Bank Statement Analyzer

Full-stack app to upload bank statements, extract transactions, categorize spending with local LLMs, and explore financial insights through dashboards and chat.

Project Structure

• frontend/ - Next.js app for authentication, uploads, dashboards, jobs, transaction management, and chat UI.
• backend/ - FastAPI API for auth, statement processing, background jobs, analytics, embeddings, and storage.

Detailed docs:

• Frontend: frontend/README.md
• Backend: backend/README.md

Architecture (High Level)

• frontend sends authenticated requests to backend.
• backend stores app data in PostgreSQL.
• File assets (uploaded PDFs and markdown artifacts) are stored in MinIO (S3-compatible).
• Background processing runs through RQ + Redis.
• Embeddings are stored/retrieved from ChromaDB.
• Ollama powers extraction/categorization and chat responses.

Quick Start

1. Start infrastructure services:

   docker compose -f docker-compose.infra.yml up

2. Run backend in one terminal:
   • Follow backend/README.md for venv, migrations, API server, and worker.
3. Run frontend in another terminal:
   • Follow frontend/README.md for install and dev server.

Full Docker Setup

To run the full stack via Docker (including app services), use:

docker compose up

Environment

• Root .env.example contains sample environment variables.
• Copy to .env and adjust values for your machine/services.

Notes

• Backend default local URL: http://localhost:8000
• Frontend default local URL: http://localhost:3000
• Frontend expects backend running locally unless configured otherwise.