Deepseek-chat:options:thinking:type:enabled thinking tools bug

[ai-llt]

Description

agent: {
deepseek-chat: {
"mode": "primary",
"model": "deepseek/deepseek-chat",
"options": {
"thinking": {
"type": "enabled"
}
}
}
}

→ Read opencode.json
Error: File not found: D:\Projects\LLM_Workspace\bear_party\opencode.json
Missing reasoning_content field in the assistant message at message index 2. For more information, please refer to https://api-docs.deepseek.com/guides/thinking_mode#tool-calls

$ find . -name "plans" -type d
FIND: Parameter format not correct
Missing reasoning_content field in the assistant message at message index 5. For more information, please refer to https://api-docs.deepseek.com/guides/thinking_mode#tool-calls

$ dir /b
Get-ChildItem: Cannot find path 'D:\b' because it does not exist.
Missing reasoning_content field in the assistant message at message index 8. For more information, please refer to https://api-docs.deepseek.com/guides/thinking_mode#tool-calls

Above is the issue where the thinking parameter cannot be used and will result in an error in thinking mode. Please help me

Plugins

No response

OpenCode version

No response

Steps to reproduce

No response

Screenshot and/or share link

No response

Operating System

No response

Terminal

No response

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• The problem with deepseek-reasoner in OpenCode #8934: The problem with deepseek-reasoner in OpenCode
• Deepseek missing "reasoning_content" field ; deepseek_reasoner model erroring out #5027: Deepseek missing "reasoning_content" field ; deepseek_reasoner model erroring out
• DeepSeek 3.2: Missing reasoning_content #4947: DeepSeek 3.2: Missing reasoning_content
• DeepSeek V3.2 : reasoning_content not cleared from message history on new turns, causing excess token usage and violating API spec #5577: DeepSeek V3.2 : reasoning_content not cleared from message history on new turns, causing excess token usage and violating API spec

Feel free to ignore if none of these address your specific case.

[tongsh6]

I can reproduce the same class of bug in a slightly different setup: a custom local subagent using deepseek/deepseek-reasoner with tools enabled.

Local agent file:

~/.config/opencode/agents/architect.md

---
description: Read-only architecture review agent
mode: subagent
model: deepseek/deepseek-reasoner
temperature: 0.1
steps: 16
tools:
  write: false
  edit: false
  bash: true
---

Observed error:

Missing reasoning_content field in the assistant message at message index 32.
For more information, please refer to https://api-docs.deepseek.com/guides/thinking_mode#tool-calls

So this does not seem limited to deepseek-chat with thinking.type=enabled; it also appears in custom subagent flows with deepseek-reasoner when tool calls are involved.

Related issue with more detail: #8934.