NyayaMitra

NyayaMitra is a legal-assistant project with:

• Backend: FastAPI + RAG (ChromaDB)
• Frontend: React + Vite

This guide helps other developers run the project locally after cloning from GitHub.

Prerequisites

• Python 3.11+ (tested with 3.12)
• Node.js 18+ and npm
• macOS/Linux shell commands below (Windows users can adapt to PowerShell)

1) Clone and enter project

git clone <your-repo-url>
cd nyayamitra

2) Backend setup

Create and activate a Python virtual environment, then install dependencies:

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

3) Environment variables

Create your .env file from the example:

cp .env.example .env

Then edit .env and add at least one provider key (recommended: GROQ_API_KEY or GEMINI_API_KEY).

If no key is set, the app can still run in context-only mode.

4) Vector DB options (Chroma)

This repo can include a refined prebuilt chroma_db/ so new users can run quickly.

Option A: Use the included/refined chroma_db/ (fast start)

No ingestion step needed. Just run backend + frontend.

Option B: Rebuild your own vector DB from scratch

From project root:

rm -rf chroma_db
source .venv/bin/activate
python scripts/prepare_rag_chunks.py --include-qa --include-criminal-primer --include-legal-vector-10k
python scripts/ingest_chroma.py

This recreates the local vector store in chroma_db/.

5) Auth (SQLite + JWT)

User accounts are stored in legal_advisor.db at the project root (nyayamitra/legal_advisor.db), not inside backend/. Open it with DB Browser for SQLite or run python scripts/inspect_users_db.py — the file is binary, so the IDE text view will not show rows. See DATABASE_URL and JWT_SECRET in .env.

• Sign up / log in in the UI before Chat or Action plan.
• API: POST /api/auth/signup, POST /api/auth/login, GET /api/auth/me (Bearer token).
• Protected routes: /api/ask, /api/chat, /api/plan require Authorization: Bearer <token>.

On Render, set a strong JWT_SECRET; SQLite data is ephemeral on free tier unless you add persistent disk.

6) Run backend

From project root:

source .venv/bin/activate
uvicorn backend.main:app --reload --host 127.0.0.1 --port 8000

Backend URLs:

• API base: http://127.0.0.1:8000
• Health: http://127.0.0.1:8000/health

7) Run frontend

Open a second terminal:

cd frontend
npm install
npm run dev

Frontend URL:

• http://localhost:5173

Vite is already configured to proxy /api and /health to backend 127.0.0.1:8000.

8) Quick smoke test (optional)

source .venv/bin/activate
python scripts/smoke_api.py

Important: include datasets and vector DB in GitHub

These files are project data and should be committed so others (and hosts like Render) have everything they need:

• ai_legal_phase4_diverse.json
• legal_vector_dataset_10k.json
• data/raw/*.jsonl (constitution + QA + criminal primer shards for prepare_rag_chunks.py)
• data/chunks/rag_chunks.jsonl (optional but speeds rebuilds; otherwise regenerate with the script)
• chroma_db/ (refined vector DB; includes Chroma’s *.sqlite3 files — .gitignore is set so those are not excluded)

You can add and commit them with:

git add README.md ai_legal_phase4_diverse.json legal_vector_dataset_10k.json data/raw chroma_db
# optional (skip if you will regenerate chunks on the server):
git add data/chunks/rag_chunks.jsonl
git commit -m "Add setup README and include required datasets and chroma DB"
git push

Production deploy (Render + Vercel)

Deploy the FastAPI backend on Render and the Vite React frontend on Vercel. The UI calls the API using VITE_API_BASE_URL (see frontend/src/api.js); locally that variable is unset so /api and /health still go through the Vite dev proxy.

Backend (Render)

1. Push this repo to GitHub (include chroma_db/ if you want RAG without running ingest on the server; see note above).
2. In Render: New → Blueprint (connect the repo and select render.yaml) or Web Service with:
   • Root directory: repository root (where requirements.txt lives).
   • Build command: pip install -r requirements.txt
   • Start command: uvicorn backend.main:app --host 0.0.0.0 --port $PORT
   • Python version: 3.11+ (this repo includes runtime.txt for Render’s native Python runtime).
3. Environment variables on the service (minimum for a working UI against prod API):
   • CORS_ORIGINS — comma-separated list including your Vercel production URL (e.g. https://your-app.vercel.app). Add http://localhost:5173 if you want a local frontend to talk to the hosted API.
   • GROQ_API_KEY / GEMINI_API_KEY / OPENAI_API_KEY — optional; without them the API stays in context-only mode (see .env.example).
4. After deploy, copy the service URL (e.g. https://nyayamitra-api.onrender.com). Check https://<that-host>/health in a browser.

Blueprint reference: render.yaml (secret keys use sync: false so Render prompts you when applying the blueprint).

Free tier: Services may spin down after idle time (cold starts). Chroma + sentence-transformers are memory-heavy; if the service crashes on startup, try a paid instance with more RAM or a smaller index.

Frontend (Vercel)

1. New Project → import the same GitHub repo.
2. Set Root Directory to frontend (Vite app; vercel.json sets build/output).
3. Under Environment Variables, add VITE_API_BASE_URL = your Render URL without a trailing slash (e.g. https://nyayamitra-api.onrender.com). Redeploy after changing it.
4. Deploy. Open the Vercel URL; the header should load /health from Render and show chunk count when chroma_db is present on the server.

See frontend/.env.example for a short reminder.

Verify

• GET https://<render-host>/health returns JSON with ok and chroma_documents.
• Vercel app loads without CORS errors in the browser console; chat or plan returns a response.

Common issues

• Backend not starting: check you used --host 127.0.0.1 --port 8000 (host and port order matters).
• Frontend shows proxy errors: backend is not running on 127.0.0.1:8000.
• Missing API keys: add provider keys in .env or use context-only mode.
• Production CORS errors: set CORS_ORIGINS on Render to your exact Vercel https://… origin (comma-separated if multiple).
• Production UI cannot reach API: set VITE_API_BASE_URL on Vercel to the Render URL and redeploy the frontend.