MiMo API Update Breaking Reasoning Content Passthrough

[JoaoPedroCampanari]

MiMo model API (Xiaomi) now requires reasoning_content to be passed back in subsequent API messages when thinking mode is active. Sub-agents spawned by OpenClaude fail because the reasoning_content field is not preserved and forwarded in the conversation history sent to the API. This was working normally before — the issue started suddenly without any changes to my configuration or setup. I have already tried downgrading OpenClaude to a previous version and the
error persists.
Steps to Reproduce

1. Configure OpenClaude to use a MiMo model (e.g., mimo-v2.5-pro) through an OpenAI-compatible
2. Trigger any action that spawns a background agent (e.g., Explore, Plan, or any Agent tool invocation)
3. The agent fails immediately with a 400 error

Expected Behavior

Sub-agents should be spawned and execute normally, preserving all required API fields (including reasoning_content) in the conversation history, as it did before this issue appeared.

Actual Behavior

The API returns a 400 error:

API Error: 400 OpenAI API error 400: data:{"error":{"code":"400","message":"Param Incorrect","param":"The reasoning_content in the thinking mode must be passed back to the API.","type":""}}

This happens because OpenClaude does not include the reasoning_content from previous assistant messages when constructing the conversation history for sub-agent API calls.

Environment

• OpenClaude version: 0.10.0
• OS: Windows 10 Pr
• Terminal: (PowerShell 5, Windows Terminal)
• Provider: https://token-plan-sgp.xiaomimimo.com/v1
• Model: mimo-v2.5-pro

API Error: 400 OpenAI API error 400: data:{"error":{"code":"400","message":"Param Incorrect","param":"The reasoning_content in the thinking mode must be passed back to the API.","type":""}}
[openai_category=unknown,host=localhost]

Additional Context

• Everything was working normally with the exact same configuration. This issue started suddenly without any changes on my side.
• I have already attempted a downgrade of OpenClaude to a previous version and the error persists, which strongly suggests the change was introduced on the MiMo API side (likely a Xiaomi server-side update).
• The behavior is similar to DeepSeek's reasoning models, which also require reasoning_content to be passed back in follow-up messages.
• The fix likely involves ensuring that the OpenAI-compatible client layer preserves and forwards the reasoning_content field when serializing assistant messages in the conversation history for sub-agent calls.
• Normal conversations work perfectly fine. The error only occurs when trying to spawn and use agents (Explore, Plan, Agent tool, etc.). Direct chat interaction with the model is unaffected.

[JoaoPedroCampanari]

#1155