openclaw - 💡(How to fix) Fix [Bug]: [Bug] DeepSeek V4/Reasoner 模型在多轮对话切换或 Telegram 环境下报错 400: reasoning_content missing [1 comments, 2 participants]

[agent/embedded] embedded run agent end: runId=8bea32c4-e762-469c-b09a-23a0b2d9703d isError=true model=deepseek-v4-pro provider=deepseek error=LLM request failed: provider rejected the request schema or tool payload. rawError=400 The reasoning_content in the thinking mode must be passed back to the API. [agent/embedded] embedded run agent end: runId=8bea32c4-e762-469c-b09a-23a0b2d9703d isError=true model=deepseek-v4-pro provider=deepseek error=LLM request failed: provider rejected the request schema or tool payload. rawError=400 The reasoning_content in the thinking mode must be passed back to the API. [agent/embedded] embedded run agent end: runId=8bea32c4-e762-469c-b09a-23a0b2d9703d isError=true model=deepseek-v4-pro provider=deepseek error=LLM request failed: provider rejected the request schema or tool payload. rawError=400 The reasoning_content in the thinking mode must be passed back to the API.

TL;DR

The issue can be fixed by ensuring the reasoning_content field is properly passed back to the API when switching to the DeepSeek V4 model.

Guidance

• Verify that the reasoning_content field is being correctly persisted and passed back to the API in the Telegram conversation flow.
• Check the OpenClaw code to ensure it is correctly handling the reasoning_content field when switching between models.
• Consider adding a default or empty reasoning_content field when switching to the DeepSeek V4 model to prevent API errors.
• Review the API documentation to ensure that the reasoning_content field is being passed in the correct format and location.

Example

No code example is provided as the issue does not contain sufficient information about the codebase.

Notes

The issue seems to be related to the handling of the reasoning_content field when switching between models, specifically when using the DeepSeek V4 model. The exact cause and solution may depend on the specific implementation details of the OpenClaw code and the API.

Recommendation

Apply a workaround to ensure the reasoning_content field is properly passed back to the API when switching to the DeepSeek V4 model, such as adding a default or empty field. This will prevent API errors and allow the conversation to continue.