AkhileshMalthi/multi-agent-system-using-langgraph — reverse-engineered prompt

Build me a production ready Python backend for a general purpose AI assistant where multiple agents work together on a user request. A user should be able to send one plain English prompt, like comparing two tools or asking for a beginner tutorial, then the system should create a task, have a research agent gather the important points, have a writing agent turn that into a clear draft, and let the user check the status and approve the final result.

Use FastAPI for the API, LangGraph and LangChain for the agent workflow, Celery for background processing, Redis for state and queueing, and PostgreSQL for saving tasks and results. Include Docker Compose so I can start the whole thing locally with an API key for OpenAI or Groq in an env file.

Please add simple endpoints to create a task, get task status, and approve or reject a draft. Include clean project structure, basic tests, and a short README with setup and curl examples. Look up current docs online if you need to.