fix: GoogleGenaiSamplingHandler leaks thought parts and gives unhelpful errors on empty responses by strawgate · Pull Request #3849 · PrefectHQ/fastmcp

strawgate · 2026-04-11T21:47:43Z

Problem

Three related bugs in GoogleGenaiSamplingHandler's response conversion:

Thought parts leak as TextContent — Line 364 has a comment "Skip thought parts" but only checks if part.text: which is truthy for thought parts too. Internal reasoning text appears in MCP responses.
Thinking-only responses crash — When all parts have thought=True, response.text is None (Google SDK filters thoughts), causing ValueError: No content in response: STOP with no indication of why.
Safety errors are unhelpful — Tool-path error says "No content in response from completion" without including finish_reason. A safety-filtered response is indistinguishable from any other empty response.

Fix

Filter part.thought == True in _response_to_result_with_tools
Include finish_reason in tool-path error message
Detect thinking-only responses and provide a specific error message

Tests

7 new unit tests covering thought filtering, safety errors, and normal response passthrough. Tests construct mock GenerateContentResponse objects directly — no API calls.

🤖 Generated with Claude Code

marvin-context-protocol · 2026-04-11T21:49:51Z

Test Failure Analysis

(Updated to reflect latest workflow run 24293068620)

Summary: The only failing job is Tests: Python 3.10 on windows-latest. A test in tests/client/test_stdio.py hit the 5-second timeout on the Windows runner — this failure is unrelated to the PR changes.

Root Cause: This is a pre-existing flaky Windows runner issue. The same windows-latest / Python 3.10 job has failed on main in two other recent runs (24285512392 and 24256002384) with the same timeout pattern, each time in a different test file. The PR only modifies google_genai.py, its test file, and a minor fix to test_task_return_types.py — none of which are in test_stdio.py.

Suggested Solution: No action needed on this PR. The failure is a known intermittent Windows CI flake, not introduced by these changes. Re-running the failed job should be sufficient.

Detailed Analysis

Failing job: Tests: Python 3.10 on windows-latest (job 70933497777)

Failure pattern: The test runner reached the 19th test in tests\client\test_stdio.py (18 passed before it), then hit the 5s timeout. A second timeout then occurred while pytest was formatting the error report — specifically inside ntpath.realpath → _getfinalpathname on Windows, which is a known Windows path-resolution hang in pytest error formatting.

Relevant log excerpt:

tests\client\test_stdio.py ..................++++++ Timeout ++++++
~~~ Stack of MainThread ~~~
  File "ntpath.py", line 708, in realpath
    if _getfinalpathname(spath) == path:
++++++++++++++++++++++++ Timeout +++++++++++++++++++++
##[error]Process completed with exit code 1.

Evidence this is pre-existing flakiness:

Run 24285512392 on main: same Windows job failed, timeout in tests\tools\tool\test_content.py
Run 24256002384 on main: same Windows job failed
All other jobs (Ubuntu 3.10, Ubuntu 3.13, lowest-direct, MCP conformance, integration) passed

PR changes are unrelated to the failure:

src/fastmcp/client/sampling/handlers/google_genai.py
tests/client/sampling/handlers/test_google_genai_handler.py
tests/server/tasks/test_task_return_types.py

Related Files

tests/client/test_stdio.py — the file where the timeout occurred; not modified by this PR
src/fastmcp/client/sampling/handlers/google_genai.py — the implementation changed by this PR; unrelated to the failure

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 27c3b2b85a

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 8eb32536a4

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

- Filter thought parts (part.thought=True) from response content instead of leaking them as TextContent in _response_to_result_with_tools - Include finish_reason in error messages when no content is found, so safety-filtered responses (SAFETY, RECITATION, etc.) are distinguishable - Add specific error message for thinking-only responses in _response_to_create_message_result Fixes #3846 Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

Tests cover: - Thought parts filtered from tool-path responses - Thought-only responses produce descriptive errors - Safety-filtered responses include finish_reason in error - Normal responses (text + function calls) unaffected Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

Remove unused ty: ignore comments from lines where isinstance() narrows the type, and add correct ty: ignore[invalid-argument-type] and ty: ignore[not-subscriptable] comments on lines in newly added test functions where ty cannot infer the union type is a list. Also apply ruff format fix in test_task_return_types.py. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

Addresses review feedback: any() would misclassify mixed responses (thought + function_call) as thinking-only, hiding the real error. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

chatgpt-codex-connector Bot reviewed Apr 11, 2026

View reviewed changes

Comment thread src/fastmcp/client/sampling/handlers/google_genai.py Outdated

strawgate changed the title ~~[DNM] Fix GoogleGenaiSamplingHandler thought part filtering and error messages~~ Fix GoogleGenaiSamplingHandler thought part filtering and error messages Apr 11, 2026

strawgate changed the title ~~Fix GoogleGenaiSamplingHandler thought part filtering and error messages~~ fix: GoogleGenaiSamplingHandler leaks thought parts and gives unhelpful errors on empty responses Apr 11, 2026

chatgpt-codex-connector Bot reviewed Apr 11, 2026

View reviewed changes

Comment thread src/fastmcp/client/sampling/handlers/google_genai.py

strawgate removed the DON'T MERGE PR is not ready for merging. Used by authors to prevent premature merging. label Apr 11, 2026

strawgate and others added 5 commits April 12, 2026 13:10

Fix ruff format and ty check issues

7ab3d35

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

Fix: use all() not any() for thinking-only detection

187823c

Addresses review feedback: any() would misclassify mixed responses (thought + function_call) as thinking-only, hiding the real error. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

jlowin force-pushed the strawgate/fix-genai-thought-parts-3846 branch from c005b0a to 187823c Compare April 12, 2026 17:10

jlowin merged commit d5a3d54 into main Apr 12, 2026
8 checks passed

jlowin deleted the strawgate/fix-genai-thought-parts-3846 branch April 12, 2026 17:10

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix: GoogleGenaiSamplingHandler leaks thought parts and gives unhelpful errors on empty responses#3849

fix: GoogleGenaiSamplingHandler leaks thought parts and gives unhelpful errors on empty responses#3849
jlowin merged 5 commits intomainfrom
strawgate/fix-genai-thought-parts-3846

strawgate commented Apr 11, 2026 •

edited

Loading

Uh oh!

marvin-context-protocol Bot commented Apr 11, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

strawgate commented Apr 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

Fix

Tests

Uh oh!

marvin-context-protocol Bot commented Apr 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Test Failure Analysis

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

strawgate commented Apr 11, 2026 •

edited

Loading

marvin-context-protocol Bot commented Apr 11, 2026 •

edited

Loading