Developer documentation

OpenAI-compatible integration

Use your Varion API key as the client API key and replace the OpenAI base URL.

Python

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_VARION_KEY",
    base_url="https://api.varion.tech/v1",
)

response = client.chat.completions.create(
    model="gpt-5-mini",
    messages=[{"role":"user","content":"Hello"}],
)

Node.js

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "YOUR_VARION_KEY",
  baseURL: "https://api.varion.tech/v1",
});

Supported endpoints

Endpoint	Status
`POST /v1/chat/completions`	Supported
`POST /v1/responses`	Supported
`GET /v1/models`	Supported

Optional headers

X-Varion-Optimize: off
X-Varion-Cache: off
X-Upstream-API-Key: sk-...

Developer downloads

Python helper Node.js helper Postman collection

Local tests call Varion’s optimise-only endpoint for the exact optimized request. Provider keys stay on the customer’s computer and are sent only to the selected AI provider.

Verified paired benchmark

For publishable measurements, send the identical original request twice with the same benchmark ID. First send X-Varion-Benchmark-Phase: baseline, X-Varion-Optimize: off and X-Varion-Cache: off. Then send phase optimized. Read X-Varion-Verified-Saved-Tokens, X-Varion-Verified-Savings-Percent and X-Varion-Savings-Verified.

True passthrough always reports zero saved tokens. Estimated and verified savings are reporting metrics only. Billing is based on successful customer input tokens processed.