Chat completions

The kenari chat completions endpoint is compatible with OpenAI, so it can be used as a direct replacement for https://api.openai.com/v1. Point the base URL to https://kenari.id/v1 and use the kn-... API key. Tokens are still billed from the prepaid balance in Rupiah.

Endpoint

POST /v1/chat/completions

Request

Two fields are required: model and messages.

The model field accepts a model id from GET /v1/models, a :free variant (for example step-3-7-flash:free), or a route name you created yourself (for example opus-hemat, see Routing).

Other supported fields:

Field	Type	Description
`stream`	boolean	Send the response as an SSE stream. See Streaming.
`max_tokens`	integer	Output token limit.
`temperature`	number	Sampling randomness level.
`top_p`	number	Nucleus sampling.
`stop`	string or array	Text sequences that stop generation.
`frequency_penalty`	number	Token frequency penalty.
`presence_penalty`	number	Token presence penalty.
`tools`	array	Tool definitions. See Tools.
`tool_choice`	string or object	How the model chooses a tool.
`response_format`	object	Output format, for example JSON.
`reasoning_effort`	string	For reasoning models. See Reasoning.
`plugins`	array	Pre-dispatch plugins. Today only `file-parser`, for reading documents. See Read documents.

Fields not listed in the table above are forwarded as-is to the upstream model.

messages

The messages field is an array of {role, content} objects. The role value is one of system, user, assistant, or tool. The content value can be a string, or an array of parts for multimodal input: text, image_url, and file. A file part carries a PDF or document image and needs the file-parser plugin, see Read documents.

{
  "model": "step-3-7-flash",
  "messages": [
    { "role": "system", "content": "Kamu asisten yang ringkas." },
    { "role": "user", "content": "Halo!" }
  ]
}

Response

The response follows the OpenAI chat.completion shape: id, object, created, model, choices[], and usage. Each choices entry has index, message{role, content}, and finish_reason. The usage object contains prompt_tokens, completion_tokens, and total_tokens. The usage.prompt_tokens_details.cached_tokens field appears when the upstream reports a cache read.

{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "created": 1717000000,
  "model": "step-3-7-flash",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "Halo! Ada yang bisa dibantu?" },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 12,
    "completion_tokens": 9,
    "total_tokens": 21
  }
}

Examples

OpenAI SDK Python, set base_url to https://kenari.id/v1:

from openai import OpenAI
client = OpenAI(base_url="https://kenari.id/v1", api_key="kn-...")
response = client.chat.completions.create(model="step-3-7-flash", messages=[{"role": "user", "content": "Halo!"}])
print(response.choices[0].message.content)

OpenAI SDK TypeScript, set baseURL to https://kenari.id/v1:

import OpenAI from "openai";
const client = new OpenAI({ baseURL: "https://kenari.id/v1", apiKey: "kn-..." });
const res = await client.chat.completions.create({ model: "step-3-7-flash", messages: [{ role: "user", content: "Halo!" }] });
console.log(res.choices[0].message.content);

curl:

curl https://kenari.id/v1/chat/completions -H "Authorization: Bearer kn-..." -H "Content-Type: application/json" -d '{"model":"step-3-7-flash","messages":[{"role":"user","content":"Halo!"}]}'

Streaming

Set "stream": true in the request to receive the response as an SSE stream token by token. Event format details and stream reading steps are in Streaming.