doocus first ver

Makefile

@@ -24,7 +24,8 @@ FLAGS ?= --embed-images --scene-detection --scene-threshold 10 --diarize --trans
 # Per-run knobs (appended after FLAGS): language, plus any one-off extra flags
 # (e.g. --skip-cache-whisper, --transcript-formats srt).
 # NB: named AUDIO_LANG, not LANG — LANG is the shell locale env var.
-AUDIO_LANG ?=
+# Defaults to English; override (e.g. AUDIO_LANG=es) for other languages.
+AUDIO_LANG ?= en
 EXTRA      ?=
 
 # Optional: restrict scanned extensions / pick the python interpreter.
@@ -32,7 +33,13 @@ EXT    ?=
 PYTHON ?=
 export PYTHON
 
-.PHONY: batch dry help
+# doocus (document extraction) knobs. DOCS_OUT defaults to docs-output; the input
+# tree is mirrored under it. DOC_EXTRA forwards one-off flags to process_doc.py
+# (e.g. --render for pdf page thumbnails, --ocr for images/scanned pdfs).
+DOCS_OUT  ?= docs-output
+DOC_EXTRA ?=
+
+.PHONY: batch dry docs docs-dry help
 
 batch:
 	@ctrl/batch.sh -i "$(IN)" -o "$(OUT)" $(if $(EXT),-e "$(EXT)") -- \
@@ -41,5 +48,16 @@ batch:
 dry:
 	@ctrl/batch.sh -i "$(IN)" -o "$(OUT)" $(if $(EXT),-e "$(EXT)") -n
 
+# Document extraction (doocus), mirrors the input tree into DOCS_OUT.
+#   make docs IN="/mnt/win/drive"                    # extract all docs
+#   make docs IN="..." DOC_EXTRA="--render --ocr"    # thumbnails + OCR
+#   make docs-dry IN="..."                           # list what would run
+docs:
+	@ctrl/batch_docs.sh -i "$(IN)" -o "$(DOCS_OUT)" $(if $(EXT),-e "$(EXT)") -- \
+		$(DOC_EXTRA)
+
+docs-dry:
+	@ctrl/batch_docs.sh -i "$(IN)" -o "$(DOCS_OUT)" $(if $(EXT),-e "$(EXT)") -n
+
 help:
 	@sed -n '3,14p' Makefile