fix(core): flush transcript for pure tool-call responses to ensure BeforeTool hooks see complete state

[krishdef7]

Summary

Extends #17996 to cover a remaining edge case: when the model emits a functionCall with no response text and no thoughts, recordMessage was never called, leaving BeforeTool hooks with no gemini entry in the transcript at dispatch time.

Details

PR #17996 introduced the hasThoughts flag to flush the transcript when thoughts precede a tool call. However, the flush condition was responseText || hasThoughts — so a pure tool call (no text, no thoughts) still skipped recordMessage.

Before this PR, transcript state at BeforeTool hook dispatch:

{
  "messages": [
    { "type": "user", "content": "analyze test.py" }
  ]
}

After this PR, transcript state at BeforeTool hook dispatch:

{
  "messages": [
    { "type": "user", "content": "analyze test.py" },
    { "type": "gemini", "content": "" }
  ]
}

Changes:

• Add hasThoughts flag tracking (same pattern as Save transcript to file before function call is performed #17996, needed since that PR isn't merged yet)
• Extend flush condition: responseText || hasThoughts || hasToolCall
• This is a 3-line change in processStreamResponse — no new dependencies, no architectural changes

Why processStreamResponse and not ToolExecutor:
An earlier approach considered flushing at the tool dispatch point in CoreToolScheduler. That would require threading chatRecordingService through CoreToolScheduler → ToolExecutor, coupling recording logic into the scheduler. The correct separation of concerns keeps all recording in GeminiChat.processStreamResponse, where it already lives.

Related Issues

Related to #17996, #17922

How to Validate

1. Create a BeforeTool hook that reads the session transcript file
2. Send a prompt that causes the model to immediately call a tool with no preamble text
3. Before this PR: transcript contains only the user message at hook time
4. After this PR: transcript contains both the user message and a gemini-type entry

Unit test added: should flush transcript before tool dispatch for pure tool call with no text or thoughts in geminiChat.test.ts — mocks a pure functionCall stream and asserts a gemini-type message is written to disk before stream completion.

Pre-Merge Checklist

• Added/updated tests
• Validated build passes (npm run build)
• No breaking changes

[gemini-code-assist]

Summary of Changes

Hello @krishdef7, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request refines the chat transcript recording mechanism to ensure that BeforeTool hooks consistently receive an up-to-date transcript, particularly when the model directly invokes a tool without generating any preceding text or thoughts. The change prevents a previous scenario where the transcript would be incomplete, thereby improving the reliability of tool execution workflows.

Highlights

• Transcript Flushing Logic: Modified the transcript flushing condition in processStreamResponse to ensure that a 'gemini' entry is recorded even when the model emits a pure tool call (without response text or thoughts). This addresses an edge case where BeforeTool hooks would not see the complete transcript state.
• New hasThoughts Flag: Introduced a hasThoughts flag to track if the model's response chunk contains any thoughts, which is now incorporated into the transcript flushing logic.
• Unit Test Coverage: Added a new unit test case (should flush transcript before tool dispatch for pure tool call with no text or thoughts) to validate the corrected transcript flushing behavior for pure tool calls.

Changelog

• packages/core/src/core/geminiChat.test.ts
  • Added a new test case to verify transcript flushing for pure tool calls without text or thoughts.
• packages/core/src/core/geminiChat.ts
  • Introduced a hasThoughts flag to track if model responses contain thoughts.
  • Updated the transcript flushing condition to include hasThoughts and hasToolCall, ensuring recordMessage is called for pure tool invocations.

Activity

• Added/updated tests to cover the new functionality.
• Validated that the build passes successfully.
• Confirmed that no breaking changes were introduced.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request addresses an important edge case where the session transcript was not updated for model responses containing only a tool call, without any accompanying text or thoughts. This could lead to BeforeTool hooks operating on stale state.

The fix correctly extends the condition for recording a model message to include hasToolCall, ensuring that a gemini entry is always added to the transcript when the model produces any output, including a pure tool call. The addition of the hasThoughts flag, while duplicated from another pending PR, is necessary here for completeness.

The new unit test effectively validates this fix by simulating a pure tool call stream and asserting that the transcript is written to disk with the new gemini message. The changes are well-contained and the logic is sound. I have no further comments.