FishID

AI-powered fish species identification — upload a photo, get ranked species matches with AI-generated facts, and keep a personal fish log. Built for divers, snorkelers, and marine enthusiasts.

Live app: fishid.vercel.app

How it works

flowchart LR
  B[Browser] -->|resize + EXIF strip| I["/api/identify"]
  I -->|top-3 candidates| F[Fishial AI]
  B -->|stream| E["/api/enrich"]
  E --> G[Groq llama-3.3-70b]
  E --> W[Wikipedia]
  E --> C[(enrichment_cache)]
  B -->|supabase-js + RLS| S[(Supabase\nAuth · Postgres · Storage)]

Loading

• One Next.js app (App Router) on Vercel — no separate backend.
• Supabase owns auth (email + Google OAuth, cookie sessions), Postgres (species catalog, fish log, rate limiting, AI cache), and Storage (log photos, private bucket).
• Two server routes are all that's needed: POST /api/identify (session-gated, rate-limited proxy to the Fishial recognition API, returns ranked candidates) and GET /api/enrich/[species] (+/stream: one cached Groq completion streamed to the client, with Wikipedia as the factual backbone).
• Photos are resized and EXIF-stripped in the browser before they go anywhere — GPS metadata never leaves the device.
• The fish log is plain supabase-js under row-level security: every table and the storage bucket are owner-scoped.

Features

• 📷 Photo upload with mobile camera capture
• 🐠 Ranked top-3 species matches with confidence bars (fish ID is genuinely ambiguous — the UI says so)
• ✨ AI-generated description, visual ID cues, and a fun fact, streamed in live and cached per species
• 📒 Personal fish log with photos, candidates, and dates
• 🔍 Searchable species catalog with habitat/region/conservation filters
• 🔐 Supabase Auth: email + password (with confirmation) and one-tap Google sign-in

Development

pnpm install
pnpm dev          # localhost:3000
pnpm typecheck && pnpm lint && pnpm test && pnpm build

.env.local:

NEXT_PUBLIC_SUPABASE_URL=...
NEXT_PUBLIC_SUPABASE_ANON_KEY=...
FISHIAL_CLIENT_ID=...
FISHIAL_SECRET=...
GROQ_API_KEY=...
SUPABASE_SERVICE_ROLE_KEY=...   # server-only

Database schema and RLS policies live in supabase/migrations/; the species seed is in supabase/seed/. CI (GitHub Actions) runs typecheck, lint, unit tests, and build on every push.

Attribution

• Fishial — fish recognition API
• Groq — LLM inference
• Wikipedia — species summaries and reference images
• Fish icons by Freepik (see docs/fish-icons-license.html)

License

MIT — © Hrishi Kabra (@hrishikabra · LinkedIn · kabrahrishi@gmail.com)