Calling your model

The base URL is https://api.opensteering.com/v1 and the request's model field selects which of your models answers — its id is on the model's page in the dashboard. Authenticate with an account API key.

Python

from openai import OpenAI

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

reply = client.chat.completions.create(
    model="{model-id}",
    messages=[
        {"role": "user", "content": "A customer wants a refund after 45 days."}
    ],
)
print(reply.choices[0].message.content)

JavaScript

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.opensteering.com/v1",
  apiKey: process.env.OPENSTEERING_API_KEY,
});

const reply = await client.chat.completions.create({
  model: "{model-id}",
  messages: [
    { role: "user", content: "A customer wants a refund after 45 days." },
  ],
});
console.log(reply.choices[0].message.content);

Listing your models

GET /v1/models returns your ready models in the OpenAI list format, so client.models.list() works too — useful for discovering model ids programmatically.

No system prompt needed

Your values are learned as an internal activation direction, not injected as a policy prompt. Requests without a system message already answer under the steering intervention, and the attestation report on the model page is the measured proof. You can still send a system message for per-request context, like the customer's name or order details.

Errors