fix: prevent infinite loops and 400 errors in message history management

[bigboateng]

Important

Fixes infinite loops and 400 errors by ensuring thinking block presence and tool use/result pairing in message history management.

• Behavior:
  • Fixes infinite loops by adding placeholder thinking blocks when Claude doesn't emit them, using ensureThinkingBlockForResponse() in loop.ts.
  • Prevents 400 errors by ensuring every assistant message starts with a thinking block when thinkingBudget is enabled.
  • Ensures tool use/result pairing by rewriting cleanMessageHistory() in utils/message-processing.ts.
• Functions:
  • Adds ensureThinkingBlockForResponse() in utils/message-processing.ts to handle missing thinking blocks.
  • Updates cleanMessageHistory() to validate tool use/result pairing on a per-message basis.
• Testing:
  • Extended thinking test passes with multiple tool uses in 27.5s without loops.
  • Properly handles missing thinking blocks by adding placeholders.

^{This description was created by for 5e7c22c. You can customize this summary. It will automatically update as commits are pushed.}

[ellipsis-dev]

Important

Looks good to me! 👍

Reviewed everything up to 5e7c22c in 58 seconds. Click for details.

• Reviewed 167 lines of code in 3 files
• Skipped 0 files when reviewing.
• Skipped posting 3 draft comments. View those below.
• Modify your settings and rules to customize what types of comments Ellipsis leaves. And don't forget to react with 👍 or 👎 to teach Ellipsis.

1. .changeset/fix-extended-thinking-blocks.md:5

• Draft comment:
  The release note now clearly lists the three major fixes (infinite loop prevention, extended thinking block validation, and tool use/result pairing) and matches the implementation changes. Make sure this documentation remains in sync with future changes.
• Reason this comment was not posted:
  Confidence changes required: 0% <= threshold 50% None

2. loop.ts:257

• Draft comment:
  The new call to ensureThinkingBlockForResponse properly injects a placeholder thinking block when missing. Using '!!thinkingBudget' to toggle this behavior is clear. Consider if deep cloning might be needed if the thinking block structure becomes more complex.
• Reason this comment was not posted:
  Confidence changes required: 33% <= threshold 50% None

3. utils/message-processing.ts:296

• Draft comment:
  In ensureThinkingBlockForResponse, the placeholder thinking block is cloned using a shallow spread. This is acceptable for now, but consider a deep clone if nested properties might be modified in the future.
• Reason this comment was not posted:
  Confidence changes required: 33% <= threshold 50% None

Workflow ID: wflow_Q8sIP977rldlW1SW

^{You can customize by changing your verbosity settings, reacting with 👍 or 👎, replying to comments, or adding code review rules.}

[ellipsis-dev]

Skipped PR review on 58a2632 because no changed files had a supported extension. If you think this was in error, please contact us and we'll fix it right away.