Add notifications crate shared between API and indexer to send Telegram
messages on scan/thumbnail/conversion completion/failure, metadata linking,
batch and refresh events. Configurable via a new Notifications tab in the
backoffice settings with per-event toggle switches grouped by category.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add a complete metadata synchronization system allowing users to search
and sync series/book metadata from external providers (Google Books,
Open Library, ComicVine, AniList, Bédéthèque). Each library can use a
different provider. Matching requires manual approval with detailed sync
reports showing what was updated or skipped (locked fields protection).
Key changes:
- DB migrations: external_metadata_links, external_book_metadata tables,
library metadata_provider column, locked_fields, total_volumes, book
metadata fields (summary, isbn, publish_date)
- Rust API: MetadataProvider trait + 5 provider implementations,
7 metadata endpoints (search, match, approve, reject, links, missing,
delete), sync report system, provider language preference support
- Backoffice: MetadataSearchModal, ProviderIcon, SafeHtml components,
settings UI for provider/language config, enriched book detail page,
edit forms with locked fields support, API proxy routes
- OpenAPI/Swagger documentation for all new endpoints and schemas
Closes#3
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace sccache with a two-stage build strategy: dummy source files cache
dependency compilation in a separate layer, and --mount=type=cache for
Cargo registry/git/target directories. Source-only changes now skip
full dependency recompilation.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CBR: extract_cbr_page extrayait TOUT le CBR sur disque pour lire une
seule page. Reécrit avec le crate unrar : listing en mémoire + extraction
ciblée de la page demandée uniquement. Zéro subprocess, zéro temp dir.
PDF: render_pdf_page utilisait pdftoppm subprocess + temp dir. Reécrit
avec pdfium-render in-process. Zéro subprocess, zéro temp dir.
CBZ: sort naturel (natord) pour l'ordre des pages.
Dockerfile API: retire unar et poppler-utils, ajoute libpdfium.so.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Déplace les migrations du service `migrate` séparé vers un entrypoint.sh
- L'API exécute `sqlx migrate run` au démarrage avant de lancer le binaire
- Gestion de la rétrocompatibilité : détecte un schéma pre-sqlx et crée
une baseline `_sqlx_migrations` pour éviter les conflits sur les instances existantes
- Installe sqlx-cli dans le builder, copie le binaire et les migrations dans l'image finale
- Supprime le service `migrate` du docker-compose.yml ; l'indexer dépend maintenant
du healthcheck de l'API (qui garantit que les migrations sont appliquées)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add CLAUDE.md at root and AGENTS.md in apps/api, apps/indexer,
apps/backoffice, crates/parsers with module-specific guidelines
- Unify all service ports to 70XX (no more internal/external split):
API 7080, Indexer 7081, Backoffice 7082
- Update docker-compose.yml, Dockerfiles, config.rs defaults,
.env.example, backoffice routes, bench.sh, smoke.sh
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Install sccache in builder stage of both api and indexer Dockerfiles
- Configure RUSTC_WRAPPER to use sccache
- Use Docker cache mount (--mount=type=cache,target=/sccache) to persist cache
- Reduces build time significantly on subsequent builds by caching compiled artifacts
- Requires Docker BuildKit (enabled by default in Docker 23.0+)
Note: First build will still be slow (installs sccache + populates cache)
Subsequent builds will be much faster as dependencies are cached
