DocuMind AI

# 1. Clone the repository

git clone https://github.com/yourusername/documind-ai.git

cd documind-ai

# 2. Install dependencies

npm install

# 3. Set up environment variables

cp .env.example .env

# Edit .env with your actual values

# Database

DATABASE_URL="postgresql://..."   # Supabase connection URL

DIRECT_URL="postgresql://..."    # Direct connection (migrations)

# AI Services

GROQ_API_KEY="gsk_..."       # groq.com - free tier

HUGGINGFACEHUB_API_KEY="hf_..."   # huggingface.co - free tier

# Auth

JWT_SECRET="min-32-chars-secret"  # openssl rand -base64 64

# Redis

UPSTASH_REDIS_REST_URL="https://..."

UPSTASH_REDIS_REST_TOKEN="..."

# App Config

NEXT_PUBLIC_APP_URL="http://localhost:3000"

ADMIN_EMAIL="admin@yourdomain.com"

FREE_TIER_MONTHLY_TOKEN_LIMIT=50000

MAX_FILE_SIZE_BYTES=10485760

VECTOR_STORE_TABLE_NAME="document_embeddings"

RETRIEVER_TOP_K=4

# 4. Set up database

# Run prisma/setup-pgvector.sql in your Supabase SQL editor first

npx prisma generate

npx prisma db push

# 5. Start development server

npm run dev

Features

User Features

• 📧 Email + password authentication (JWT, httpOnly cookies)
• 📄 PDF upload with drag-and-drop interface
• 🤖 AI-powered chat with streaming responses
• 📍 Citation tracking (page numbers, relevance scores)
• 💬 Chat history per document
• 📊 Token usage dashboard

Technical Features

• 🏗️ RAG Pipeline: LangChain → HuggingFace → pgvector → Groq
• 🔐 Multi-tenant isolation: userId filtering at every layer
• ⚡ Rate limiting: Redis sliding window (5 req/min)
• 🛡️ Prompt injection prevention: Input sanitization
• 📈 Admin dashboard: Platform-wide analytics with Recharts
• 🧪 Evaluation script: Precision@K, Groundedness, Relevance

Environment Requirements

ServiceFree Tier LimitsSupabase500MB DB, 1GB bandwidth/monthVercel100GB bandwidth, 100hr compute/monthGroq14,400 req/day, 30 req/minHuggingFaceRate limited, ~1 req/secUpstash Redis10,000 commands/day, 256MB