diff --git a/docs/architecture.md b/docs/architecture.md index b0a778a1..5de21ae5 100644 --- a/docs/architecture.md +++ b/docs/architecture.md @@ -114,7 +114,7 @@ Last.fm schemas (separate, not Ecto-persisted to main DB): | Module | Purpose | |--------|---------| | `Records.SearchParser` | Parses search syntax: `artist:X`, `album:X`, `genre:"Y"`, `format:cd`, `type:album`, `purchase_year:2024`, `release_year:2024`, free text | -| `Records.Similarity` | Embedding generation (OpenAI, enriched with Last.fm tags), cosine-distance search (sqlite-vec) | +| `Records.Similarity` | Embedding generation (OpenAI, enriched with Last.fm tags, skips API call when text representation unchanged), cosine-distance search (sqlite-vec) | | `Records.TracklistPdf` | Generates 120mm×120mm PDF tracklist from record + release data (Typst) | | `Batch` | Generic batch runner: stream + transaction + error accumulation | | `Records.Batch` | Batch operations: refresh all MusicBrainz data, generate all embeddings (uses `Batch`) | @@ -186,7 +186,7 @@ stubbed via `Req.Test` (configured in `config/test.exs`). | `FetchArtistImage` | heavy_writes | Artist info fetched | | `RefreshCover` | heavy_writes | Manual action / import | | `PopulateGenres` | heavy_writes | Manual action (chains → GenerateRecordEmbedding) | -| `GenerateRecordEmbedding` | heavy_writes | Manual / after genre population | +| `GenerateRecordEmbedding` | heavy_writes | Manual / after genre population (delegates to Similarity, skips unchanged) | | `RecordRefreshMusicBrainzData` | music_brainz | Manual / batch | | `ArtistRefreshMusicBrainzData` | music_brainz | Manual / batch | | `ArtistRefreshDiscogsData` | discogs | Manual / batch |