call_llm

call_llm asks the configured Qwen backend for a focused completion from inside a session tool flow.

Parameters

Field	Type	Description
`prompt`	string	Instructions for the model
`attachments`	array	Optional files on disk
`model`	string	Optional Qwen model ID or short name
`systemPrompt`	string	Optional system prompt
`maxTokens`	number	Maximum output tokens
`temperature`	number	Sampling temperature
`thinking`	boolean	Request extended reasoning when supported
`outputFormat`	string	Optional structured output style
`outputSchema`	object	Optional JSON schema

Examples

json

{
  "prompt": "Summarize the current implementation risk in three bullets.",
  "model": "qwen3-coder-flash"
}

json

{
  "prompt": "Extract the API endpoints from this file.",
  "attachments": [{ "path": "/path/to/server.ts", "startLine": 1, "endLine": 200 }],
  "outputFormat": "extraction"
}