Feature

OpenAI Compatibility

Aniron's API matches the OpenAI request and response shape field for field. Point your existing SDK at a new base URL and key, and everything you already built — streaming, tool calls, retries — keeps working against 20+ models.

How it works

Point the SDK at Aniron

Change the base URL to api.aniron.ai and swap in an Aniron key. The official OpenAI client libraries need no other configuration.

Pick a model string

Requests use a model identifier like any OpenAI call. Aniron resolves it to one of 20+ open-source or frontier models behind the same endpoint.

Response shape is unchanged

Fields, error codes, and streaming chunks match the OpenAI API, so parsing, retry, and logging code you already wrote does not change.

client.ts

import OpenAI from 'openai';

const client = new OpenAI({
  baseURL: 'https://api.aniron.ai/v1',
  apiKey: process.env.ANIRON_API_KEY,
});

// Same call, same response shape. Only the model
// string and base URL changed.
const completion = await client.chat.completions.create({
  model: 'aniron/llama-3.3-70b',
  messages: [{ role: 'user', content: 'Explain TCP.' }],
});

2 lines Typical migration diff

Most integrations only change the base URL and the API key. The rest of the request pipeline is untouched.

When compatibility matters

Migrating an existing product

Move an OpenAI-only integration to a multi-model gateway without rewriting the request-building or response-parsing code around it.

Testing models against a production integration

Swap the model string in a staging environment and compare cost, latency, and output quality without touching application code.

Using OpenAI tooling and frameworks

Libraries and agent frameworks built against the OpenAI client work unmodified, since they only ever depend on the API shape, not the provider.

Avoiding a rewrite for fallback logic

Add model fallbacks or rate-limit handling through Aniron features instead of hand-rolling retry logic across multiple provider SDKs.

Frequently asked questions

Do I need a different SDK to use Aniron?

No. The official OpenAI SDKs for every language work unmodified. Point the client at the Aniron base URL and use an Aniron API key, and every existing call keeps working.

Is the response shape identical to the OpenAI API?

Yes. Chat completions, streaming, tool calls, and embeddings all return the same fields your code already parses. Nothing downstream of the API call needs to change.

Can I use non-OpenAI models through the same interface?

Yes. The compatibility layer covers 20+ open-source and frontier models from multiple providers. You select a model with a model string, not a different API shape.

Do streaming responses work the same way?

Yes. Server-sent event streaming, including delta chunks and the final usage payload, matches the OpenAI streaming format exactly.

What is not compatible?

Account and billing endpoints, such as balance and usage queries, are Aniron-specific since they do not exist in the OpenAI API. Every model-inference endpoint follows the OpenAI shape.

Related resources