🤖 Finary AI - Core Engine

FastAPI + LangChain SQL Agent

This is the AI engine powering Finary. It manages the communication between the Gemini LLM, the PostgreSQL database, and the frontend clients.

🧠 AI Integration

• LLM: Gemini 2.5 Flash Lite.
• Agent Framework: LangChain SQL Agent with SQLDatabaseToolkit.
• Privacy Logic: Every AI query is strictly isolated using x-user-id headers to ensure multi-tenant data security.

🛣️ API Endpoints

• POST /api/v1/chat: Secure SQL Agent chat.
• GET /api/v1/proactive-insight: Gemini-generated spending analysis.
• POST /api/v1/scan-receipt: Vision-based receipt extraction.
• POST /api/v1/voice-entry: Audio-to-transaction processing.

🛠️ Tech Stack

• Framework: FastAPI (Python 3.13)
• AI: Google GenAI SDK + LangChain
• ORM: SQLAlchemy + Psycopg2
• Hosting: Render

⚙️ Environment Variables

Create a .env file:

SUPABASE_URL=your_url
SUPABASE_SERVICE_ROLE_KEY=your_key
GEMINI_API_KEY=your_google_ai_studio_key
DATABASE_URL=postgresql://postgres:[password]@[aws-1-ap-south-1.pooler.supabase.com:6543/postgres](https://aws-1-ap-south-1.pooler.supabase.com:6543/postgres)

# To run 
pip install -r requirements.txt
python main.py