diff --git a/docker-compose.localdev.yml b/docker-compose.localdev.yml
index 88dab27d..202ca5b8 100644
--- a/docker-compose.localdev.yml
+++ b/docker-compose.localdev.yml
@@ -1,138 +1,54 @@
-# Self-contained local development stack — Postgres + API + scheduler, all on this machine.
+# Self-contained local development stack — Postgres + API + scheduler + worker, all on this machine.
+# An OVERRIDE of docker-compose.yml, not a copy of it: only what local development changes lives
+# here, so the two cannot drift apart (they had — the worker here lost its wait for the API's
+# migrations).
 #
-# Fully decoupled: its own local database (a Docker volume) and its own scheduler, so there
-# is no shared DB and no migration coupling with any other instance. Edit code, rebuild, and
-# everything (including the background sync) runs right here.
+#   docker compose -f docker-compose.yml -f docker-compose.localdev.yml up -d --build
 #
-#   docker compose -f docker-compose.localdev.yml up -d --build
-#
-# Browse at http://localhost:8080 — Google login works because the OAuth redirect is
-# http://localhost:8080/auth/callback. The API runs `alembic upgrade head` on startup, so it
-# applies local migrations to THIS local DB only. Seed the catalog once with a pg_dump from
-# another instance if you want real data (see RUNBOOK).
-
-# Bounded container logs. Docker's json-file driver keeps every line forever by default, and the API
-# now writes one structured line per request (E22 S2) — 3×50 MB per service is days of history on a
-# busy instance and a hard ceiling on the disk it can take.
-x-logging: &logging
-  driver: json-file
-  options:
-    max-size: "50m"
-    max-file: "3"
+# Fully decoupled: its own local database (a Docker volume) and its own scheduler, so there is no
+# shared DB and no migration coupling with any other instance. Browse at http://localhost:8080 —
+# Google login works because the OAuth redirect is http://localhost:8080/auth/callback. The API runs
+# `alembic upgrade head` on startup, so it applies local migrations to THIS local DB only. Seed the
+# catalog once with a pg_dump from another instance if you want real data (see RUNBOOK).
 
 services:
   db:
-    image: postgres:16-alpine
-    environment:
-      POSTGRES_USER: ${POSTGRES_USER:-siftlode}
-      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-siftlode}
-      POSTGRES_DB: ${POSTGRES_DB:-siftlode}
-    volumes:
-      - pgdata:/var/lib/postgresql/data
+    # Harmless on Docker Desktop; needed on some nested/hardened Docker hosts (every service below).
     security_opt:
       - apparmor:unconfined
-    healthcheck:
-      test: ["CMD-SHELL", "pg_isready -U ${POSTGRES_USER:-siftlode} -d ${POSTGRES_DB:-siftlode}"]
-      interval: 5s
-      timeout: 5s
-      retries: 12
-    logging: *logging
-    restart: unless-stopped
 
   api:
-    build:
-      context: .
-      dockerfile: Dockerfile
-      args:
-        APP_VERSION: ${APP_VERSION:-dev}
-        GIT_SHA: ${GIT_SHA:-unknown}
-        BUILD_DATE: ${BUILD_DATE:-}
-    env_file: .env
     environment:
-      # Own local database (overrides whatever DATABASE_URL is in .env).
-      DATABASE_URL: postgresql+psycopg://${POSTGRES_USER:-siftlode}:${POSTGRES_PASSWORD:-siftlode}@db:5432/${POSTGRES_DB:-siftlode}
-      # This instance owns its scheduler now (its own DB, so no double-write concern).
+      # This instance owns its scheduler (its own DB, so no double-write concern).
       SCHEDULER_ENABLED: "true"
-      # Download center: the API serves finished files (it does NOT run the worker loop).
-      DOWNLOAD_ROOT: /downloads
-      WORKER_ENABLED: "false"
       # Plex HLS remux scratch on fast container-local storage (the /downloads bind-mount is a slow
       # Windows Docker Desktop mount in dev; prod uses fast local disk so it keeps the default).
       PLEX_HLS_DIR: /var/tmp/plex-hls
-    depends_on:
-      db:
-        condition: service_healthy
-    # Harmless on Docker Desktop; needed on some nested/hardened Docker hosts.
     security_opt:
       - apparmor:unconfined
-    ports:
-      - "${APP_PORT:-8080}:8000"
     volumes:
-      # Downloaded media lives on the host so you can inspect the generated Plex tree.
-      # Gitignored. The worker writes here; the API reads it to serve local downloads.
+      # Downloaded media lives on the host so you can inspect the generated Plex tree. Gitignored.
       - ./downloads:/downloads
-      # Plex media, read-only (see the plex_media volume note below). Maps to plex_path_map.
       - plex_media:/plex-media:ro
-    logging: *logging
-    restart: unless-stopped
 
-  # Dedicated download worker: same image, runs the yt-dlp job loop instead of the API.
-  # It shares the DB (job queue) and the downloads mount with the API.
   worker:
-    build:
-      context: .
-      dockerfile: Dockerfile
-      args:
-        APP_VERSION: ${APP_VERSION:-dev}
-        GIT_SHA: ${GIT_SHA:-unknown}
-        BUILD_DATE: ${BUILD_DATE:-}
-    command: ["python", "-m", "app.worker"]
-    # tini as PID 1: reaps orphaned ffmpeg/yt-dlp grandchildren when the watchdog hard-kills a
-    # download child's process group (the worker reaps its own direct children via subprocess.wait).
-    init: true
-    env_file: .env
-    environment:
-      DATABASE_URL: postgresql+psycopg://${POSTGRES_USER:-siftlode}:${POSTGRES_PASSWORD:-siftlode}@db:5432/${POSTGRES_DB:-siftlode}
-      # The worker must not also run the scheduler; the API owns that.
-      SCHEDULER_ENABLED: "false"
-      DOWNLOAD_ROOT: /downloads
-      WORKER_ENABLED: "true"
-    depends_on:
-      db:
-        condition: service_healthy
-      bgutil-pot:
-        condition: service_started
     security_opt:
       - apparmor:unconfined
     volumes:
       - ./downloads:/downloads
       - plex_media:/plex-media:ro
-    logging: *logging
-    restart: unless-stopped
 
-  # PO-token provider: mints YouTube Proof-of-Origin tokens on demand so the worker bypasses
-  # bot-detection and gets high-quality formats. The worker reaches it at http://bgutil-pot:4416
-  # (DOWNLOAD_POT_BASE_URL). Pin the tag to match the plugin version in requirements.txt.
   bgutil-pot:
-    image: brainicism/bgutil-ytdlp-pot-provider:1.3.1
-    init: true
     security_opt:
       - apparmor:unconfined
-    logging: *logging
-    restart: unless-stopped
 
 volumes:
-  pgdata:
   # OPTIONAL (Plex integration dev testing): the Plex media, mounted READ-ONLY so the container can
   # play the physical file locally (this dev box isn't the Plex host). Only referenced when the Plex
-  # module is enabled; harmless otherwise (Docker mounts it lazily on first container use). On prod
-  # (LXC on the Plex host) this is a plain host bind-mount.
-  #
-  # Local bind to a few test movies/shows kept on this box — the full nas.lan SMB share was too slow.
-  # Point PLEX_MEDIA_HOST_PATH (.env, gitignored, so no site path lives in the tracked compose) at the
-  # Docker-Desktop host path (e.g. /run/desktop/mnt/host/d/plex-test-media for D:\plex-test-media),
-  # mirroring the Plex server's /data/<lib>/… layout so plex_path_map (/data=/plex-media) resolves.
-  # (The prior CIFS/PLEX_SMB_* approach for the whole nas.lan share lives in git history.)
+  # module is enabled; harmless otherwise. Point PLEX_MEDIA_HOST_PATH (.env, gitignored, so no site
+  # path lives in the tracked compose) at the Docker-Desktop host path (e.g.
+  # /run/desktop/mnt/host/d/plex-test-media for D:\plex-test-media), mirroring the Plex server's
+  # /data/<lib>/… layout so plex_path_map (/data=/plex-media) resolves.
   plex_media:
     driver: local
     driver_opts: