Repository navigation
feat(shell): stream thinking content to terminal in real-time - #1538
FuryMartin wants to merge 1 commit into
Conversation
![devin-ai-integration[bot]](https://avatars.githubusercontent.com/in/811515?s=60&v=4){kind=small}
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: bdc780b3ac
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
| if self.is_think and self.raw_text: | ||
| return BulletColumns( | ||
| Markdown(self.raw_text, style="grey50 italic"), | ||
| bullet_style="grey50", |
There was a problem hiding this comment.
In the normal shell path we read the wire with merge=False (src/kimi_cli/ui/shell/__init__.py:467-468), so a reasoning turn arrives as many ThinkParts. _LiveView.append_content() only calls refresh_soon() when the block is created or its type changes, not after later append()s (src/kimi_cli/ui/shell/visualize.py:1298-1311). Because this code now materializes Markdown(self.raw_text, ...) inside compose(), the live view caches a snapshot of the first thought fragment and stays stale until some unrelated event (tool call, step end, etc.) forces a recomposition. In practice, long tool-free reasoning steps will appear to stop updating after the first chunk, so the new “real-time” thinking stream is incomplete in the default interactive shell.
Useful? React with 👍 / 👎.
Related Issue
Resolve #(issue_number)
Description
Checklist
make gen-changelogto update the changelog.make gen-docsto update the user documentation.