Chat Completions¶

POST /v1/chat/completions

Generate a model response for a conversation. This endpoint is OpenAI-compatible.

Request¶

POST https://api.mangoboost.io/v1/chat/completions
Authorization: Bearer <your-api-key>
Content-Type: application/json

Body parameters¶

The request body follows the OpenAI Chat Completions schema. The parameters below are the ones these docs cover; an OpenAI-shaped field not listed here is passed through to the model server rather than validated by Mango Inference, so treat anything outside this table as unsupported until you have tested it.

Parameter	Type	Required	Description
`model`	string	Yes	Model ID. See Models and pricing.
`messages`	array	Yes	Conversation messages (`role` + `content`).
`max_tokens`	integer	No	Maximum tokens to generate.
`temperature`	number	No	Sampling temperature.
`top_p`	number	No	Nucleus sampling cutoff.
`stream`	boolean	No	Stream tokens as server-sent events. Defaults to `false`.
`stop`	string / array	No	Stop sequence(s).
`tools`	array	No	Tool definitions. See Tool calling.
`tool_choice`	string / object	No	How the model selects tools.
`response_format`	object	No	Constrain output. See Structured output.

Sampling parameters you omit fall back to the serving defaults for the model. Those defaults are not published and can differ per model, so set the ones your application depends on explicitly rather than relying on them.

Response¶

{
  "id": "e9c0b646861d4c0085503fe54d796afa",
  "object": "chat.completion",
  "created": 1785742283,
  "model": "zai-org/GLM-5.3",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Hello! How can I help you today?",
        "reasoning_content": "…",
        "tool_calls": null
      },
      "logprobs": null,
      "finish_reason": "stop",
      "matched_stop": 154827
    }
  ],
  "usage": {
    "prompt_tokens": 14,
    "total_tokens": 261,
    "completion_tokens": 247,
    "prompt_tokens_details": null,
    "reasoning_tokens": 237
  },
  "metadata": {
    "weight_version": "default"
  }
}

Response fields¶

Field	Description
`choices[].message.content`	The generated text.
`choices[].message.reasoning_content`	Reasoning the model produced before its answer, on reasoning-capable models. Not an OpenAI field. See Reasoning.
`choices[].message.tool_calls`	Tool calls the model wants you to run, or `null`. See Tool calling.
`choices[].finish_reason`	Why generation stopped (`stop`, `length`, `tool_calls`, …).
`choices[].matched_stop`	Which stop condition ended generation: a token ID, or the stop string that matched. Not an OpenAI field.
`usage`	Token counts for the request. `reasoning_tokens` counts reasoning separately, and is included in `completion_tokens`.
`metadata.weight_version`	The model weights that served the request. Not an OpenAI field.

The three fields marked "not an OpenAI field" are additions. OpenAI SDKs ignore unknown response fields, so they do not break existing client code. However, code that round-trips a response through a strict schema may need to allow them.

Errors¶

See Reliability and error handling.