FIX: gemini tool call hanging #3898
Merged
Conversation
This file contains hidden or 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
jamadeo
approved these changes
Aug 6, 2025
Collaborator
jamadeo
left a comment
jamadeo
left a comment
There was a problem hiding this comment.
Good find! This is a pretty nasty loop and could use some refactoring probably. But before taking that on, we should get some streaming-compatible scenario tests -- then we could add this as a case to test for
alexhancock
approved these changes
Aug 7, 2025
michaelneale
added a commit
that referenced
this pull request
Aug 8, 2025
* main: chore: rmcp 0.5.0 (#3935) add gpt-5 to openai provider format (#3924) added gpt5 context limit (#3927) show status of osx codesigning and increase timeout (#3926) Bump auto-compact threshold to 80% (#3925) FIX: gemini tool call hanging (#3898) feat(deps): upgrade rmcp to 0.4.1 (#3918) Fix dark mode rendering of config form and centered providers grid for wider screens. (#3837) fix: extension list not refreshing after installing from deeplink (#3878)
katzdave
added a commit
that referenced
this pull request
Aug 8, 2025
* 'main' of github.com:block/goose: remove fallback routing to hub/home for unknown routes (#3954) Use cross in linux bundle workflow (#3950) fix: disable signing for release branches until we figure out keys for this flow (#3951) Sanitize Tags Unicode Block (#3920) Add a message about DCO to CONTRIBUTING.md (#3741) Move hardcoded LLM prompts to template files (#3934) docs: migrate streamable config to consolidated component (#3936) feat: streamline list args on cli (#3937) mcp/developer: Refactor to use tokio SplitStream (#3894) feat: first time automated ollama install experience and openrouter (#3881) chore: rmcp 0.5.0 (#3935) add gpt-5 to openai provider format (#3924) added gpt5 context limit (#3927) show status of osx codesigning and increase timeout (#3926) Bump auto-compact threshold to 80% (#3925) FIX: gemini tool call hanging (#3898) feat(deps): upgrade rmcp to 0.4.1 (#3918) Fix dark mode rendering of config form and centered providers grid for wider screens. (#3837) fix: extension list not refreshing after installing from deeplink (#3878)
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.
Issue
The response_to_streaming_message function in the OpenAI format provider was not returning tool call messages to the stream when AI providers send complete tool calls in a single chunk with finish_reason: "tool_calls".
Root Cause: The original logic assumed all tool calls would be streamed across multiple chunks (like OpenAI's behavior) and always entered a loop to wait for additional tool call data. When providers like Gemini send complete tool calls in one chunk, the function would:
4.Break out of the entire function without yielding the tool call message
This resulted in tool calls being processed but never returned to the caller, causing streaming responses to appear empty.
Fix
Added a check to detect when tool calls are already complete in the initial chunk