Agent Service Toolkit

Active
GitHub Python MIT

Description

Full toolkit for running an AI agent service built with LangGraph, FastAPI, and Streamlit, providing a complete reference architecture for agent service deployment.

Key Features

  • LangGraph-based agent with human-in-the-loop via interrupt(), long-term memory with Store, and flow control with Command
  • FastAPI service with both streaming and non-streaming endpoints for flexible agent serving
  • Streamlit chat interface with voice input/output for interactive agent communication
  • Multiple agent support — run and call different agents by URL path with model selection
  • RAG agent implementation using ChromaDB for document retrieval and question answering
  • Docker Compose setup with PostgreSQL, hot-reload, and full testing suite

Use Cases

💡 Deploying a production-ready AI agent service with API endpoints and chat UI
💡 Building a RAG assistant that answers questions from custom document collections
💡 Prototyping multi-agent workflows with human oversight and approval gates
💡 Creating a voice-enabled AI assistant with Streamlit's interactive interface
💡 Serving multiple specialized agents behind a single FastAPI gateway

Quick Start

Clone the repo, set OPENAI_API_KEY in .env, run 'uv sync --frozen' to install dependencies, then start with 'python src/run_service.py'. Launch the Streamlit UI in another shell with 'streamlit run src/streamlit_app.py'. Alternatively, use 'docker compose watch' for one-command setup.

Related Projects