Fix: memory leak and exit problems

[PWZER]

Motivation and Context

• Here handle_sse calls app.run() every time it processes a sse request, but app.run() keeps running and does not stop automatically, which causes handle_sse to never end, and finally causes a memory leak.
  

  python-sdk/examples/servers/simple-tool/mcp_simple_tool/server.py

  Lines 67 to 73 in b4c7db6

  	async def handle_sse(request):
  	async with sse.connect_sse(
  	request.scope, request.receive, request._send
  	) as streams:
  	await app.run(
  	streams[0], streams[1], app.create_initialization_options()
  	)

• And this is also the key reason for the following exit problems issues#541, Fixed.

  INFO:     Waiting for background tasks to complete. (CTRL+C to force quit)

Types of changes

• Bug fix (non-breaking change which fixes an issue)

Checklist

• I have read the MCP Documentation
• My code follows the repository's style guidelines
• New and existing tests pass locally
• I have added appropriate error handling
• I have added or updated documentation as needed

[dsp-ant]

Is this needed?

[PWZER]

Yes, sse-starlette supports client_close_handler_callable since version 2.3.0.

[nadeemcite]

This looks great, Thanks @PWZER

[ihrpr]

Unfortunately we can't upgrade sse-starlette in SDK now.

We can achieve similar functionality as client_close_handler_callable provides in this case by using disconnect_listener

Something like this should work:

 async with anyio.create_task_group() as tg:
            response = EventSourceResponse(
                content=sse_stream_reader,
                data_sender_callable=sse_writer,
            )

            async def disconnect_listener():
                try:
                    await response.listen_for_disconnect(receive)
                    logger.error(f"Disconnect detected for session: {session_id}")
                    await client_close_handler()  # defined earlier in the PR - name should be changed probably
                except Exception as e:
                    logger.error(f"Error in disconnect listener: {e}")

            tg.start_soon(disconnect_listener)
            logger.debug("Starting SSE response task")
            tg.start_soon(response, scope, receive, send)

            ...
            yield (read_stream, write_stream)

[ihrpr]

Suggested change

	del self._read_stream_writers[session_id]
	self._read_stream_writers.pop(session_id, None)