joshuac215/agent-service-toolkit ? reverse-engineered prompt

Build me a complete AI agent toolkit that I can run locally and extend later.

I want a Python app with an agent backend, an API service, and a simple chat UI. The backend should use LangGraph, expose FastAPI endpoints for normal and streaming chat, and let me keep separate agents that can be called by different URL paths. The UI should be a Streamlit chat app with conversation history and a clean way to switch between agents. If it makes sense, add support for voice input and output too.

Please include a sensible project structure, config via environment variables, and a good default setup that works with either an API key based model or a local model if I choose one. If useful, add a basic example agent and a simple retrieval based agent so I can see how to customize things. Also make it easy to run with Docker or directly in Python, and include tests and a README that explains how to get started.