Back to Qwen Code

call_llm

packages/desktop/apps/electron/resources/docs/llm-tool.md

0.18.01013 B
Original Source

call_llm

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

Parameters

FieldTypeDescription
promptstringInstructions for the model
attachmentsarrayOptional files on disk
modelstringOptional Qwen model ID or short name
systemPromptstringOptional system prompt
maxTokensnumberMaximum output tokens
temperaturenumberSampling temperature
thinkingbooleanRequest extended reasoning when supported
outputFormatstringOptional structured output style
outputSchemaobjectOptional 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"
}