After GAP-5 removed the data-advanced-section placeholder div, clicking 'ver JSON'
or 'ver pipeline' in ResultNarrated would open the accordion via URL state but
not scroll to it. Restore by adding id='translator-advanced-section' to the
AdvancedSection root and using getElementById + scrollIntoView (with rAF defer
so React commits the open state first).
- Move useTranslateSession() to TranslatorPageClient (shell level) for PipelineView to receive real steps
- Build PipelineStep[] from session result (detected/intermediate/translated/response)
- Pass session as prop down to TranslateTab; internal hook kept for isolated test compatibility
- Dedup useProviderOptions: only TranslateTab calls the hook, props pass to SimpleControls
- Remove unused data-advanced-section/data-input-text placeholder (DOM data leak GAP-5)
- Add aria-expanded + aria-controls to AutoFeaturesCard toggle (a11y GAP-2)
Addresses code review RISCO-4, GAP-2, GAP-3, GAP-5.
- store.ts: wrap markMemoryNeedsReindex in safeMarkNeedsReindex helper that swallows
errors when the DB is no longer available (e.g. test teardown after the parent
promise resolved). Prevents fire-and-forget vector upserts from triggering
unhandledRejection in tests.
- memory-store.test.ts: drain setImmediate in afterEach/after hooks so pending
vector upsert tasks settle before DATA_DIR is removed.
- memory-settings.test.ts: extend deepEqual expected shape with the 7 new fields
introduced by plan 21 F5 (embeddingSource, embeddingProviderModel,
transformersEnabled, staticEnabled, rerankEnabled, rerankProviderModel,
vectorStore).
New routes:
- POST /api/memory/retrieve-preview (dry-run playground)
- GET /api/memory/embedding-providers
- GET /api/memory/engine-status
- POST /api/memory/summarize
- POST /api/memory/reindex
- GET/PUT /api/settings/qdrant
- GET /api/settings/qdrant/health
- POST /api/settings/qdrant/search
- POST /api/settings/qdrant/cleanup
Modified:
- PUT /api/memory/[id] added (Hard Rule #12 sanitize)
- /api/memory/route.ts: Hard Rule #12 fix (sanitizeErrorMessage)
- /api/settings/memory/route.ts: MemorySettingsExtendedSchema (D9 7 new fields)
Tests: 7 integration test files (33 tests total) all passing.
Hard Rules #5, #7, #8, #12 verified.
The previous expression was parsed as (A && B && C) || D, allowing D to evaluate
with creds possibly null. Wrap (apiKey || accessToken) in parens so creds-narrowing
covers the whole disjunction.
ChatPlayground.tsx → ChatTab (markdown + system prompt + metrics).
SearchPlayground.tsx → superseded by /dashboard/search-tools Studio (F8).
grep confirmed zero external imports before deletion.
Refactors ChatPlayground.tsx into ChatTab — multi-turn SSE chat with:
markdown rendering via MarkdownMessage (F1), system prompt from config pane,
token/cost per message via useStreamMetrics (F5), regenerate button, and
stop/cancel support. Hard Rule compliance: no useCallback to satisfy
react-hooks/preserve-manual-memoization rule.
hookBufferStart was calling recordRequestStart without sourceModel,
causing the field to be null even when extractSourceModel returned
a value from the body. Now forwards the extracted model so the
Traffic Inspector buffer entry is populated correctly.
GitLawB's OpenGateway API does not expose a /models endpoint per
provider-path. Previously the models route fell through to the generic
fallback which returned static catalog models with the misleading
'API unavailable — using local catalog' warning.
Now gitlawb and gitlawb-gmi are handled as static model providers
(same pattern as reka and qwen OAuth) — models are served from the
provider registry without any warning, since all registered models
are functional via POST /chat/completions.
