Skip to content

CLAUDE.md

Latest commit

 

History

History
54 lines (43 loc) · 2.59 KB

CLAUDE.md

File metadata and controls

54 lines (43 loc) · 2.59 KB
Raw

Nagare

Subtitle mining tool for Emby / Jellyfin / Plex. A Rust (axum + tokio) backend watches media-server playback sessions and AnkiConnect, then enriches Anki cards with audio, screenshots, and sentence/translation context. A Svelte frontend provides the player-side UI and configuration.

Layout

  • src/ — Rust backend
    • api.rs — axum HTTP/WebSocket routes, enrichment pipeline (perform_enrichment)
    • config.rs — Config and all persisted settings (serde). Stored as JSON in SQLite (app_config)
    • anki.rs — AnkiConnect client + the poller/AnkiBeacon push loop (run_anki_poller)
    • mining.rs — SQLite persistence (config, media history, mined notes)
    • session.rs — media-server session tracking, history entries, subtitle loading
    • media.rs — ffmpeg audio/screenshot/AVIF generation
    • subtitle.rs — subtitle parsing + sentence-to-line matching
    • media_server/ — per-server adapters: mediabrowser.rs (Emby/Jellyfin), plex.rs, shared types in mod.rs
  • frontend/src/ — Svelte app
    • lib/ConfigPage.svelte — settings UI (tabbed: Server / Anki & Media / Frontend)
    • lib/EnrichDialog.svelte — the card enrichment dialog
    • lib/stores.js, lib/api.js — client state and backend calls
  • api_reference/ — third-party media-server OpenAPI specs (see below)

API reference

When working with the media-server adapters in src/media_server/, consult the upstream OpenAPI specifications in api_reference/ rather than guessing endpoint shapes or field names:

  • api_reference/emby_openapi.json — Emby
  • api_reference/jellyfin-openapi-stable.json — Jellyfin
  • api_reference/plex_openapi.json — Plex

These are large; search them for the specific endpoint/field instead of reading them whole.

Build & check

  • Backend: cargo check / cargo build / cargo test (run from repo root)
  • Frontend: npm run build (or npm run dev) from frontend/

Notes

  • Config is persisted via SQLite, not a flat file; Config::save_to is legacy. The mining.auto_approve field is a migrated client-local setting and is stripped server-side.
  • Series tagging (anki.series_tag_enabled / anki.series_tag_parent): when enabled, enrichment adds a per-series tag in perform_enrichment. With a non-empty parent it is parent::Series_Name; with an empty parent it is just Series_Name. The series name comes from the media server's metadata (NowPlaying.series_name), plumbed through NowPlayingState / HistoryEntry (persisted in the media_history table) and MediaContext; movies fall back to the bare title.