openclaw - 💡(How to fix) Fix Subagent spawn fails: thinking enabled but reasoning_content missing [1 participants]

FailoverError: LLM request rejected: thinking is enabled but reasoning_content is missing in assistant tool call message at index 2

TL;DR

The most likely fix is to modify the subagent system to propagate reasoning_content in tool call messages when thinking mode is enabled in the parent session.

Guidance

• Verify that the reasoning_content is indeed missing in the tool call messages sent by the subagent by inspecting the request payload.
• Check the subagent system's implementation to see where the reasoning_content is supposed to be added to the tool call messages.
• Consider modifying the subagent system to include the reasoning_content in the tool call messages when thinking mode is enabled.
• As a temporary workaround, use the suggested exec + direct CLI approach for tasks instead of subagents.

Example

No code snippet is provided as the issue does not contain sufficient implementation details.

Notes

The provided analysis suggests that the issue is specific to the subagent system and the propagation of reasoning_content when thinking mode is enabled. The workaround provided in the issue may not be suitable for all use cases.

Recommendation

Apply the workaround by using exec + direct CLI for tasks instead of subagents, as modifying the subagent system may require significant changes and testing.