RecallBridge is a source-linked product recall search and watchlist application. It brings recent consumer product, food, and medical device recall records into one normalized interface, then explains why a saved product may match a recall.
RecallBridge is an educational open-source project, not an official alerting service. Always confirm a result with the publishing agency before taking action.
Recall information is published across separate agency systems with different schemas. RecallBridge provides one practical workflow for:
- browsing recent official records;
- searching by product, manufacturer, model, UPC, or hazard;
- checking a product with explainable match confidence;
- keeping a browser-local watchlist without creating an account; and
- opening the original agency record for verification.
| Source | Coverage used in this project | Connector |
|---|---|---|
| U.S. Consumer Product Safety Commission | Consumer products | SaferProducts.gov recall API |
| openFDA Food Enforcement | Food enforcement reports | openFDA API |
| openFDA Device Recall | Medical device recalls | openFDA API |
The repository includes six clearly labeled fictional records so the app works before its first sync. A successful official-data sync removes those demo records. The generated SQLite database is intentionally excluded from Git.
Search, filter, inspect hazards and remedies, and follow every result back to its publishing source.
Enter a product name or identifier. RecallBridge ranks possible matches and shows the evidence behind each confidence level.
The operational layout collapses into a compact mobile workflow without hiding source or safety information.
RecallBridge uses deterministic, inspectable rules rather than a black-box classifier:
- Exact: a model, UPC, or lot identifier matches.
- Strong: multiple product, brand, manufacturer, or category signals agree.
- Possible: a smaller number of descriptive signals overlap.
Every result contains human-readable match reasons. A match is a lead for verification, not proof that a particular item is recalled.
Official APIs
|
v
Source connectors -> normalized Pydantic records -> SQLite index
|
React client <- FastAPI search and match endpoints+
|
Browser-local watchlist
recallbridge/
├── backend/
│ ├── data/ # fictional seed records; generated DB is ignored
│ ├── recallbridge/ # API, storage, matching, and source connectors
│ └── tests/
├── e2e/ # Playwright user-flow tests
├── scripts/sync_recalls.py # official source synchronization
├── src/ # React and TypeScript client
├── CONTRIBUTING.md
├── SECURITY.md
└── README.md
Requirements: Python 3.12+, Node.js 22+, and npm.
git clone https://github.com/gokul-debugger/recallbridge.git
cd recallbridge
python3.12 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
pip install -r backend/requirements-dev.txt
npm ciSynchronize recent official records:
python scripts/sync_recalls.py --days 365 --limit 100Start the API:
uvicorn recallbridge.app:app --app-dir backend --reload --port 8000In a second terminal, start the client:
npm run devOpen http://127.0.0.1:5173. API documentation is available at http://127.0.0.1:8000/docs.
RecallBridge ships as one Docker service: FastAPI serves both the API and the built React application. In hosted deployments, an asynchronous refresh task updates the SQLite recall index from the official sources every six hours while keeping the last successful records available if a source is temporarily down.
Live application: recallbridge.onrender.com
API documentation: recallbridge.onrender.com/docs
Render's free web services use an ephemeral filesystem and may spin down while
idle. The recall index is therefore rebuilt after a restart. A production
deployment should set RECALLBRIDGE_DB to a persistent volume path.
npm run lint
npm run test
npm run build
pytest -q
ruff check backend scripts
npx playwright install chromium
npm run test:e2eCurrent automated coverage includes connector normalization, search and matching APIs, exact and possible match behavior, browser-local watchlist persistence, recall browsing, and responsive product-save flows.
- Watchlist products stay in the current browser's local storage.
- No user account, analytics tracker, or product identifier is sent to a third party by the client.
- The backend requests only public recall records from the listed agencies during an explicit local sync or a configured hosted refresh.
- openFDA notes that enforcement reports should not be treated as an official alerting or recall lifecycle service. RecallBridge preserves source links and presents the data as a searchable index.
Please report security concerns using the process in SECURITY.md.
- Initial coverage is limited to the three listed U.S. data feeds.
- Source schemas and publication delays differ.
- Text matching can miss aliases or produce possible matches that require manual review.
- The watchlist does not perform background notifications. Recheck saved products after syncing data.
- RecallBridge does not replace agency guidance, manufacturer instructions, or professional advice.
- Add source adapters for more countries and agencies.
- Add scheduled sync metadata and stale-source warnings.
- Support watchlist import and export without an account.
- Improve identifier normalization and add source-specific contract tests.
- Add optional, privacy-preserving notifications as a separate service.
Contributions are welcome. Start with CONTRIBUTING.md and an issue describing the user problem you want to solve.
MIT © 2026 Gokul Krishna