use httpx.AsyncClient for async http requests #34
Conversation
| @@ -768,7 +790,7 @@ uvicorn.run(app, host="0.0.0.0", port=8000) | |||
| </Step> | |||
| <Step title="Run tests"> | |||
| ```bash | |||
| uv add --dev pytest pytest-asyncio | |||
There was a problem hiding this comment.
httpx and starlette both depend on anyio which ships with an async test framework - so you don't need to install an extra one here
There was a problem hiding this comment.
The change to httpx looks great—thank you!
Re: anyio, although we use it in our SDK, it may make examples or getting started a bit more confusing than using asyncio, because the latter is part of the standard library. We've purposely opted to stick with asyncio for now. Do you mind reverting those bits? 🙏
| @@ -108,20 +108,20 @@ Let's build your first MCP server in Python! We'll create a weather server that | |||
| Add this functionality: | |||
|
|
|||
| ```python | |||
| # Create reusable session | |||
| http = requests.Session() | |||
There was a problem hiding this comment.
There's nowhere to run a context manager and expose it to request handlers, so we're forced to create a new client each request: modelcontextprotocol/python-sdk#67
|
Happy to drop the use of anyio.run, but having multiple async plugins for pytest installed is overkill |
To stay consistent across the example, the implementation should use one async runtime for now. I would think it's best if for now we keep asyncio and create an issue to follow up if we should just switch the whole example, including the implementation itself, to anyio given that it is transient a dependency anyway. |
Yes I configured it to use asyncio only |
requests.Session does blocking IO which blocks the event loop! No other requests can be processed concurrently