Available when the endpoint opens

API

The Cold Serve API follows the OpenAI chat completions format. Any OpenAI-compatible client works by changing the base URL and the key.

Cold Serve is in private testing. This page describes the shape of the public endpoint; keys, models and prices are published when it opens.

Base URL

https://api.coldservelabs.com/v1

Authentication

Send your API key as a bearer token in the Authorization header of every request.

Authorization: Bearer YOUR_API_KEY

Endpoints

POST /chat/completions: Chat completions. Set "stream": true to receive the reply as server-sent events. Every response reports token usage in the usage field.
GET /models: The models we serve, each with its price.

Example

curl https://api.coldservelabs.com/v1/chat/completions \
  -H "Authorization: Bearer $COLDSERVE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MODEL_ID",
    "messages": [{"role": "user", "content": "Hello"}],
    "stream": true
  }'

With an OpenAI client library, set the base URL and key:

import os
from openai import OpenAI

client = OpenAI(base_url="https://api.coldservelabs.com/v1",
                api_key=os.environ["COLDSERVE_API_KEY"])

Your data

Requests run under zero data retention: prompts and outputs are not stored, not logged and never used for training.