Generate with Chat Completions

POST /v1/chat/completions

OpenAI Chat Completions-compatible generation over messages[]. Read choices[].message.content for text; tool-capable responses can instead carry tool_calls. A content item can be text, image_url with a data URL, or input_video with base64 video bytes on macOS and Linux builds with native video support, which also requires ffmpeg and ffprobe on the service machine. Streaming uses chat.completion.chunk events terminated by data: [DONE]; do not apply that parser to the other protocols. Each endpoint requires a model whose supported_endpoints includes that exact path. These APIs forward the selected model's protocol, so optional parameters and response details can vary by local worker or remote provider. Local models keep inputs on the device; a configured remote model receives the request sent to it.

Guide and examplesLanguage and vision·Local Gateway API·All endpoints

Request body

application/jsonRequired

Responses

{
  "id": "chatcmpl_example",
  "object": "chat.completion",
  "model": "Qwen/Qwen3.5-4B",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Hello!"
      },
      "finish_reason": "stop"
    }
  ]
}