Change base_url to https://api.revoye.com/v1, use your Revoye key as the API key, and set the model to revoye/auto. The SDK call waits for the response and returns it in choices[0].message.content, exactly as it would from OpenAI. Everything else about your request code stays as it is.

Python

import os
from openai import OpenAI
 
client = OpenAI(
    base_url="https://api.revoye.com/v1",
    api_key=os.environ["REVOYE_KEY"],
    timeout=660,                        # seconds; a request can be held for up to 10 minutes
)
 
completion = client.chat.completions.create(
    model="revoye/auto",
    messages=[{"role": "user", "content": "Rewrite this paragraph in a friendlier tone: …"}],
)
 
print(completion.choices[0].message.content)

TypeScript

import OpenAI from "openai";
 
const client = new OpenAI({
  baseURL: `${process.env.REVOYE_API_BASE}/v1`,
  apiKey: process.env.REVOYE_KEY,
  timeout: 660_000, // milliseconds; a request can be held for up to 10 minutes
});
 
const completion = await client.chat.completions.create({
  model: "revoye/auto",
  messages: [{ role: "user", content: "Summarise this support ticket in two sentences: …" }],
});
 
console.log(completion.choices[0].message.content);

Most responses arrive in 3 to 90 seconds. Set the client timeout above the 10-minute hold so the SDK never gives up before Revoye answers.

What differs

A slow request comes back as an error with its id. If a request is still being processed after 10 minutes, the call returns 504 JOB_TIMEOUT, which the SDK raises as an API error. The request keeps running: its id is in details.job_id, and the response can be collected from GET /v1/completions/{id}. A request that fails or is cancelled is raised the same way, with the same envelope as the native endpoint.

You can accept now and collect later. Pass revoye_wait: false through extra_body and the call returns 202 straight away, with the job id in id and choices[0].message.content set to null. Collect the response by id, or use the native endpoint with a callback_url to take a webhook.

Messages are flattened. system messages come first and the other turns follow as context, all in a single request. To follow up on an earlier response, use the native endpoint with conversation_ref and mode: "continue".

stream: true returns 400. Revoye returns complete responses. Any UI code that assumes a stream needs a non-streaming path before it can talk to Revoye.

usage is null. Guard any code that does arithmetic on it — cost estimation, budget enforcement, context-window maths — because None + int is the failure you will hit first. Your token use is in the dashboard and in quota on GET /v1/status.

Desk or Cloud

To pin a request to one channel, set a default header on the client — default_headers={"Revoye-Channel": "cloud"} in Python, defaultHeaders: { 'Revoye-Channel': 'cloud' } in TypeScript — or pass revoye_channel through extra_body per request. Leave both out and your plan and routing preference decide. A channel your plan does not include is refused with CHANNEL_NOT_ENTITLED, never re-routed.

Model pickers

GET /v1/models returns the models your plan can use in OpenAI's list shape, with revoye/auto first. A picker built against an OpenAI client populates itself with no changes.

What you give up

The compatibility layer covers the everyday call. The rest of what Revoye offers lives on /v1/completions: webhooks, per-attempt timeouts, deadlines, priority, metadata, and continuing a conversation. Use the OpenAI SDK to reuse code you already have, and the native endpoint when you need those.

Using an OpenAI SDK with Revoye