> ## Documentation Index
> Fetch the complete documentation index at: https://docs.datafold.com/llms.txt
> Use this file to discover all available pages before exploring further.

# OpenAI-compatible

The universal provider for any service that speaks the OpenAI `/chat/completions` protocol: Together, Fireworks, Groq, OpenRouter, Mistral, DeepInfra, Cerebras, and self-hosted runtimes like vLLM, LM Studio, or Ollama.

## Configure in Datafold

1. Navigate to **Settings** → **Integrations**, click **Add New Integration**, and choose **OpenAI-compatible**.
2. Fill in:

| Field              | Description                                                                                                                                           |
| ------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Base URL**       | The provider's OpenAI-compatible base, e.g. `https://api.mistral.ai/v1` or `https://api.together.xyz/v1`.                                             |
| **API Key**        | The provider's API key. Sent as `Authorization: Bearer` unless a custom header is set.                                                                |
| **API Key Header** | Optional. Only needed for providers that expect the key in a non-standard header.                                                                     |
| **Models**         | Fetched automatically from the provider's `/models` endpoint once the Base URL and key are entered. Select the models to expose, or add ids manually. |

3. Click **Save**.

<Info>
  Reasoning controls: leave the **Thinking** selector off for models that don't accept OpenAI's `reasoning_effort` parameter (e.g. Mistral's Magistral family reasons implicitly and rejects unknown parameters).
</Info>