The /api/v1/media endpoint was loading ALL media items from the database
with no pagination, causing timeouts with large libraries (1800+ items
per source). Now supports:
- ?limit=N (default 50)
- ?offset=N (default 0)
- ?downloaded=true (filter to only downloaded media)
- Ordered by uploaded_at desc
- Remove unused aliases that caused warnings-as-errors in EX_CHECK mode
- Run mix format to fix formatter check
- Add root-level yarn install to CI for prettier (used by mix check)
- All 1007 tests pass, zero warnings
- Add FallbackController for proper 404 JSON responses on missing resources
- Add non-bang get_source/1 and get_media_item/1 for safe lookups
- Fix Task JSON encoder: remove Repo.preload call (not available in encoder scope),
handle NotLoaded associations by returning nil
- Fix media item index: use list_media_items_for_source/1 instead of
list_pending_media_items_for/1 which had overly restrictive filtering
- Add missing extras_directory to test config
- All 1007 tests pass
- Add /api/v1 namespace with Bearer token auth (PINCHFLAT_API_TOKEN env var)
- API controllers for sources, media items, media profiles, settings, tasks
- Support for all source actions: create, update, delete, force_download_pending,
force_redownload, force_index, force_metadata_refresh, sync_files_on_disk
- Media item endpoints: list, show, search, force_download, update, delete
- Media profile CRUD endpoints
- Settings show/update and app_info endpoints
- Task listing endpoints
- Add Jason.Encoder for Task schema
- Comprehensive test suite for all API endpoints
- Gitea Actions CI workflow (runs existing checks + API-specific tests)
- API documentation in docs/API.md
hermes-agent