# AI Providers

Sentinel currently uses an AI provider for one feature: [AI Security Rules](/docs/sentinel/integrations/ai-security-rules/), which evaluates requests with an LLM instead of pattern-matching rules.

## Supported providers

Set `AI_PROVIDER` to one of:

- `anthropic`
- `azure` (Azure OpenAI)
- `google`
- `mistral`
- `ollama`
- `openai`

## Configuration

```
AI_PROVIDER=mistral
AI_PROVIDER_MODEL=mistral-large-latest
AI_PROVIDER_OPTIONS={"apiKey":"abc123..."}
```

`AI_PROVIDER_OPTIONS` is JSON-encoded and accepts provider-specific options — for example, a custom base URL to point at a self-hosted or proxied endpoint:

```
AI_PROVIDER_OPTIONS={"apiKey":"abc123...","baseURL":"https://api.openai.com/v1"}
```

## Adjusting request parameters

`AI_PROVIDER_REQUEST_OPTIONS` is also JSON-encoded, and passes through parameters on the underlying API request — useful for controlling response length, randomness, or stop sequences:

```
AI_PROVIDER_REQUEST_OPTIONS={"max_tokens":300,"temperature":1,"stop":["\n\n"]}
```

## Related

- **[AI Security Rules](/docs/sentinel/integrations/ai-security-rules/)** — the feature this powers, including prompt-writing tips and self-hosted-model recommendations.
- **[ENV Variables](/docs/sentinel/operations/env-variables/#ai-providers)** — variable reference.