Chat Completions

The chat completions endpoint creates a model response from a list of messages.

Endpoint

POST https://api.glbwintk.com/v1/chat/completions

Request

curl https://api.glbwintk.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "your-model-id",
    "messages": [
      {
        "role": "system",
        "content": "You are a concise assistant."
      },
      {
        "role": "user",
        "content": "Write one sentence about GLBWinTK."
      }
    ],
    "temperature": 0.2,
    "max_tokens": 80
  }'

Message Roles

Role	Purpose
system	Sets behavior or instructions for the response.
user	Contains user input.
assistant	Represents prior assistant output in a conversation.

Response Shape

Successful responses use an OpenAI-compatible structure:

{
  "id": "chatcmpl-example",
  "object": "chat.completion",
  "model": "your-model-id",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "GLBWinTK provides one OpenAI-compatible API endpoint for AI applications."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 20,
    "completion_tokens": 12,
    "total_tokens": 32
  }
}

Production Notes

• Start with short prompts while validating a model id.
• Set reasonable max_tokens values for predictable usage.
• Log request ids and status codes when diagnosing failures.