Rasoi

India-first homemade cooking recommendations from the ingredients you already have. Deterministic recipe matching backed by PostgreSQL — not AI-generated recipes.

Stack

• Next.js 15, TypeScript, Tailwind CSS
• Supabase (PostgreSQL, free tier)
• Optional: Ollama + Qwen for explanation text only

Quick start

1. Supabase project

1. Create a free project at supabase.com.
2. Copy Project URL, anon key, and service role key.
3. In the SQL Editor, run the migration file: supabase/migrations/001_initial.sql

2. Environment

cp .env.example .env.local

Fill in your Supabase values in .env.local.

3. Install and seed

npm install
npm run db:seed

db:seed loads the curated seed catalog plus any generated catalog in data/generated/. This repo includes 102 hand-curated home recipes and can expand to 1600+ sourced Indian recipe entries after running npm run recipes:import.

4. Run locally

npm run dev

Open http://localhost:3000. Try: aloo, pyaz, tamatar.

5. Deploy (Vercel)

1. Push to GitHub and import in Vercel.
2. Add the same env vars (NEXT_PUBLIC_SUPABASE_*, SUPABASE_SERVICE_ROLE_KEY).
3. Keep AI_ENABLED=false on Vercel unless you host Ollama on a reachable server.

---

Optional: Ollama (AI explanations only)

The app works fully without Ollama. AI only adds short “why these recipes” text on the results page — it does not create or rank recipes.

Install Ollama (Windows)

1. Download from https://ollama.com/download and install.
2. Open PowerShell and pull a Qwen model:

ollama pull qwen2.5:7b

3. Verify Ollama is running:

ollama list
curl http://127.0.0.1:11434/api/tags

Enable in the app

In .env.local:

AI_ENABLED=true
OLLAMA_BASE_URL=http://127.0.0.1:11434
OLLAMA_MODEL=qwen2.5:7b
NEXT_PUBLIC_AI_ENABLED=true

Restart npm run dev. After searching ingredients, the results page may show an AI explanation block.

Note: Vercel cannot reach localhost. For production AI you would need Ollama on a VPS or a tunnel — not required for MVP.

---

Project structure

app/                 # Pages + API routes
components/          # UI
lib/intelligence/    # Normalize, match, score (deterministic)
lib/ai/              # Ollama client (optional)
lib/db/              # Supabase queries
data/seeds/          # Ingredients JSON + recipe catalog
supabase/migrations/ # SQL schema

Adding more recipes

Edit data/seeds/recipe-catalog.ts (use only canonical names from data/seeds/ingredients.json), then run npm run db:seed again.

For a larger sourced catalog, run:

npm run recipes:import

This downloads Anupam007/indian-recipe-dataset from Hugging Face and writes normalized recipe/ingredient data into data/generated/. The dataset page has no explicit license, so generated entries keep original source URLs and do not copy full method text into the app.

Scripts

Command	Description
npm run dev	Development server
npm run build	Production build
npm run db:seed	Load DB from seeds
npm run recipes:import	Generate expanded sourced catalog
npm run typecheck	TypeScript check