API reference

Understand the request shape, response fields, and common parameters.

POST/chat/completions

Append this path to your configured base URL. The reference follows a common chat-completions request shape; supported fields depend on the selected model and endpoint.

Authentication

Send a bearer key in the Authorization header and JSON with Content-Type: application/json. Keep the credential on your server.

Request body

Field	Type	Description
`model`	string · optional	An enabled model ID. If omitted, your saved default model is used.
`messages`	array · required	An ordered conversation. Each message includes a role and content.
`messages[].role`	string	A supported role, commonly system, user, assistant, or tool.
`messages[].content`	string	Text content. This release supports text and code; image and audio input are not accepted.
`stream`	boolean · optional	Requests streaming where the chosen model and endpoint support it.
`temperature`	number · optional	Controls sampling for models that support it. Check model-specific limits.
`tools`	array · optional	Tool definitions for compatible models. Your application validates and executes calls.

TypeScript · chat completions

const response = await fetch(
  `${process.env.PRISMHUB_BASE_URL}/chat/completions`,
  {
    method: "POST",
    headers: {
      Authorization: `Bearer ${process.env.PRISMHUB_API_KEY}`,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      model: "provider/model-id",
      messages: [
        { role: "user", content: "Build something brilliant." },
      ],
    }),
  },
);
if (!response.ok) throw new Error(`HTTP ${response.status}`);
const result = await response.json();

Reference template. Set a configured base URL ending in /api/v1, a server-side key, and a supported model identifier. This page sends no API requests.

Example response

This is an illustrative response shape. Actual responses include the selected model, PrismHub request identifier, and metered usage.

Illustrative JSON response

{
  "id": "example-response-id",
  "object": "chat.completion",
  "model": "provider/model-id",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "What would you like to build?"
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 24,
    "completion_tokens": 9,
    "total_tokens": 33
  }
}

Streaming and tools

Streaming responses commonly use server-sent events. Accumulate content carefully and handle interruptions. Tool definitions describe actions; the model does not execute them. Validate tool arguments and permissions in your own application.