Skip to main content

About the project

Recall Hunter helps identify recalled products that may still be for sale online. Its browser extension, Recall Guard, checks shopping listings against official U.S. recall data and warns shoppers about possible matches, while the dashboard helps safety teams review evidence and track sellers. The system uses Bright Data to find listings, Gemini to assess potential matches, and Tiger Data to store results and monitoring history. It shows the official recall source and uncertainty instead of claiming a product is guaranteed safe.

Built with

typescriptpythoncsshtmlgeminiopenaiclaudellama

AI tools

Coded withClaude CodeCodexCursorGemini 3.5 Flash LiteGemini 4 HydrogenOlder Gemini

Prize Categories Entered

  • Gemini Prize Category Q1 2025 - Q1 2026

    Best Use of Gemini API

    Submitted

    We used Flash-Lite to screen product titles and confirm possible recall matches, Flash and 4-hydrogen to compare product photos and identify labels, and Embedding 2 to support recall and image search. Gemini also helped rank alternatives; 3.1 Pro was configured as a judge, but the live logs show no successful Pro response.

  • DigitalOcean

    Best Use of DigitalOcean

    Submitted

    We used Serverless Inference to plan searches and draft seller notices, Spaces to store evidence, and App Platform to host the live API and dashboard.

  • Tiger Data Q3 2026 - Q2 2027

    Best Use of Tiger Data

    Submitted

    We stored recalls, listings, verdicts, and time-stamped observations in TimescaleDB. Its BM25 search and vectorscale index support fast matching and image similarity.

  • Sentry

    Sentry Raffle

    Submitted

    We added tracing around backend pipeline stages and error capture to help diagnose failed checks and sponsor calls.

  • Bright Data

    Best Use of Bright Data

    Submitted

    We used its Web Scraper API to find marketplace listings, its SERP API to discover similar products, and Web Unlocker to re-check whether listings were still available. The image also lists OfferUp, but I could not verify that integration in the code.

  • HyperFrames

    Best Demo Video Using HyperFrames

    Submitted

    The image identifies HyperFrames as the demo-video tool. I could not find a rendered video or HyperFrames output in the repository, so claim it as used only if that video was actually made.

  • Event prize

    Best Overall Hack

    Submitted

    Recall Hunter connects the full path from an official product recall to a marketplace listing that may still be for sale. Recall Shield warns shoppers and shows the evidence, while the dashboard helps safety teams review listings, track sellers, and monitor removal.

  • Event prize

    Best Use of System One Models

    Submitted

    The project uses a fast first stage to screen product listings against an indexed recall database, then sends possible matches to Gemini for deeper review. This lets shoppers get timely warnings without running an expensive model on every product they scroll past.

  • Event prize

    Cleanest Use of a Software Factory

    Submitted

    We divided the build into clear components—the recall engine, marketplace collection, browser extension, backend, and console—with shared contracts between them. The repository’s FACTORY.md documents those responsibilities and the verification process, making the work easier to integrate and audit.