Chat Completions

POST /v1/chat/completions is the most widely compatible chat endpoint for existing OpenAI SDKs and clients.

curl

# Send a non-streaming POST request to the complete Chat endpoint.
curl https://sprelaytoken.com/v1/chat/completions \
  -H "Authorization: Bearer $SPRELAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.6-sol",
    "messages": [
      {"role": "system", "content": "Answer concisely."},
      {"role": "user", "content": "What is an API gateway?"}
    ]
  }'

-H adds headers and -d supplies the JSON body. The system message sets behavior; the following user message asks the question. JSON cannot contain comments, so preserve its punctuation when editing.

Python SDK

# Read the secret from the environment and import the compatible SDK.
import os
from openai import OpenAI

# The SDK appends /chat/completions to this base URL.
client = OpenAI(
    api_key=os.environ["SPRELAY_API_KEY"],
    base_url="https://sprelaytoken.com/v1",
    timeout=60.0,
)

# create() waits for one complete non-streaming response.
response = client.chat.completions.create(
    model="gpt-5.6-sol",
    messages=[
        {"role": "system", "content": "Answer concisely."},
        {"role": "user", "content": "What is an API gateway?"},
    ],
)

# Print only the text from the first candidate.
print(response.choices[0].message.content)

JavaScript SDK

// Import the SDK after installing the openai package.
import OpenAI from 'openai'

// Read the key from the Node.js process environment.
const client = new OpenAI({
  apiKey: process.env.SPRELAY_API_KEY,
  baseURL: 'https://sprelaytoken.com/v1',
  timeout: 60_000,
})

// await pauses until the non-streaming response is complete.
const response = await client.chat.completions.create({
  model: 'gpt-5.6-sol',
  messages: [
    { role: 'system', content: 'Answer concisely.' },
    { role: 'user', content: 'What is an API gateway?' },
  ],
})

// Print only the first assistant message.
console.log(response.choices[0].message.content)

model and messages are required. Add optional fields such as stream, temperature, max_tokens, or tools only after confirming that the selected model supports them. For multi-turn chat, send the relevant history again and control its size to limit latency and cost.