[Feature Request] Async Support for FastAPI Routers and Adapter Invocations

[ilkersigirci]

Description
First, thank you for this great library! While examining the code, I noticed that the FastAPI routers are not using async endpoints, and the adapter.invoke method runs synchronously for all underlying LangGraph operations. This appears to be a potential bottleneck for multi-user serving scenarios, as the server processes requests sequentially.

Current Behavior

• FastAPI routes are synchronous
• adapter.invoke blocks the event loop
• Server processes requests one at a time

Expected Behavior

• FastAPI routes should support async operations
• Adapter should support ainvoke for asynchronous execution
• Server should be able to handle multiple concurrent requests efficiently

Proposed Solution

1. Convert FastAPI routers to use async endpoints
2. Implement and use await adapter.ainvoke for LangGraph operations
3. Maintain backward compatibility with sync operations

Benefits

• Improved scalability for multi-user scenarios
• Better utilization of FastAPI's async capabilities
• Non-blocking request processing

[samuelint]

Hi @ilkersigirci thanks for your feedback!

I invite you to contribute to the project by making a pull request.
Make sure to have tests for every features you implement (I mostly look at that).

[uesleilima]

@samuelint @ilkersigirci I opened #48 with a possible solution for that, a review would be appreciated, so far it looks good in my tests with langraph api server using custom routes

[ilkersigirci]

@uesleilima It seems a major refactor, good job. Unfortunately, I am not using this library anymore since I have built one that works for me, langgraph-openai-serve. Hence, I don't think I am not suitable for reviewing this major refactor