ChatCompletionRequest

Request for chat completion with tool calling.

messages object[]required

List of messages in the conversation

Array [

roleRole (string)required

Message role: 'user', 'assistant', 'system', or 'tool'

content object

Message content. Usually a string, but may be a multimodal content list (e.g. text + image_url) or null for assistant turns that only request tool calls.

anyOf

0-item-properties

null

null

tool_call_id object

For tool messages, the ID of the tool call

anyOf

string

string

null

null

name object

For tool messages, the tool name

anyOf

string

string

null

null

tool_calls object

For assistant messages replayed from prior turns, the OpenAI-format tool calls the model previously requested. Replaying these verbatim keeps the conversation prefix byte-for-byte identical so the model server's prompt cache hits on follow-up turns.

anyOf

array

Array [

object

]

null

null

]

max_tool_iterationsMax Tool Iterations (integer)

Maximum number of tool call iterations (default: 5)

Possible values: >= 1 and <= 10

Default value: 5

streamStream (boolean)

If true, stream the final assistant response in the body as newline-delimited JSON.

Default value: false

enable_thinking object

Per-request thinking toggle. None means use the provider default. Ignored by providers that do not expose a per-request thinking switch.

anyOf

boolean

boolean

null

null

ChatCompletionRequest

{
  "messages": [
    {
      "role": "string",
      "tool_call_id": "string",
      "name": "string",
      "tool_calls": [
        {}
      ]
    }
  ],
  "max_tool_iterations": 5,
  "stream": false,
  "enable_thinking": true
}