RAG-powered chatbot for analyzing financial documents with LangChain tool calling.
- Frontend: Next.js 14 + Tailwind + shadcn/ui
- Backend: FastAPI + LangChain + LangGraph
- Database: PostgreSQL + pgvector
- Deploy: Vercel + Supabase
- Upload CSV/PDF financial documents
- Natural language queries about spending
- Tool calling (search, analyze, compare)
- Session memory for multi-turn conversations
# Backend
cd backend
pip install -r requirements.txt
uvicorn main:app --reload
# Frontend
cd frontend
npm install
npm run dev"How much did I spend on restaurants this month?"
"What are my top 5 expense categories?"
"Compare July vs August spending"
"Show transactions over $1,000"
See PROJECT_SPECS.md for detailed specs.
This project represents work built for a client under NDA. The architecture and patterns are identical - only the domain and data have been changed.