Problem
The LangGraph multi-agent chat proposal is useful, but it should not be a second full
FastAPI starter. It is an orchestration capability layered onto a FastAPI app.
It also overlaps with the simple chat extension if both define the same route shape.
Solution
Create extensions/fastapi-langgraph-chat compatible with fastapi-backend.
Scope:
- Add a small LangGraph state graph with typed state.
- Include simple nodes such as router, responder, and tool executor.
- Expose a route that does not conflict with
fastapi-ai-chat unless explicitly
intended; if it conflicts, define symmetric incompatibleWith.
- Include test-only fake LLM/tool implementations.
- Support native tools through typed functions.
- Leave MCP, RAG, and MLflow as optional follow-up extensions or documented
integration points.
- Add docs:
template/docs/LANGGRAPH_CHAT_GUIDE.md
template/docs/README.md.append
.env.example.append
- Add partial
pyproject.toml dependencies.
Constraints:
- No external LLM calls in tests.
- No frontend in the first implementation.
- No
.github/workflows in this extension.
- Keep the graph small and readable; this should be a baseline, not a full agent
framework.
Acceptance criteria:
- Generated tests verify graph routing and final response construction with fake
providers.
- The guide explains how to add a new node/tool safely.
Problem
The LangGraph multi-agent chat proposal is useful, but it should not be a second full
FastAPI starter. It is an orchestration capability layered onto a FastAPI app.
It also overlaps with the simple chat extension if both define the same route shape.
Solution
Create
extensions/fastapi-langgraph-chatcompatible withfastapi-backend.Scope:
fastapi-ai-chatunless explicitlyintended; if it conflicts, define symmetric
incompatibleWith.integration points.
template/docs/LANGGRAPH_CHAT_GUIDE.mdtemplate/docs/README.md.append.env.example.appendpyproject.tomldependencies.Constraints:
.github/workflowsin this extension.framework.
Acceptance criteria:
providers.