feat: phrase memory foundation + semantic search workspace shell (#7)#205
Open
nikazzio wants to merge 20 commits into
Open
feat: phrase memory foundation + semantic search workspace shell (#7)#205nikazzio wants to merge 20 commits into
nikazzio wants to merge 20 commits into
Conversation
Contributor
There was a problem hiding this comment.
Pull request overview
This PR lays the groundwork for the “phrase memory” feature by introducing a new SQLite schema (workspaces + presets + embedding-related tables), wiring a first-run Workspace Wizard in the React app, and adding a Rust-side sqlite-vec auto-extension registration with a vec_ping Tauri command.
Changes:
- Adds Phrase Memory–related TypeScript domain types and services (workspaces + phrase memory presets), plus initial Zustand workspace store.
- Extends DB initialization to create the new tables, add
projects.workspace_id, seed built-in presets, and storeactive_workspace_idinapp_settings. - Adds a first-run WorkspaceWizard gate in
App.tsxand introduces Rustsqlite-vecintegration +vec_pingcommand.
Reviewed changes
Copilot reviewed 15 out of 15 changed files in this pull request and generated 25 comments.
Show a summary per file
| File | Description |
|---|---|
| src/types.ts | Adds Phrase Memory types (workspaces, presets, config enums). |
| src/stores/workspaceStore.ts | New Zustand store for workspace list/active workspace + loading state. |
| src/services/workspaceService.ts | New workspace CRUD + active-workspace setting helpers. |
| src/services/workspaceService.test.ts | Unit tests for workspaceService. |
| src/services/phraseMemoryPresetService.ts | Built-in preset seeding + preset CRUD/listing. |
| src/services/phraseMemoryPresetService.test.ts | Unit tests for preset service. |
| src/services/dbService.ts | Adds Phrase Memory schema creation, projects.workspace_id migration, active_workspace_id seeding, and preset seeding. |
| src/components/workspace/WorkspaceWizard.tsx | First-run wizard UI to create a workspace and select embedding model. |
| src/App.tsx | Adds workspace guard/wizard gating. |
| src-tauri/src/vector/mod.rs | Registers sqlite-vec auto-extension and adds vec_ping command. |
| src-tauri/src/lib.rs | Wires vector module init + exposes vec_ping command. |
| src-tauri/Cargo.toml | Adds rusqlite (bundled) and sqlite-vec dependencies. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
…olons, catch specifico, workspaceStore try/finally (#7)
nikazzio
added a commit
that referenced
this pull request
Jun 2, 2026
- vec_upsert_source_phrase: workspace_id → project_id (schema mismatch) - vec_save_locked_phrases: add project_id, source_language, target_language; fix DELETE to use project_id; fix saved counter via rows-changed - vec_search_phrase_memory: CTE elimina doppio vec_distance_cosine; propaga errori row invece di filter_map(ok) - split_phrases_llm: aggiunge controllo HTTP status come get_embeddings - phraseMemoryService: pre-filtra frasi per minPhraseLength prima dell'embed; aggiunge projectId, sourceLanguage, targetLanguage a SaveLockedPhrasesOptions - DocumentView: passa projectId e lingue a saveLockedPhrases - workspaceStore: auto-seleziona primo workspace se active_id non trovato - App: gestisce promise rejection di loadWorkspaces
nikazzio
changed the title
) - phraseMemoryStore: nuovi tipi PhraseMemoryMatch/ChunkPhraseMatches, Map-based state, toggleMatchEnabled, setEnabledMatchIds, conversione distance→score - uiStore: aggiunge 'memory' a ChunkDrawerTab - usePhraseMemoryMatches: hook per match + selezione per chunk - buildMemoryInjection: funzione pura per blocco stage-instructions - checkAllChunksHaveEnabledMatches: check pre-lancio pipeline - MemoryTab: lista match con checkbox, Applica (clipboard), Rielabora - ExtractTermDialog: dialog LLM term suggestion + inserimento glossario - InsightsDrawer: tab Memoria, badge match nell'IndexTab - usePipeline: rerunChunkWithMemory (injection temporanea stage prompt), warning pre-lancio per match tutti disabilitati - glossaryService: addGlossaryEntry wrapper - llmService: extractTermFromPhrase stub (TODO: piano 4 Tauri command) - i18n: chiavi memory.*, document.insightsTabMemory, glossary.*, common.optional Co-authored-by: nikazzio <nikazzio@users.noreply.github.com>
…#228) * fix(phrase-memory): address Piano 3 PR review feedback - ExtractTermDialog: granular Zustand selectors instead of full config + fix generateId prefix 'gle' - usePipeline: restore original prompts by stage ID after phrase memory injection (Map-based, race-safe) - glossaryService: direct SQL INSERT/ON CONFLICT instead of upsertGlossaryEntries - phraseMemoryInjection: JSON.stringify source/target phrases for correct escaping - phraseMemoryStore: new Set(ids) instead of raw array in enabledMatchIds - InsightsDrawer: remove unused matchesByChunk selector, use i18n key for match badge - i18n: add memory.matchBadge key (en + it) * feat(phrase-memory): Piano 4 — Preset Management UI + Pipeline Config Types: - Add PhraseMemoryOverrides to types.ts - Extend PipelineConfig with usePhraseMemory, phraseMemoryPresetId, phraseMemoryOverrides DB: - ALTER TABLE pipelines: add use_phrase_memory, phrase_memory_preset_id, phrase_memory_overrides columns Services: - phraseMemoryPresetService: add updateCustomPreset + clonePreset - pipelineService: DbPipeline + rowToPipelineConfig + savePipelineConfig + saveFullState + duplicatePipeline now persist phrase memory fields - pipelineService.test: phrase memory persistence tests Components: - PresetForm: form crea/modifica preset custom (splitter, threshold, maxResults, minPhraseLength) - PhraseMemoryPresetManager: lista preset built-in/custom con clone/edit/delete - PhraseMemoryConfig: sezione pipeline con toggle, dropdown preset, avanzate collassabili Integration: - SettingsModal: tab "Phrase Memory" con PhraseMemoryPresetManager - SettingsTabPanel: PhraseMemoryConfig in fondo alla tab Settings - PipelineConfig: passa phrase memory props a SettingsTabPanel * fix(phrase-memory): address PR #228 review comments - PhraseMemoryConfig: sync phraseMemoryPresetId to presets[0] when toggle enabled with null presetId - SettingsModal: move hardcoded strings to i18n (phraseMemoryTab, phraseMemoryPresetsTitle, phraseMemoryPresetsHint) - en.json/it.json: add matchBadge_one/matchBadge_other plural keys - en.json/it.json: add settings.phraseMemory* i18n keys --------- Co-authored-by: nikazzio <nikazzio@users.noreply.github.com>
Comment on lines
+276
to
+290
| let rows = conn | ||
| .execute( | ||
| "INSERT OR IGNORE INTO phrase_memory \ | ||
| (id, workspace_id, source_phrase, target_phrase, \ | ||
| source_language, target_language, embedding, created_at) \ | ||
| VALUES (lower(hex(randomblob(16))), ?1, ?2, ?3, ?4, ?5, ?6, datetime('now'))", | ||
| rusqlite::params![ | ||
| workspace_id, | ||
| pair.source_phrase, | ||
| pair.target_phrase, | ||
| source_language, | ||
| target_language, | ||
| floats_to_blob(&pair.source_embedding) | ||
| ], | ||
| ) |
Comment on lines
+495
to
+502
| try { | ||
| await conn.execute( | ||
| `ALTER TABLE projects ADD COLUMN workspace_id TEXT REFERENCES workspaces(id)` | ||
| ); | ||
| } catch (err) { | ||
| const msg = err instanceof Error ? err.message : String(err); | ||
| if (!msg.includes('duplicate column') && !msg.includes('already exists')) throw err; | ||
| } |
Comment on lines
+7
to
+13
| export function checkAllChunksHaveEnabledMatches( | ||
| matchesByChunk: Map<string, ChunkPhraseMatches>, | ||
| ): string[] { | ||
| const blocked: string[] = []; | ||
| for (const [chunkId, data] of matchesByChunk) { | ||
| if (data.matches.length > 0 && data.enabledMatchIds.size === 0) { | ||
| blocked.push(chunkId); |
…pread, embedding prompt
- backupService: add use_phrase_memory/preset_id/overrides to pipelines whitelist
- en.json / it.json: remove duplicate phraseMemory* keys in settings section
- PhraseMemoryConfig: fix spread on nullable overrides (overrides ?? {})
- embedding.rs split_phrases_llm: align prompt with json_object response_format ({"phrases":[...]})
…ection, workspace scoping Audit PR #205 fixes (issues 1-5): - fix: runSingleChunk/executePipelineForChunk/runJudgeForChunk read config from store at invocation time — memory patch in rerunChunkWithMemory now reaches prompt - fix: handleLockToggle guards use_phrase_memory flag; resolves splitter/minPhraseLength from active preset + overrides instead of hardcoded values - fix: phrase_memory_presets workspace-scoped via ALTER TABLE migration; listPresets/createCustomPreset/deleteCustomPreset/updateCustomPreset/clonePreset all require workspaceId; PhraseMemoryPresetManager + PhraseMemoryConfig updated - fix: split_phrases_llm parser simplified — removes unreachable as_array/sentences branches since response_format:json_object guarantees an object Phrase memory flow wiring (issue 3): - feat: searchPhraseMemoryBatch — one fetchEmbeddings API call for all chunks, then N local SQLite vec_search_phrase_memory queries; no N×API overhead - feat: saveAllCompletedPhrases — bulk variant of saveLockedPhrases with progress cb - feat: usePhraseMemoryAutoSearch — background hook triggered on project open and when use_phrase_memory toggled; populates matchesByChunk store automatically - feat: useSaveToMemory — explicit bulk "save to memory" action with progress; filters translationLocked/completed chunks; resolves preset+overrides config - feat: executePipelineForChunk accepts memoryBlock param; getChunkMemoryBlock reads enabled matches from store; runPipeline/runSingleChunk/runDryRun inject automatically - feat: MemoryTab shows "Salva in memoria" button (cold start + footer); progress inline - feat: handleLockToggle triggers runSearch after saveLockedPhrases for incremental refresh - i18n: saveToMemoryButton/savedToMemory/saveToMemoryFailed keys (it + en)
…e-memory pipeline fixes
Dashboard redesign:
- Replace flat area cards with unified tab panel (Traduzioni/Biblioteca/Trascrizioni)
- Active tab shares background with content below — visual continuity
- 2px accent bottom indicator on active tab; disabled tabs at 40% opacity
- WorkspaceSettingsModal: new modal with 3 tabs (Generale, Phrase Memory, Backup)
mirroring SettingsModal pattern (EditorialModalShell + AnimatePresence + useFocusTrap)
- SettingsModal: rename "Impostazioni" tab → "Traduzioni", add disabled Library/
Transcriptions tabs, remove Backup section (moved to WorkspaceSettingsModal)
- Apply UI polish: text-wrap balance/pretty, tabular-nums on metrics, concentric
border-radius (28→20→16px), transition-colors duration-150 throughout
Phrase Memory pipeline:
- Fix search pipeline lifecycle and workspace scoping
- Fix memory injection into prompt context
- Fix embedding search + workspace shell gating
- Add usePhraseMemoryAutoSearch and useSaveToMemory hooks with tests
- Add phraseMemoryService and workspaceService tests
i18n: add workspace.settings.{eyebrow,generalTab,memoryTab,backupTab} (en + it)
docs: update ARCHITECTURE.md with workspace/phrase-memory store and component map
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Cosa introduce questa PR
Phrase Memory — sistema di memoria semantica per le traduzioni. L'app ricorda le frasi già tradotte, le indicizza con embedding vettoriali e le suggerisce automaticamente durante la traduzione per garantire coerenza terminologica tra sessioni e documenti.
La PR raccoglie 4 piani di sviluppo implementati su branch dedicati e integrati qui.
Piano 1 — Fondamenta DB + Workspace
sqlite-vecviarusqliteper ricerca vettoriale localeworkspaces,phrase_memory,phrase_memory_presets,source_phrase_embeddings,historical_techniques,technique_tagsPiano 2 — Embedding search + Shell gating
Piano 3 — Tab Memoria + Injection nel prompt
Piano 4 — Preset Management UI + Pipeline Config
PhraseMemoryOverrides+ 3 campi suPipelineConfig(usePhraseMemory,phraseMemoryPresetId,phraseMemoryOverrides)pipelinesviaALTER TABLEidempotenteupdateCustomPreset+clonePresetsuphraseMemoryPresetServicePresetForm: crea/modifica preset custom (splitter, soglia, maxResults, minPhraseLength)PhraseMemoryPresetManager: lista preset built-in (sola lettura, clonabili) e custom (edit/delete)PhraseMemoryConfig: toggle + dropdown preset + sezione Avanzate collassabile con override per-pipelineNote architetturali
static → blob → stage-instructions) rimane invariato per preservare il prefix cachingPipelineConfig(già inpipelineStore)Test plan
npm test— tutti i test verdinpm run typecheck— nessun errorecargo clippy— zero warningScope fuori da questa PR