Ingatini: A Personal Knowledge Search Engine

A lightweight Retrieval-Augmented Generation (RAG) application for uploading documents and asking AI-powered questions about them. Built with FastAPI, LangChain, PostgreSQL, and Gemini.

🎯 Core Idea

1. User uploads documents (PDF, DOCX, TXT)
2. System extracts text and creates embeddings
3. Embeddings stored in PostgreSQL with pgvector
4. User asks questions via chat interface
5. RAG pipeline retrieves relevant chunks and augments LLM response

📋 Project Status

Phase	Task	Status
1	Backend structure & DB schema	✅ Complete
2	Embedding pipeline & document processing	✅ Complete
3	RAG query engine & retrieval	✅ Complete
4	Frontend UI (React/Vite)	⏳ Next
5	Deployment & automation	📅 Post-launch

🚀 Quick Start

# Clone and setup
cd /home/fazrigading/Projects/ingatini

# Copy environment template
cp .env.example .env

# Start with Docker
docker compose up

# Or use the helper script
./dev start

API Documentation: http://localhost:8000/docs
Health Check: curl http://localhost:8000/api/health

See GETTING_STARTED.md for detailed setup instructions.

🏗️ Architecture

Backend

• Framework: FastAPI (Python)
• ORM: SQLAlchemy with PostgreSQL
• Vector Storage: pgvector (768-dim embeddings)
• RAG Pipeline: LangChain with Gemini
• API: RESTful with automatic docs

Database

• Users — User accounts
• Documents — Document metadata
• Chunks — Text segments with embeddings
• QueryLogs — Query history for analytics

Frontend (TODO)

• Framework: React/Vite
• UI: Clean, minimal interface
• Features: Document upload, chat with source attribution

📁 Project Structure

ingatini/
├── backend/              # FastAPI application
│   ├── app/
│   │   ├── core/        # Config & database
│   │   ├── api/         # Endpoints
│   │   ├── schemas/     # Pydantic models
│   │   ├── models/      # DB models
│   │   └── services/    # Business logic
│   ├── main.py
│   └── requirements.txt
├── frontend/            # React/Vite (TODO)
├── docker-compose.yml   # Dev environment
├── GETTING_STARTED.md   # Setup guide
├── IMPLEMENTATION.md    # Progress tracking
└── dev                  # Development helper

🔧 Development

Using Dev Helper

./dev start       # Start development
./dev logs        # View logs
./dev shell       # Access container
./dev test        # Run tests
./dev format      # Format code

API Endpoints

POST   /api/users/                  # Create user
GET    /api/users/{id}              # Get user
POST   /api/documents/upload        # Upload document
GET    /api/documents/{user_id}     # List documents
POST   /api/query/                  # Query documents (RAG)

📚 Stack

Layer	Technology
API	FastAPI 0.109
Database	PostgreSQL 15 + pgvector
ORM	SQLAlchemy 2.0
RAG	LangChain + Gemini
Embedding	embedding-001
LLM	gemini-pro
Frontend	React/Vite (TODO)
Containerization	Docker + Docker Compose

⚙️ Configuration

Create .env from .env.example:

DATABASE_URL=postgresql://user:password@localhost:5432/ingatini_db
GEMINI_API_KEY=your_key_here
GEMINI_EMBEDDING_MODEL=models/embedding-001
GEMINI_LLM_MODEL=gemini-pro
DEBUG=True

📖 Documentation

• GETTING_STARTED.md — Setup & usage guide
• IMPLEMENTATION.md — Progress & architecture
• backend/README.md — Backend development
• frontend/README.md — Frontend development

🛠️ Automation (Post-Development)

After MVP completion, integrate with n8n:

• Trigger: New file uploaded
• Action: Call embedding pipeline API
• Result: Auto-insert embeddings to database

📝 License

MIT

---

Next: See GETTING_STARTED.md to begin developing!