fix: validate extended thinking blocks and tool use/result pairing

[bigboateng]

Important

Fixes critical message history issues by validating extended thinking blocks and tool use/result pairing to prevent 400 errors.

• Behavior:
  • Added ensureThinkingBlocksForExtendedThinking() to filter out assistant messages without thinking blocks when thinkingBudget is enabled.
  • Rewrote cleanMessageHistory() to ensure tool_result blocks have corresponding tool_use blocks in the immediately previous message.
• Errors Fixed:
  • Fixed errors: "Expected thinking or redacted_thinking, but found text" and "unexpected tool_use_id found in tool_result blocks".
• Testing:
  • Extended thinking test passes with 15+ tool calls across multiple turns without errors.

^{This description was created by for 2a28db5. You can customize this summary. It will automatically update as commits are pushed.}

[ellipsis-dev]

Important

Looks good to me! 👍

Reviewed everything up to 2a28db5 in 39 seconds. Click for details.

• Reviewed 23 lines of code in 1 files
• Skipped 0 files when reviewing.
• Skipped posting 4 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:9

• Draft comment:
  Extended Thinking Block Validation fix is clearly described. Consider adding a reference to related tests or documentation for easier context lookup.
• Reason this comment was not posted:
  Confidence changes required: 33% <= threshold 50% None

2. .changeset/fix-extended-thinking-blocks.md:11

• Draft comment:
  Tool Use/Result Pairing fix is well-documented. It might help to clarify any backwards compatibility implications if the new per-message pairing changes behavior.
• Reason this comment was not posted:
  Confidence changes required: 33% <= threshold 50% None

3. .changeset/fix-extended-thinking-blocks.md:14

• Draft comment:
  Error message quotes are clear. Double-check that these strings match the actual runtime error messages exactly.
• Reason this comment was not posted:
  Confidence changes required: 0% <= threshold 50% None

4. .changeset/fix-extended-thinking-blocks.md:17

• Draft comment:
  Testing note is helpful. Consider including a reference or link to the extended thinking test results for future maintainability.
• Reason this comment was not posted:
  Confidence changes required: 33% <= threshold 50% None

Workflow ID: wflow_PJUz2wtI3HjbKZFg

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