Note Agent (FastAPI + LangGraph + OpenAI + Supabase)

A simple “note-taking agent” web app that uses an LLM (via LangGraph) to understand natural-language instructions and save/read notes.
Deployed on Vercel with a Supabase Postgres database for persistent storage.

Live Demo

• Vercel URL: https://agentd-phi.vercel.app/

Features

• Web UI (FastAPI + Jinja2 template)
• Agent powered by LangGraph ReAct + OpenAI (ChatOpenAI)
• Save notes to a Supabase table (notes) with fields:
  • filename (text)
  • content (text)
  • created_at (timestamp)

Tech Stack

• FastAPI
• LangGraph + LangChain
• OpenAI API (gpt-4o-mini)
• Supabase (Postgres)
• Vercel (serverless deployment)

How it works (high level)

1. User enters a request (e.g., “save a note: human biology”).
2. Backend calls the agent.
3. Agent decides which tool to use (write/read).
4. Note is stored in Supabase for persistence.

Local Setup

1) Clone and install

pip install -r requirements.txt
### Environment Variables (required)
Set these in your environment (Vercel: Project → Settings → Environment Variables).

- OPENAI_API_KEY
- SUPABASE_URL
- SUPABASE_ANON_KEY
# (or) use a Postgres connection string if your backend uses that:
- SUPABASE_POSTGRES_URL