_job_id: str = "", run_type: str = "initial"):
+def set_run_context(
+    run_id: str = "",
+    parent_job_id: str = "",
+    run_type: str = "initial",
+    log_level: str = "INFO",
+):
     """Set the run context for all subsequent events in this pipeline invocation."""
-    global _run_context
+    global _run_context, _run_log_level
     _run_context = {
         "run_id": run_id,
         "parent_job_id": parent_job_id,
         "run_type": run_type,
     }
+    _run_log_level = log_level.upper()
+    _stage_log_levels.clear()
+
+
+def set_stage_log_level(stage: str, level: str):
+    """Override log level for a specific stage."""
+    _stage_log_levels[stage] = level.upper()
+
+
+def clear_stage_log_level(stage: str):
+    """Remove per-stage log level override."""
+    _stage_log_levels.pop(stage, None)
 
 
 def clear_run_context():
-    global _run_context
+    global _run_context, _run_log_level
     _run_context = {}
+    _run_log_level = "INFO"
+    _stage_log_levels.clear()
+
+
+def _should_emit(level: str, stage: str) -> bool:
+    """Check if this log level should be emitted given run/stage settings."""
+    effective = _stage_log_levels.get(stage, _run_log_level)
+    return _LEVEL_ORDER.get(level.upper(), 1) >= _LEVEL_ORDER.get(effective, 1)
 
 
 def _inject_context(payload: dict) -> dict:
@@ -45,6 +78,8 @@ def _inject_context(payload: dict) -> dict:
 def log(job_id: str | None, stage: str, level: str, msg: str) -> None:
     if not job_id:
         return
+    if not _should_emit(level, stage):
+        return
     payload = {
         "level": level,
         "stage": stage,

detect/graph.py

@@ -236,7 +236,8 @@ def node_escalate_vlm(state: DetectState) -> dict:
 
     existing = state.get("detections", [])
 
-    _emit_transition(state, "escalate_vlm", "done")
+    vlm_skipped = os.environ.get("SKIP_VLM", "").strip() == "1"
+    _emit_transition(state, "escalate_vlm", "skipped" if vlm_skipped else "done")
     return {
         "detections": existing + vlm_matched,
         "unresolved_candidates": still_unresolved,
@@ -268,7 +269,8 @@ def node_escalate_cloud(state: DetectState) -> dict:
 
     existing = state.get("detections", [])
 
-    _emit_transition(state, "escalate_cloud", "done")
+    cloud_skipped = os.environ.get("SKIP_CLOUD", "").strip() == "1"
+    _emit_transition(state, "escalate_cloud", "skipped" if cloud_skipped else "done")
     return {"detections": existing + cloud_matched, "stats": stats}
 
 
@@ -302,11 +304,33 @@ _CHECKPOINT_ENABLED = os.environ.get("MPR_CHECKPOINT", "").strip() == "1"
 _frames_manifest: dict[str, dict[int, str]] = {}  # job_id → manifest (cached per job)
 
 
+class PipelineCancelled(Exception):
+    """Raised when a pipeline run is cancelled."""
+    pass
+
+
+# Cancellation hook — set by the run endpoint, checked before each node
+_cancel_check: dict[str, callable] = {}
+
+
+def set_cancel_check(job_id: str, fn):
+    _cancel_check[job_id] = fn
+
+
+def clear_cancel_check(job_id: str):
+    _cancel_check.pop(job_id, None)
+
+
 def _checkpointing_node(node_name: str, node_fn):
     """Wrap a node function to auto-checkpoint after completion."""
     stage_index = NODES.index(node_name)
 
     def wrapper(state: DetectState) -> dict:
+        job_id = state.get("job_id", "")
+        check = _cancel_check.get(job_id)
+        if check and check():
+            raise PipelineCancelled(f"Cancelled before {node_name}")
+
         result = node_fn(state)
 
         job_id = state.get("job_id", "")

detect/inference/client.py

@@ -34,10 +34,16 @@ def _encode_image(image: np.ndarray) -> str:
 class InferenceClient:
     """HTTP client for the GPU inference server."""
 
-    def __init__(self, base_url: str | None = None, timeout: float = 60.0):
+    def __init__(self, base_url: str | None = None, timeout: float = 60.0,
+                 job_id: str = "", log_level: str = "INFO"):
         self.base_url = (base_url or DEFAULT_URL).rstrip("/")
         self.timeout = timeout
+        self.job_id = job_id
+        self.log_level = log_level
         self.session = requests.Session()
+        if job_id:
+            self.session.headers["X-Job-Id"] = job_id
+            self.session.headers["X-Log-Level"] = log_level
 
     def health(self) -> ServerStatus:
         """Check server health and loaded models."""

detect/stages/frame_extractor.py

@@ -8,6 +8,7 @@ Emits log + stats_update SSE events as it works.
 from __future__ import annotations
 
 import tempfile
+import time
 from pathlib import Path
 
 import ffmpeg
@@ -53,6 +54,8 @@ def extract_frames(
     emit.log(job_id, "FrameExtractor", "INFO",
              f"Starting extraction: {Path(video_path).name} "
              f"({duration:.1f}s, {probe.width}x{probe.height}, fps={config.fps})")
+    emit.log(job_id, "FrameExtractor", "DEBUG",
+             f"Probe: codec={probe.video_codec}, bitrate={probe.video_bitrate}, max_frames={config.max_frames}")
 
     with tempfile.TemporaryDirectory() as tmpdir:
         pattern = str(Path(tmpdir) / "frame_%06d.jpg")
@@ -65,14 +68,24 @@ def extract_frames(
             .overwrite_output()
         )
 
+        t0 = time.monotonic()
         try:
             stream.run(capture_stdout=True, capture_stderr=True, quiet=True)
         except ffmpeg.Error as e:
             stderr = e.stderr.decode() if e.stderr else "unknown error"
             emit.log(job_id, "FrameExtractor", "ERROR", f"FFmpeg failed: {stderr[:200]}")
             raise RuntimeError(f"FFmpeg failed: {stderr}") from e
+        ffmpeg_ms = (time.monotonic() - t0) * 1000
+        emit.log(job_id, "FrameExtractor", "DEBUG", f"FFmpeg decode: {ffmpeg_ms:.0f}ms")
 
+        t0 = time.monotonic()
         frames = _load_frames(Path(tmpdir), config.fps)
+        load_ms = (time.monotonic() - t0) * 1000
+        if frames:
+            h, w = frames[0].image.shape[:2]
+            mem_mb = sum(f.image.nbytes for f in frames) / (1024 * 1024)
+            emit.log(job_id, "FrameExtractor", "DEBUG",
+                     f"Loaded {len(frames)} frames ({w}x{h}) in {load_ms:.0f}ms, {mem_mb:.1f}MB in memory")
 
     emit.log(job_id, "FrameExtractor", "INFO", f"Extracted {len(frames)} frames")
     emit.stats(job_id, frames_extracted=len(frames))

detect/stages/ocr_stage.py

@@ -13,6 +13,7 @@ Model instances are cached at module level so they survive across pipeline runs.
 from __future__ import annotations
 
 import logging
+import time
 from typing import TYPE_CHECKING
 
 import numpy as np
@@ -91,7 +92,8 @@ def run_ocr(
     # Build these once per pipeline run, not per crop
     if inference_url:
         from detect.inference import InferenceClient
-        client = InferenceClient(base_url=inference_url)
+        from detect.emit import _run_log_level
+        client = InferenceClient(base_url=inference_url, job_id=job_id or "", log_level=_run_log_level)
     else:
         model = _get_local_model(config.languages[0])
 
@@ -108,12 +110,19 @@ def run_ocr(
             if crop.size == 0:
                 continue
 
+            t0 = time.monotonic()
             if inference_url:
                 raw_results = client.ocr(image=crop, languages=config.languages)
                 texts = [{"text": r.text, "confidence": r.confidence} for r in raw_results]
             else:
                 raw = model.ocr(crop)
                 texts = _parse_ocr_raw(raw, config.min_confidence)
+            ocr_ms = (time.monotonic() - t0) * 1000
+
+            h, w = crop.shape[:2]
+            text_preview = ", ".join(t["text"][:30] for t in texts) if texts else "(none)"
+            emit.log(job_id, "OCRStage", "DEBUG",
+                     f"Frame {seq} box {box.x},{box.y} ({w}x{h}): {ocr_ms:.0f}ms → {text_preview}")
 
             for t in texts:
                 candidates.append(TextCandidate(

detect/stages/scene_filter.py

@@ -9,6 +9,8 @@ CV stages without losing unique visual content.
 
 from __future__ import annotations
 
+import time
+
 import imagehash
 from PIL import Image
 
@@ -63,8 +65,16 @@ def scene_filter(
     emit.log(job_id, "SceneFilter", "INFO",
              f"Filtering {len(frames)} frames (hamming_threshold={config.hamming_threshold})")
 
+    t0 = time.monotonic()
     hashes = _compute_hashes(frames)
+    hash_ms = (time.monotonic() - t0) * 1000
+    emit.log(job_id, "SceneFilter", "DEBUG",
+             f"Computed {len(hashes)} perceptual hashes in {hash_ms:.0f}ms ({hash_ms/max(len(hashes),1):.1f}ms/frame)")
+
+    t0 = time.monotonic()
     kept = _dedup(frames, hashes, config.hamming_threshold)
+    dedup_ms = (time.monotonic() - t0) * 1000
+    emit.log(job_id, "SceneFilter", "DEBUG", f"Dedup pass: {dedup_ms:.0f}ms")
 
     dropped = len(frames) - len(kept)
     pct = (dropped / len(frames) * 100) if frames else 0

detect/stages/vlm_cloud.py

@@ -13,6 +13,8 @@ from __future__ import annotations
 import base64
 import io
 import logging
+import os
+import time
 
 import numpy as np
 from PIL import Image
@@ -108,6 +110,11 @@ def escalate_cloud(
     if not candidates:
         return []
 
+    if os.environ.get("SKIP_CLOUD", "").strip() == "1":
+        emit.log(job_id, "CloudLLM", "INFO",
+                 f"SKIP_CLOUD=1, skipping {len(candidates)} crops")
+        return []
+
     if not has_api_key():
         emit.log(job_id, "CloudLLM", "WARNING",
                  f"No API key set for cloud provider, skipping {len(candidates)} crops")
@@ -120,7 +127,7 @@ def escalate_cloud(
     matched: list[BrandDetection] = []
     total_cost = 0.0
 
-    for candidate in candidates:
+    for i, candidate in enumerate(candidates):
         crop = _crop_image(candidate)
         if crop.size == 0:
             continue
@@ -133,11 +140,15 @@ def escalate_cloud(
         prompt = vlm_prompt_fn(crop_context)
         image_b64 = _encode_crop(crop)
 
+        t0 = time.monotonic()
         try:
             result = _call_cloud_api(image_b64, prompt)
         except Exception as e:
-            logger.warning("Cloud LLM failed for '%s': %s", candidate.text, e)
+            call_ms = (time.monotonic() - t0) * 1000
+            emit.log(job_id, "CloudLLM", "DEBUG",
+                     f"[{i+1}/{len(candidates)}] FAILED '{candidate.text[:30]}': {e} ({call_ms:.0f}ms)")
             continue
+        call_ms = (time.monotonic() - t0) * 1000
 
         stats.cloud_llm_calls += 1
         model_info = provider.models.get(provider.model)
@@ -148,6 +159,11 @@ def escalate_cloud(
         brand = result["brand"]
         confidence = result["confidence"]
 
+        emit.log(job_id, "CloudLLM", "DEBUG",
+                 f"[{i+1}/{len(candidates)}] '{candidate.text[:30]}' → "
+                 f"{'✓ ' + brand if brand else '✗'} "
+                 f"(conf={confidence:.2f}, {result['tokens']}tok, ${call_cost:.4f}, {call_ms:.0f}ms)")
+
         if brand and confidence >= min_confidence:
             detection = BrandDetection(
                 brand=brand,

detect/stages/vlm_local.py

@@ -9,6 +9,8 @@ objects for crops the VLM can identify.
 from __future__ import annotations
 
 import logging
+import os
+import time
 
 import numpy as np
 
@@ -61,6 +63,11 @@ def escalate_vlm(
     if not candidates:
         return [], []
 
+    if os.environ.get("SKIP_VLM", "").strip() == "1":
+        emit.log(job_id, "VLMLocal", "INFO",
+                 f"SKIP_VLM=1, skipping {len(candidates)} crops")
+        return [], candidates
+
     emit.log(job_id, "VLMLocal", "INFO",
              f"Processing {len(candidates)} unresolved crops with moondream2")
 
@@ -69,9 +76,10 @@ def escalate_vlm(
 
     if inference_url:
         from detect.inference import InferenceClient
-        client = InferenceClient(base_url=inference_url)
+        from detect.emit import _run_log_level
+        client = InferenceClient(base_url=inference_url, job_id=job_id or "", log_level=_run_log_level)
 
-    for candidate in candidates:
+    for i, candidate in enumerate(candidates):
         crop = _crop_image(candidate)
         if crop.size == 0:
             still_unresolved.append(candidate)
@@ -84,6 +92,7 @@ def escalate_vlm(
         )
         prompt = vlm_prompt_fn(crop_context)
 
+        t0 = time.monotonic()
         try:
             if inference_url:
                 result = client.vlm(image=crop, prompt=prompt)
@@ -93,9 +102,16 @@ def escalate_vlm(
             else:
                 brand, confidence, reasoning = _vlm_local(crop, prompt)
         except Exception as e:
-            logger.warning("VLM failed for candidate '%s': %s", candidate.text, e)
+            vlm_ms = (time.monotonic() - t0) * 1000
+            emit.log(job_id, "VLMLocal", "DEBUG",
+                     f"[{i+1}/{len(candidates)}] FAILED '{candidate.text[:30]}': {e} ({vlm_ms:.0f}ms)")
             still_unresolved.append(candidate)
             continue
+        vlm_ms = (time.monotonic() - t0) * 1000
+        emit.log(job_id, "VLMLocal", "DEBUG",
+                 f"[{i+1}/{len(candidates)}] '{candidate.text[:30]}' → "
+                 f"{'✓ ' + brand if brand else '✗ unresolved'} "
+                 f"(conf={confidence:.2f}, {vlm_ms:.0f}ms)")
 
         if brand and confidence >= min_confidence:
             detection = BrandDetection(

detect/stages/yolo_detector.py

@@ -14,6 +14,7 @@ from __future__ import annotations
 import base64
 import io
 import logging
+import time
 
 from PIL import Image
 
@@ -32,10 +33,11 @@ def _frame_to_b64(frame: Frame) -> str:
     return base64.b64encode(buf.getvalue()).decode()
 
 
-def _detect_remote(frame: Frame, config: DetectionConfig, inference_url: str) -> list[BoundingBox]:
+def _detect_remote(frame: Frame, config: DetectionConfig, inference_url: str,
+                    job_id: str = "", log_level: str = "INFO") -> list[BoundingBox]:
     """Call the inference server over HTTP."""
     from detect.inference import InferenceClient
-    client = InferenceClient(base_url=inference_url)
+    client = InferenceClient(base_url=inference_url, job_id=job_id, log_level=log_level)
     results = client.detect(
         image=frame.image,
         model=config.model_name,
@@ -99,15 +101,24 @@ def detect_objects(
     all_boxes: dict[int, list[BoundingBox]] = {}
     total_regions = 0
 
-    for frame in frames:
+    for i, frame in enumerate(frames):
+        t0 = time.monotonic()
         if inference_url:
-            boxes = _detect_remote(frame, config, inference_url)
+            from detect.emit import _run_log_level
+            boxes = _detect_remote(frame, config, inference_url,
+                                   job_id=job_id or "", log_level=_run_log_level)
         else:
             boxes = _detect_local(frame, config)
+        det_ms = (time.monotonic() - t0) * 1000
 
         all_boxes[frame.sequence] = boxes
         total_regions += len(boxes)
 
+        emit.log(job_id, "YOLODetector", "DEBUG",
+                 f"Frame {frame.sequence}: {len(boxes)} regions in {det_ms:.0f}ms"
+                 f" [{', '.join(b.label for b in boxes)}]" if boxes else
+                 f"Frame {frame.sequence}: 0 regions in {det_ms:.0f}ms")
+
         if boxes and job_id:
             box_dicts = [{"x": b.x, "y": b.y, "w": b.w, "h": b.h,
                           "confidence": b.confidence, "label": b.label}

gpu/emit.py

@@ -0,0 +1,52 @@
+"""
+Lightweight event emitter for the GPU inference server.
+
+Pushes debug logs to the same Redis stream as the pipeline orchestrator,
+so GPU-side details (model load, VRAM, inference timing) appear in the
+same log panel.
+
+Only active when the request includes X-Job-Id header.
+No dependency on the detect package.
+"""
+
+from __future__ import annotations
+
+import json
+import os
+from datetime import datetime, timezone
+
+import redis
+
+REDIS_URL = os.environ.get("REDIS_URL", "redis://localhost:6379/0")
+EVENTS_PREFIX = "detect_events"
+
+_LEVEL_ORDER = {"DEBUG": 0, "INFO": 1, "WARN": 2, "ERROR": 3}
+
+_redis_client = None
+
+
+def _get_redis():
+    global _redis_client
+    if _redis_client is None:
+        _redis_client = redis.from_url(REDIS_URL, decode_responses=True)
+    return _redis_client
+
+
+def log(job_id: str, stage: str, level: str, msg: str, log_level: str = "INFO"):
+    """Push a log event to Redis if the level meets the threshold."""
+    if not job_id:
+        return
+    if _LEVEL_ORDER.get(level.upper(), 1) < _LEVEL_ORDER.get(log_level.upper(), 1):
+        return
+
+    r = _get_redis()
+    key = f"{EVENTS_PREFIX}:{job_id}"
+    event = json.dumps({
+        "event": "log",
+        "level": level,
+        "stage": stage,
+        "msg": msg,
+        "ts": datetime.now(timezone.utc).isoformat(),
+    })
+    r.rpush(key, event)
+    r.expire(key, 3600)

gpu/requirements.txt

@@ -2,6 +2,7 @@ fastapi>=0.109.0
 uvicorn[standard]>=0.27.0
 rapidfuzz>=3.0.0
 Pillow>=10.0.0
+redis>=5.0.0
 
 # --- GPU-specific installs (mcrn: RTX 3080, CUDA toolkit 12.8) ---
 #

gpu/server.py

@@ -14,13 +14,16 @@ import base64
 import io
 import logging
 import os
+import time
 from contextlib import asynccontextmanager
 
 import numpy as np
-from fastapi import FastAPI, HTTPException
+from fastapi import FastAPI, HTTPException, Request
 from PIL import Image
 from pydantic import BaseModel
 
+from emit import log as emit_log
+
 from config import get_config, get_device, update_config
 from models import registry
 from models.yolo import detect as yolo_detect
@@ -36,6 +39,19 @@ def _decode_image(b64: str) -> np.ndarray:
     return np.array(img)
 
 
+def _job_ctx(request: Request) -> tuple[str, str]:
+    """Extract job_id and log_level from request headers."""
+    job_id = request.headers.get("x-job-id", "")
+    log_level = request.headers.get("x-log-level", "INFO")
+    return job_id, log_level
+
+
+def _gpu_log(job_id: str, log_level: str, stage: str, level: str, msg: str):
+    """Emit a log event if job context is present."""
+    if job_id:
+        emit_log(job_id, stage, level, msg, log_level=log_level)
+
+
 # --- Request/Response models ---
 
 class DetectRequest(BaseModel):
@@ -160,19 +176,31 @@ def unload_model(body: dict):
 
 
 @app.post("/detect", response_model=DetectResponse)
-def detect(req: DetectRequest):
+def detect(req: DetectRequest, request: Request):
+    job_id, log_level = _job_ctx(request)
+
     try:
+        t0 = time.monotonic()
         image = _decode_image(req.image)
+        decode_ms = (time.monotonic() - t0) * 1000
+        h, w = image.shape[:2]
+        _gpu_log(job_id, log_level, "GPU:YOLO", "DEBUG",
+                 f"Decoded {w}x{h} image in {decode_ms:.0f}ms")
     except Exception as e:
         raise HTTPException(status_code=400, detail=f"Bad image: {e}")
 
     try:
+        t0 = time.monotonic()
         results = yolo_detect(
             image,
             model_name=req.model,
             confidence=req.confidence,
             target_classes=req.target_classes,
         )
+        infer_ms = (time.monotonic() - t0) * 1000
+        _gpu_log(job_id, log_level, "GPU:YOLO", "DEBUG",
+                 f"Inference: {len(results)} detections in {infer_ms:.0f}ms "
+                 f"(model={req.model}, conf={req.confidence})")
     except Exception as e:
         raise HTTPException(status_code=500, detail=f"Detection failed: {e}")
 
@@ -180,14 +208,22 @@ def detect(req: DetectRequest):
 
 
 @app.post("/ocr", response_model=OCRResponse)
-def ocr(req: OCRRequest):
+def ocr(req: OCRRequest, request: Request):
+    job_id, log_level = _job_ctx(request)
+
     try:
         image = _decode_image(req.image)
+        h, w = image.shape[:2]
     except Exception as e:
         raise HTTPException(status_code=400, detail=f"Bad image: {e}")
 
     try:
+        t0 = time.monotonic()
         results = ocr_run(image, languages=req.languages)
+        infer_ms = (time.monotonic() - t0) * 1000
+        texts = [r["text"][:20] for r in results]
+        _gpu_log(job_id, log_level, "GPU:OCR", "DEBUG",
+                 f"OCR {w}x{h}: {infer_ms:.0f}ms → {len(results)} results {texts}")
     except Exception as e:
         raise HTTPException(status_code=500, detail=f"OCR failed: {e}")
 
@@ -223,14 +259,22 @@ def preprocess_image(req: PreprocessRequest):
 
 
 @app.post("/vlm", response_model=VLMResponse)
-def vlm(req: VLMRequest):
+def vlm(req: VLMRequest, request: Request):
+    job_id, log_level = _job_ctx(request)
+
     try:
         image = _decode_image(req.image)
+        h, w = image.shape[:2]
     except Exception as e:
         raise HTTPException(status_code=400, detail=f"Bad image: {e}")
 
     try:
+        t0 = time.monotonic()
         result = vlm_query(image, req.prompt)
+        infer_ms = (time.monotonic() - t0) * 1000
+        _gpu_log(job_id, log_level, "GPU:VLM", "DEBUG",
+                 f"VLM {w}x{h}: {infer_ms:.0f}ms → "
+                 f"brand='{result.get('brand', '')}' conf={result.get('confidence', 0):.2f}")
     except Exception as e:
         raise HTTPException(status_code=500, detail=f"VLM failed: {e}")
 

ui/common/types/generated.ts

@@ -170,19 +170,6 @@ export interface SourceBrandSighting {
   created_at: string | null;
 }
 
-export interface SourceJob {
-  job_id: string;
-  source_type: string;
-  chunk_count: number;
-  total_bytes: number;
-}
-
-export interface ChunkInfo {
-  filename: string;
-  key: string;
-  size_bytes: number;
-}
-
 export interface CreateJobRequest {
   source_asset_id: string;
   preset_id: string | null;
@@ -220,6 +207,19 @@ export interface WorkerStatus {
   gpu_available: boolean;
 }
 
+export interface SourceJob {
+  job_id: string;
+  source_type: string;
+  chunk_count: number;
+  total_bytes: number;
+}
+
+export interface ChunkInfo {
+  filename: string;
+  key: string;
+  size_bytes: number;
+}
+
 export interface ChunkEvent {
   sequence: number;
   status: string;

ui/detection-app/src/App.vue

@@ -9,19 +9,27 @@ import FramePanel from './panels/FramePanel.vue'
 import BrandTablePanel from './panels/BrandTablePanel.vue'
 import TimelinePanel from './panels/TimelinePanel.vue'
 import CostStatsPanel from './panels/CostStatsPanel.vue'
+import SourceSelector from './panels/SourceSelector.vue'
 import type { StatsUpdate, RunContext } from './types/sse-contract'
 import { usePipelineStore } from './stores/pipeline'
 
 const pipeline = usePipelineStore()
 
-const jobId = ref(new URLSearchParams(window.location.search).get('job') || 'test-job')
+const jobParam = new URLSearchParams(window.location.search).get('job')
+const jobId = ref(jobParam || '')
 const stats = ref<StatsUpdate | null>(null)
 const runContext = ref<RunContext | null>(null)
 const status = ref<'idle' | 'live' | 'processing' | 'error'>('idle')
+const logPanel = ref<{ clear: () => void } | null>(null)
+
+// No job selected → open source selector
+if (!jobParam) {
+  pipeline.openSourceSelector()
+}
 
 const source = new SSEDataSource({
   id: 'detect-stream',
-  url: `/api/detect/stream/${jobId.value}`,
+  url: jobId.value ? `/api/detect/stream/${jobId.value}` : '',
   eventTypes: ['graph_update', 'stats_update', 'frame_update', 'detection', 'log', 'job_complete', 'waiting'],
 })
 
@@ -37,6 +45,12 @@ source.on<StatsUpdate>('stats_update', (e) => {
   }
 })
 
+source.on<{ report?: { status?: string, error?: string } }>('job_complete', (e) => {
+  if (e.report?.status === 'failed') {
+    status.value = 'error'
+  }
+})
+
 // Resizable splits
 const pipelineWidth = ref(320)
 const detectionsFlex = ref(3)  // ratio for detections vs stats
@@ -71,7 +85,36 @@ const statusMap: Record<string, 'idle' | 'live' | 'processing' | 'error'> = {
 const checkStatus = () => { status.value = statusMap[source.status.value] ?? 'idle' }
 setInterval(checkStatus, 500)
 
+if (jobId.value) {
   source.connect()
+}
+
+async function stopPipeline() {
+  if (!jobId.value) return
+  try {
+    await fetch(`/api/detect/stop/${jobId.value}`, { method: 'POST' })
+  } catch { /* ignore — UI will see the cancel event via SSE */ }
+}
+
+function onJobStarted(newJobId: string) {
+  jobId.value = newJobId
+  // Reset UI state
+  stats.value = null
+  runContext.value = null
+  status.value = 'processing'
+  logPanel.value?.clear()
+  pipeline.reset()
+  pipeline.setStatus('running')
+  // Update URL without reload
+  const url = new URL(window.location.href)
+  url.searchParams.set('job', newJobId)
+  window.history.pushState({}, '', url.toString())
+  // Connect SSE to new job
+  source.disconnect()
+  source.setUrl(`/api/detect/stream/${newJobId}`)
+  source.connect()
+  // Switch to normal layout (reset sets it to normal already)
+}
 </script>
 
 <template>
@@ -82,7 +125,18 @@ source.connect()
       <span v-if="runContext" class="run-info">
         {{ runContext.run_type }} · run: {{ runContext.run_id }}
       </span>
-      <span class="job-id">job: {{ jobId }}</span>
+      <button class="header-btn" title="Select source" @click="pipeline.openSourceSelector()">
+        <svg width="14" height="14" viewBox="0 0 16 16" fill="none" stroke="currentColor" stroke-width="1.5">
+          <path d="M2 4h4l2 2h6v8H2V4z"/><path d="M2 4V2h12v2"/>
+        </svg>
+      </button>
+      <button
+        v-if="jobId && (status === 'live' || status === 'processing')"
+        class="header-btn stop-btn"
+        title="Stop pipeline"
+        @click="stopPipeline"
+      >■</button>
+      <span class="job-id">job: {{ jobId || '—' }}</span>
     </header>
 
     <div class="main-layout">
@@ -168,6 +222,11 @@ source.connect()
           </Panel>
         </template>
 
+        <!-- === SOURCE SELECTOR MODE === -->
+        <template v-else-if="pipeline.layoutMode === 'source_selector'">
+          <SourceSelector @job-started="onJobStarted" />
+        </template>
+
       </div>
     </div>
 
@@ -183,7 +242,7 @@ source.connect()
         </Panel>
       </template>
       <template v-else>
-        <LogPanel :source="source" :status="status" />
+        <LogPanel ref="logPanel" :source="source" :status="status" />
       </template>
     </div>
   </div>
@@ -200,12 +259,11 @@ body {
 }
 
 .app {
-  height: 100vh;
+  min-height: 100vh;
   display: grid;
   grid-template-rows: auto 1fr auto;
   padding: var(--space-4);
   gap: var(--space-2);
-  overflow: hidden;
 }
 
 header {
@@ -235,6 +293,34 @@ header h1 { font-size: var(--font-size-lg); font-weight: 600; }
   font-size: var(--font-size-sm);
 }
 
+
+.header-btn {
+  background: var(--surface-2);
+  border: 1px solid var(--surface-3);
+  border-radius: 4px;
+  color: var(--text-secondary);
+  width: 28px;
+  height: 28px;
+  cursor: pointer;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  transition: all 0.15s;
+}
+.header-btn:hover {
+  background: var(--surface-3);
+  color: var(--text-primary);
+}
+.stop-btn {
+  background: var(--status-error);
+  color: #000;
+  font-size: 12px;
+  font-weight: 700;
+}
+.stop-btn:hover {
+  opacity: 0.8;
+}
+
 .job-id { color: var(--text-dim); font-size: var(--font-size-sm); margin-left: auto; }
 
 /* Main layout: pipeline left, content right — both same height */
@@ -328,11 +414,10 @@ header h1 { font-size: var(--font-size-lg); font-weight: 600; }
 .stat .label { color: var(--text-dim); font-size: var(--font-size-sm); }
 .stat .value { font-weight: 600; }
 
-/* Log: full width bottom, fixed height */
+/* Log: full width bottom */
 .log-row {
   flex-shrink: 0;
-  height: 160px;
+  height: 200px;
-  overflow: hidden;
 }
 
 .empty { color: var(--text-dim); padding: var(--space-6); text-align: center; }
@@ -399,4 +484,50 @@ header h1 { font-size: var(--font-size-lg); font-weight: 600; }
   text-align: center;
   font-size: var(--font-size-sm);
 }
+
+/* Source selector */
+.source-selector {
+  display: flex;
+  flex-direction: column;
+  height: 100%;
+  gap: var(--space-3);
+  padding: var(--space-3);
+}
+
+.source-info {
+  font-size: var(--font-size-sm);
+  color: var(--text-secondary);
+}
+
+.source-hint {
+  color: var(--text-dim);
+  margin-top: var(--space-1);
+}
+
+.source-hint code {
+  background: var(--surface-2);
+  padding: 1px 4px;
+  border-radius: 3px;
+}
+
+.source-list {
+  flex: 1;
+  overflow-y: auto;
+  background: var(--surface-2);
+  border-radius: var(--panel-radius);
+  padding: var(--space-2);
+}
+
+.source-loading {
+  color: var(--text-dim);
+  text-align: center;
+  padding: var(--space-4);
+  font-size: var(--font-size-sm);
+}
+
+.source-actions {
+  flex-shrink: 0;
+  display: flex;
+  justify-content: flex-end;
+}
 </style>

ui/detection-app/src/panels/LogPanel.vue

@@ -21,6 +21,12 @@ props.source.on<LogEvent>('log', (e) => {
     ts: e.ts,
   })
 })
+
+function clear() {
+  entries.value = []
+}
+
+defineExpose({ clear })
 </script>
 
 <template>

ui/detection-app/src/panels/PipelineGraphPanel.vue

@@ -26,6 +26,21 @@ props.source.on<{ nodes: GraphNode[] }>('graph_update', (e) => {
   nodes.value = e.nodes
 })
 
+props.source.on<{ report?: { status?: string } }>('job_complete', (e) => {
+  const status = e.report?.status
+  if (status === 'failed' || status === 'cancelled') {
+    nodes.value = nodes.value.map(n => ({
+      ...n,
+      status: n.status === 'running' ? 'error' : n.status,
+    }))
+  } else {
+    nodes.value = nodes.value.map(n => ({
+      ...n,
+      status: n.status === 'running' ? 'done' : n.status,
+    }))
+  }
+})
+
 function onOpenRegionEditor(stage: string) {
   pipeline.openBBoxEditor(stage)
 }

ui/detection-app/src/stores/pipeline.ts

@@ -22,7 +22,7 @@ export const usePipelineStore = defineStore('pipeline', () => {
   const error = ref<string | null>(null)
 
   // Layout mode
-  const layoutMode = ref<string>('normal')  // normal | bbox_editor | stage_editor
+  const layoutMode = ref<string>('normal')  // normal | bbox_editor | stage_editor | source_selector
   const editorStage = ref<string | null>(null)  // which stage's editor is open
 
   const isRunning = computed(() => status.value === 'running')
@@ -59,6 +59,11 @@ export const usePipelineStore = defineStore('pipeline', () => {
     if (msg) status.value = 'error'
   }
 
+  function openSourceSelector() {
+    layoutMode.value = 'source_selector'
+    editorStage.value = null
+  }
+
   function openBBoxEditor(stage: string) {
     layoutMode.value = 'bbox_editor'
     editorStage.value = stage
@@ -91,6 +96,6 @@ export const usePipelineStore = defineStore('pipeline', () => {
     checkpoints, error, layoutMode, editorStage,
     isRunning, isPaused, canReplay, isEditing,
     setJob, setStatus, updateNodes, setRunContext, setCheckpoints, setError,
-    openBBoxEditor, openStageEditor, closeEditor, reset,
+    openSourceSelector, openBBoxEditor, openStageEditor, closeEditor, reset,
   }
 })

ui/framework/src/components/Panel.vue

@@ -9,6 +9,7 @@ defineProps<{
   <div class="panel">
     <div class="panel-header">
       <span class="panel-title">{{ title }}</span>
+      <span class="panel-actions"><slot name="actions" /></span>
       <span class="panel-status" :class="status ?? 'idle'" />
     </div>
     <div class="panel-body">
@@ -51,11 +52,17 @@ defineProps<{
   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%;
-  margin-left: auto;
 }
 .panel-status.idle { background: var(--status-idle); }
 .panel-status.live { background: var(--status-live); }

ui/framework/src/datasources/SSEDataSource.ts

@@ -69,10 +69,9 @@ export class SSEDataSource extends DataSource {
       })
     }
 
-    // Also listen to the generic 'done' terminal event
+    // Terminal event — pipeline finished (success, failure, or cancel)
     this.es.addEventListener('done', () => {
       this.status.value = 'idle'
-      this.disconnect()
     })
   }
 

ui/framework/src/renderers/GraphRenderer.vue

@@ -6,7 +6,7 @@ import '@vue-flow/core/dist/theme-default.css'
 
 export interface GraphNode {
   id: string
-  status: 'pending' | 'running' | 'done' | 'error'
+  status: 'pending' | 'running' | 'done' | 'error' | 'skipped'
 }
 
 const props = defineProps<{
@@ -29,6 +29,7 @@ const statusColors: Record<string, string> = {
   running: 'var(--status-processing)',
   done: 'var(--status-live)',
   error: 'var(--status-error)',
+  skipped: '#4a6fa5',
 }
 
 const flowNodes = computed(() =>