JoshuaC215/agent-service-toolkit

Full toolkit for running an AI agent service built with LangGraph, FastAPI and Streamlit

What it solves

This toolkit provides a complete, production-ready template for building and deploying AI agents. It eliminates the need to manually set up the boilerplate for the backend service, the API layer, and the user interface, allowing developers to focus on defining the agent's logic and tools.

How it works

The project implements a full-stack architecture consisting of four main components:

  1. Agent Logic: Built using the LangGraph framework, supporting advanced features like human-in-the-loop interrupts, long-term memory (Store), and supervisor patterns.
  2. Backend Service: A FastAPI server that exposes the agent via streaming and non-streaming endpoints, including support for the AG-UI protocol for compatibility with frontends like CopilotKit.
  3. Client Library: A Python AgentClient that allows other applications to interact with the agent service programmatically.
  4. Frontend: A Streamlit application that provides a chat interface with voice input and output capabilities.

Who it’s for

Developers who want to quickly prototype or deploy LangGraph-based AI agents with a professional API and a functional web interface without building the infrastructure from scratch.

Highlights

  • Full-Stack Template: Includes everything from agent definition to a Streamlit UI and Docker deployment.
  • Advanced LangGraph Integration: Implements interrupt(), Command, and langgraph-supervisor.
  • Flexible Streaming: Supports both token-based and message-based streaming.
  • Multi-Agent Support: Ability to host and call multiple different agents via unique URL paths.
  • Built-in RAG: Includes a basic RAG implementation using ChromaDB.
  • Enterprise-Ready Tooling: Integrated content moderation (via Safeguard), feedback systems (via LangSmith), and robust unit/integration testing.