feat: pipeline multiple per progetto con PipelineBar (#162)

[nikazzio]

Summary

• Nuova tabella pipelines come source of truth per configurazione, stage, stato di esecuzione e chunk tradotti per ogni pipeline; il testo sorgente è condiviso nel record projects
• pipelineService.ts nuovo servizio che gestisce CRUD pipeline, saveFullState, loadTranslations, restoreTranslations, saveChunkCheckpoint
• PipelineBar UI con cerchi numerati (1, 2…), nome inline editabile con font font-display italic (stesso stile di "Testo Originale"/"Traduzione Candidata"), pulsante "+" per aggiungere pipeline — si inserisce tra header e contenuto principale
• Test completamente aggiornati: nuovo pipelineService.test.ts, projectService.test.ts e projectStore.test.ts riscritti per la nuova architettura

Architettura

Il testo sorgente vive in projects (condiviso tra pipeline). Ogni pipeline ha: configurazione completa, stage, stato di esecuzione e chunk tradotti isolati. La colonna source_display_text nelle pipeline è nullable e predispone a un futuro override di sorgente per-pipeline senza schema changes.

Le migrazioni DB sono idempotenti (INSERT OR IGNORE, ensureColumn) e non rompono istanze esistenti — i dati vengono migrati da pipeline_configs → pipelines al primo avvio.

Test plan

• Aprire un progetto esistente → si apre correttamente con una sola pipeline
• Aggiungere una nuova pipeline con "+" → compare nella PipelineBar
• Rinominare una pipeline (doppio clic sul nome) → il nome si aggiorna
• Passare da una pipeline all'altra → i chunk tradotti cambiano correttamente
• Avviare la traduzione → i chunk vengono salvati nella pipeline attiva
• Riaprire il progetto → si ripristinano i chunk della pipeline attiva
• npm test → 341/341 test passano
• npx tsc --noEmit → nessun errore

[Copilot]

Pull request overview

This PR introduces multi-pipeline support per project by moving pipeline configuration/state/translations into a new pipelines table (and a new pipelineService), adds a PipelineBar UI to switch/manage pipelines, and updates persistence logic + tests to align with the new architecture.

Changes:

• Add pipelines as the source of truth for per-pipeline config/run-state/translations; keep source document text on projects.
• Introduce pipelineService.ts and refactor stores/hooks to load/save pipeline state by pipeline_id.
• Add PipelineBar to the document UI and update/replace service/store tests accordingly.

Reviewed changes

Copilot reviewed 21 out of 22 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
src/types.ts	Add Pipeline entity type used across store/service/UI.
src/stores/uiStore.ts	Adjust default showDocumentDrawer behavior when switching view modes.
src/stores/projectStore.ts	Track pipelines + active pipeline; refactor open/save to use pipelineService.
src/stores/projectStore.test.ts	Rewrite tests to mock pipeline/project/db services with new persistence flow.
src/stores/pipelineStore.ts	Remove activePipelineId from pipeline store (now owned by project store).
src/services/projectService.ts	Split out project “source text” persistence and create default pipeline on project creation.
src/services/projectService.test.ts	Update tests to validate getProjectSource/saveProjectSource.
src/services/projectService.sql.test.ts	Point SQL structural validation at pipelineService.ts.
src/services/pipelineService.ts	New service for pipeline CRUD + run state + translations persistence/restore.
src/services/pipelineService.test.ts	Add unit tests for the new pipeline service.
src/services/dbService.ts	Add pipelines table + migrations for multi-pipeline + translations.pipeline_id.
src/i18n/it.json / src/i18n/en.json	Add strings used by the new document empty-state UI cards.
src/hooks/usePipeline.ts	Switch checkpointing/run-state updates to pipelineService + active pipeline id from project store.
src/components/layout/PipelineBar.tsx	New UI to select/rename/delete/add pipelines.
src/components/layout/index.ts	Export PipelineBar.
src/components/document/DocumentView.tsx	Refresh empty state layout and add placeholder cards.
src/components/common/RunResumeBanner.tsx	Update run-state persistence to use pipelineService and project’s active pipeline id.
src/components/common/ConfirmDialog.tsx	Visual/layout adjustments for confirm modal.
src/App.tsx	Insert PipelineBar above DocumentView in document mode.
src-tauri/Cargo.lock	Bump app crate version.
CLAUDE.md	Add Git workflow guidance.

Comments suppressed due to low confidence (1)

src/services/pipelineService.ts:355

• Same issue as above: this bulk translation INSERT/UPSERT also omits project_id, which can leave rows orphaned on project deletion because only project_id has a cascading FK. Ensure new translation rows keep project_id populated (or add a proper FK/cascade on pipeline_id).

      `INSERT INTO translations (
         id, pipeline_id, original_text, final_translation, position, chunk_status, stage_results,
         judge_status, judge_rating, translation_locked, judge_issues, coherence_result, footnotes,
         source_display_text, source_processing_text, translation_display_text, translation_processing_text,
         blob_id, blob_order, blob_reference_chunk_ids
       ) VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16, $17, $18, $19, $20)

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 21 out of 22 changed files in this pull request and generated 5 comments.

Comments suppressed due to low confidence (1)

src/services/pipelineService.ts:484

• In restoreTranslations, parseJson<CoherenceResult>(row.coherence_result, undefined as unknown as CoherenceResult) again relies on an unsafe cast to allow undefined. This makes it easy to accidentally treat coherenceResult as always-present in future refactors. Prefer a parseJson<T>(value): T | undefined overload (or a dedicated parseJsonOptional) so optional JSON columns remain correctly typed without casts.

    const judgeResult = restoreJudgeResult(row);
    const stageResults = parseJson<Record<string, PipelineResult>>(row.stage_results, {});
    const coherenceResult = parseJson<CoherenceResult>(row.coherence_result, undefined as unknown as CoherenceResult);
    const footnotes = row.footnotes ? parseJson<Footnote[]>(row.footnotes, []) : undefined;
    const blobReferenceChunkIds = row.blob_reference_chunk_ids
      ? (parseJson<unknown>(row.blob_reference_chunk_ids, []) as string[]).filter((v): v is string => typeof v === 'string')
      : undefined;

[Copilot]

Pull request overview

Copilot reviewed 37 out of 38 changed files in this pull request and generated 3 comments.

Comments suppressed due to low confidence (1)

src/components/pipeline/ProductionStream.tsx:274

• recommendedChunks is computed with recommendChunkCount(inputText) (default 700 words/chunk), but the UI’s “Use recommendation” button applies chunkPresetMedium to wordsPerChunk. This can easily make the displayed recommendation inconsistent with what the button applies (especially if presets are user-tuned). Consider passing chunkPresetMedium into recommendChunkCount so the recommendation matches the preset baseline, or deriving a recommended wordsPerChunk from the computed chunk count.

  const { glossaryHighlightEnabled, setGlossaryHighlightEnabled, chunkPresetMedium } = useUiStore();
  const { t } = useTranslation();

  const stats = useMemo(() => estimateTextStats(inputText), [inputText]);
  const recommendedChunks = useMemo(() => recommendChunkCount(inputText), [inputText]);
  const enabledStages = useMemo(() => config.stages.filter((s) => s.enabled), [config.stages]);
  const hasGlossary = config.glossary.length > 0;
  const showHighlight = glossaryHighlightEnabled && hasGlossary;

[nikazzio]

Copilot review — all items addressed

Fixed in earlier commits:

• ✅ `translations` INSERT missing `project_id` — included in both `saveChunkCheckpoint` and `saveTranslationsInternal`
• ✅ `createPipeline` not passing `sourceLanguage`/`targetLanguage` — both params now passed
• ✅ `saveProjectSource` inside `runInTransaction` (deadlock) — moved outside the transaction
• ✅ PipelineBar confirm/delete dialog strings — all use `t()` keys
• ✅ `projectService.sql.test.ts` test name — corrected to match tested table
• ✅ `parseJson` unsafe cast — replaced with typed overloads
• ✅ PipelineBar aria-labels — all use `t()` keys
• ✅ `key={label}` translated string — replaced with stable keys (`'pipeline'`, `'translations'`, `'quality'`)
• ✅ `runInTransaction` unused import — removed

Fixed in commit 4578444:

• ✅ `PipelineBar` `aria-label="Pipeline"` hard-coded — now uses `t('pipeline.ariaGroup')`
• ✅ `restoreTranslations` — `blob_reference_chunk_ids` guarded with `Array.isArray` before `.filter()`

Design decision — `ON CONFLICT(id)` in translations:

Chunk IDs are generated as `chunk-${Date.now()}-${index}` and are globally unique by construction — cross-pipeline collision is not possible. `ON CONFLICT(id)` on the primary key is therefore semantically correct. Adding `pipeline_id` to the conflict target would require dropping and recreating the primary key (not supported by SQLite `ALTER TABLE`), introducing a risky migration with no practical benefit given the current ID generation strategy.

[Copilot]

Pull request overview

Copilot reviewed 55 out of 57 changed files in this pull request and generated 6 comments.