The top-level await redisClient.connect() in redis-client.js was
blocking module import when no Redis server is available (CI env).
Add NODE_ENV guard + reconnect strategy with max retries.
- Lazy-load tesseract.js only in OCR route (skips 30MB WASM download)
- Conditional tracing.js import (skip OTel gRPC in test mode)
- Remove unused barcode-detector import
- Add testTimeout/openHandlesTimeout to jest.config.js
- Add --forceExitTimeout=30000 to test script
- Split CI test jobs into parallel workflows (backend/frontend/mobile)
- Add timeout-minutes: 15 to all test jobs
- Use npm ci --ignore-scripts + selective native rebuild
The backend test was hanging in CI because server.js imports off-service.js
which pulls in redis-client.js. Without a mock, it tries to connect to Redis
at import time, hanging indefinitely in CI where no Redis server exists.
Also added the missing searchOTC mock for cima-service.js to fix the
SyntaxError from the export mismatch.
- READ operations don't require auth per OFF docs (only User-Agent)
- Auth credentials were causing 503 errors
- Keep User-Agent header which is required
- Switch from /cgi/search.pl (503 errors) to /api/v2/search with brand filter
- Add required User-Agent header per OFF documentation
- Add OFF_USER_AGENT env var for customization
- Update .env.example with User-Agent documentation
- Retry logic continues to work with new endpoint
- Add retry logic with exponential backoff (2 retries)
- Remove custom User-Agent that was being blocked by OFF
- Fix syntax error (missing catch block)
- OFF API blocks anonymous users during high demand
- Add OFF_USERNAME and OFF_PASSWORD env vars
- Use Basic Auth when credentials are provided
- Improves rate limits for authenticated requests
- Auth is optional (works without credentials)
- Validate response is JSON before parsing (OFF returns HTML on 503)
- Don't cache empty results to avoid blocking future searches
- Add validateStatus to only accept 200 responses
- Better logging for API failures
- Changed from /api/v2/search (returning 503) to /cgi/search.pl (working)
- Removed strict categories_tags=baby-food filter that was too restrictive
- Fixed test file to use dynamic imports for ES module compatibility
- Increased timeout to 10s for reliability
- GET /api/products/search?q={query} - parallel search across both sources
- GET /api/products/:source/:id - product detail by source (cima/openfoodfacts)
- Uses Promise.allSettled for fault-tolerant parallel search
- Added imports for searchOTC and OFF service functions
- transformOFFProduct: maps OFF products to unified Product model
- searchBabyProducts: searches baby products via OFF API (1h cache)
- getBabyProductDetails: fetches product details by barcode (1h cache)
- 9 tests passing for transform logic and input validation
- Added avatar selection popup with predefined avatar options
- Redesigned ProfileView (web + mobile) with avatar upload and editable fields
- Added AvatarSelectionModal for mobile profile
- Fixed medicine search: parse dosage terms (e.g. 'Paracetamol 1G')
to query CIMA only by name and filter by dosage field
- Updated styles for profile, search, and medicine results
Antoni Nuñez Romeu