Straydar

A map-based tool for tracking, reporting, and getting help for stray and lost cats in a neighborhood — built as a React + Supabase full-stack app.

What it does

Drop a pin, report a cat, and it's shared live with everyone else using the app — not just saved to your own browser. Straydar handles the parts that make community pet-tracking actually useful in practice:

• Duplicate detection — before creating a new cat record, it checks sightings within 150m and scores them against the new report's temperament and description, so the same colony cat doesn't get re-added every time someone spots it (src/services/matching.js).
• Live shared state — cats and sightings live in Postgres via Supabase and sync in real time across everyone's browser (src/context/DataContext.jsx).
• Lost-cat reports with owner contact info and a dedicated "missing" flow, separate from general stray sightings.
• An AI assistant for on-the-spot rescue guidance (TNR, approaching a scared cat, kitten feeding, wound triage) that works with or without an API key — canned expert guidance by default, live Claude responses if you add your own Anthropic key.
• Nearest emergency vets, pulled live from OpenStreetMap with a static fallback list.
• A /demo walkthrough — an isolated version of the app (/demo, /demo/feed, /demo/missing, /demo/emergency) seeded with local demo cats ringed around your real location, so you can try the reunion flow without touching shared production data (src/hooks/useDemoMode.js, src/services/demoData.js).

AI rescue assistant	Missing-cat report

Stack

React 19, Vite, React Router 7, Tailwind CSS v4, Leaflet/react-leaflet for the map, Supabase (Postgres + Realtime + Auth-ready), the Anthropic SDK for the assistant, Vitest for tests, oxlint for linting.

See DECISIONS.md for the reasoning behind the less obvious choices (why Supabase, why RLS is permissive for now, why the duplicate matcher is a hand-rolled score instead of ML, why the AI assistant runs client-side).

Running it locally

npm install
cp .env.example .env   # fill in your Supabase project URL + anon key
npm run dev

Full backend setup (creating the Supabase project, running the migrations) is in CLAUDE.md.

npm run lint    # oxlint
npm test        # vitest
npm run build   # production build

Project structure

• src/context/DataContext.jsx — the single source of truth for cats/ sightings; every component reads and writes through useData().
• src/services/ — persistence (db.js, supabaseClient.js), the duplicate-matching algorithm (matching.js), geo math (geo.js), and the AI assistant (aiAssistant.js).
• src/pages/ — MapPage, FeedPage, MissingPage, EmergencyPage.
• supabase/migrations/ — the Postgres schema and seed data.

CLAUDE.md has the full architecture writeup, including the trickiest cross-file flow (the duplicate-cat detection pipeline) if you want to dig in further.