Document Analyzer RAG

Production-ready chat-first RAG application for document Q&A.

Overview

Document Analyzer lets users:

1. Sign in with Firebase social auth (Google or GitHub)
2. Upload PDF documents
3. Wait for indexing to complete
4. Chat with grounded, document-based answers

Frontend and backend are deployed separately:

• Frontend: https://docanalyzer.app
• Backend API: https://api.docanalyzer.app

Core Features

• Firebase social login (Google + GitHub)
• Backend-owned JWT cookie sessions
• Chat-first workflow with streaming responses (SSE)
• Document upload, processing, indexing, and chat attachment flow
• RAG pipeline with retrieval safeguards and graceful fallback behavior
• Daily usage quota and infrastructure limits endpoint
• Strict CORS policy and layered rate limiting

Tech Stack

• Frontend: Next.js, React, TypeScript, Tailwind CSS, Framer Motion, Firebase Web SDK
• Backend: Node.js, Express, Firebase Admin SDK
• Database: PostgreSQL (Supabase)
• Storage: Supabase Storage
• AI: Google Gemini
• Deploy: Vercel (frontend), Render (backend)

Repo Structure

.
├── frontend/              # Next.js app
├── src/                   # Express API
├── database/schema.sql    # Database schema bootstrap
├── render.yaml            # Render blueprint
├── BACKEND_DOCS.md        # API contract documentation
└── ARCHITECTURE.md        # System architecture and flows

Screenshots

Architecture visuals currently included:

• System architecture: docs/architecture.svg
• RAG flow: docs/rag-architecture.svg

Authentication Model (Final)

• Only Firebase social providers are supported:
  • Google
  • GitHub
• Backend auth exchange endpoint remains:
  • POST /api/v1/auth/google
• Legacy email/password/OTP auth is removed from active flows.

Local Setup

1) Install dependencies

npm install
cd frontend && npm install

2) Configure Firebase

In Firebase Console:

1. Create/select project
2. Enable Auth providers:
   • Google
   • GitHub
3. Create a Web App and copy client config values
4. Create a Service Account for backend and copy:
   • project ID
   • client email
   • private key

3) Configure environment variables

Backend:

cp .env.example .env

Frontend:

cd frontend
cp .env.example .env.local

Backend required env vars

• DATABASE_URL
• SUPABASE_URL
• SUPABASE_SERVICE_KEY
• GEMINI_API_KEY
• JWT_SECRET
• FIREBASE_PROJECT_ID
• FIREBASE_CLIENT_EMAIL
• FIREBASE_PRIVATE_KEY

Frontend required env vars

• NEXT_PUBLIC_API_URL
• NEXT_PUBLIC_FIREBASE_API_KEY
• NEXT_PUBLIC_FIREBASE_AUTH_DOMAIN
• NEXT_PUBLIC_FIREBASE_PROJECT_ID
• NEXT_PUBLIC_FIREBASE_APP_ID

4) Initialize database (fresh/disposable DB only)

npm run db:schema

5) Run locally

Backend:

npm run dev

Frontend:

cd frontend
npm run dev

Production Deployment

Backend on Render

1. Connect repository to Render
2. Use blueprint file: render.yaml
3. Set sync: false secrets in Render dashboard
4. Confirm health endpoint:
   • GET https://api.docanalyzer.app/api/v1/health/live

Frontend on Vercel

1. Import frontend/ as project root
2. Set production env vars (NEXT_PUBLIC_*)
3. Set NEXT_PUBLIC_API_URL=https://api.docanalyzer.app
4. Deploy and validate auth + chat flows

Security & Stability Checks

• Strict CORS allowlist for docanalyzer.app domains
• Global rate limit + route-specific auth/chat/upload limiters
• Standard API response envelope:
  • success: { ok: true, data: ... }
  • error: { ok: false, error: { code, message, details? } }
• Firebase Admin env validation at boot
• No JSON service account file dependency in backend runtime

Verification Commands

npm test
npm run frontend:test
npm run frontend:lint
npm run frontend:build

Additional Documentation

• Backend contract: BACKEND_DOCS.md
• System architecture: ARCHITECTURE.md
• Deployment checklist: DEPLOYMENT_CHECKLIST.md