Repository navigation
Only return tool_call parameters but can not run the tool with local LLM #7486
Description
Description
- Air-gapped env, start a local llm “Qwen3-32B” or “Qwen3-Coder-30B-A3B-Instruct” with vllm
- Windows10 vscode, install opencode plugin to use this provider and give a simple prompt "introduce this project"
It only returns the the tool_call parameters but doesnot run the tool.
response:
<tool_call>
{"name": "glob", "arguments": {"pattern": "*/"}}
</tool_call>
configuration:
{
"$schema": "https://opencode.ai/config.json",
"provider": {
"lmstudio": {
"npm": "@ai-sdk/openai-compatible",
"name": "LM Studio (local)",
"options": {
"baseURL": "https://10.200.10.122:8088/v1"
},
"models": {
"Qwen/Qwen3-Coder-30B-A3B-Instruct": {
"name": "Qwen/Qwen3-Coder-30B-A3B-Instruct",
"tool_call": true,
"reasoning": true
}
}
}
}
}
Metadata
Metadata
Assignees
Labels
No labels
Activity
sorokinmax commentedon Mar 13, 2026
The same problem occurs when using the Step 3.5 Flash model via the connected OpenRouter in the OpenCode desktop application.
supperman009 commentedon May 13, 2026
Has the problem been solved? I'm having the same issue as you. The model works fine when accessed in Hermes, but the same problem occurs when accessing it in OpenCode — the tool call content is directly displayed in the conversation window.
supperman009 commentedon May 13, 2026
hermes is ok
github-actions commentedon Jul 13, 2026
To stay organized issues are automatically closed after 60 days of no activity. If the issue is still relevant please open a new one.