DeepSeek-V4 reasoning_content error still reproducible on v0.1.5.post3 with WebUI + exec tool on Windows

[e16a]

Bug Description

I can still reproduce a DeepSeek-V4 reasoning_content must be passed back error on v0.1.5.post3.

This seems similar to #3469, but the trigger path is different:

• OS: Windows 10
• nanobot: v0.1.5.post3
• Channel: WebUI / websocket
• Tool involved: exec
• Provider: DeepSeek
• Model: deepseek-v4-pro
• Frontend: official WebUI
• Node: v22.14.0

Steps to Reproduce

1. I created a local Windows TTS script:
   D:\nanobot-015-test\TOOLS\say.bat "测试成功，可以说话了。"
   The script works correctly when run manually in CMD.

2. Then I asked nanobot in WebUI to test this TTS tool. The agent used several tool calls, including checking the file, running the script, and attempting to adjust/debug it.

3. After several tool calls, the next request failed with:

Error: {'message': 'The reasoning_content in the thinking mode must be passed back to the API.', 'type': 'invalid_request_error', 'param': None, 'code': 'invalid_request_error'}

Expected Behavior

Expected Behavior

Tool-heavy sessions with DeepSeek-V4 thinking mode should preserve and replay reasoning_content correctly.

Actual Behavior

After several exec/tool calls, DeepSeek rejects the next request because reasoning_content was not passed back.

Why this may be different from #3469

#3469 reports this issue with Linux + QQ channel / CLI multi-round chat.

This reproduction happens on Windows 10 with the official WebUI/websocket channel and the exec tool, on v0.1.5.post3.

Relevant Logs

Relevant Notes

The local command itself works correctly in CMD:

D:\nanobot-015-test\TOOLS\say.bat "测试成功， 可以说话了。"

So the issue does not appear to be caused by the TTS script itself. It seems related to DeepSeek reasoning history after tool calls.

nanobot Version

v0.1.5.post3

Python Version

3.12

Operating System

Windows

Channel / Platform

WebSocket

LLM Provider

DeepSeek

Configuration (Optional)

Additional Context

No response