fix: Watch startup errors for stdio client #289
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The character encoding of the stdin/stdout streams in Python is platform- dependent. On Windows it will be something weird, like CP437 or CP1252, depending on the locale. This change ensures that no matter the platform, UTF-8 is used.
…col#218) Adds sampling and list roots callbacks to the ClientSession, allowing the client to handle requests from the server. Co-authored-by: TerminalMan <[email protected]> Co-authored-by: David Soria Parra <[email protected]>
…/jerome/fix/request-context-typing Updated typing on request context for the server to use server session
…ontextprotocol#222) * feat: allow lowlevel servers to return a list of resources The resource/read message in MCP allows of multiple resources to be returned. However, in the SDK we do not allow this. This change is such that we allow returning multiple resource in the lowlevel API if needed. However in FastMCP we stick to one, since a FastMCP resource defines the mime_type in the decorator and hence a resource cannot dynamically return different mime_typed resources. It also is just the better default to only return one resource. However in the lowlevel API we will allow this. Strictly speaking this is not a BC break since the new return value is additive, but if people subclassed server, it will break them. * feat: lower the type requriements for call_tool to Iterable
|
We encountered this bad behavior while writing a synchronous MCP client for Letta. The uncatchable error stalls our server event loop. cc @dsp-ant |
|
cc @Kludex @allenporter for suggestions here |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
Improves
stdio_clienterror handling to prevent indefinite hangs when subprocess failures occur, particularly when wrapping MCP servers with a synchronous interface. Adds structured error propagation to make failures catchable instead of failing silently insideanyio.This is a minimal fix, but more long term, it would be great if either:
Motivation and Context
When integrating MCP servers with a sync interface, failures in
anyio.open_process()or unexpected subprocess exits could result in the process hanging indefinitely instead of raising an exception. Previously, errors insideanyio's process handling were not properly surfaced, making debugging difficult.This PR ensures:
RuntimeErrorinstead of hanging.startup_wait_time) allows for quick failure detection instead of silent hangs.This is particularly useful when MCP servers are used in synchronous mode, where an uncatchable error would otherwise stall execution.
How Has This Been Tested?
RuntimeErrorimmediately.pytestanduv run pytest.Breaking Changes
RuntimeErrorexceptions.Note that the field defaults to 0.0, so should not have any effect unless explicitly set.
Types of changes
Checklist
Additional context
startup_wait_time(default0.0) to allow detection of immediate failures without forcing all cases to wait.ExceptionGroupissues by ensuring a singleRuntimeErroris raised when subprocess failures occur.