AI Providers

Wagtail AI uses django-ai-core and the any-llm library to support multiple AI providers and models. This unified interface allows integrations with OpenAI, Anthropic Claude, local models, and many other AI services.

Note

Providers will replace the previous concept of AI Backends introduced in Wagtail AI 2.x. If you're upgrading from Wagtail AI 2.x, see the migration instructions.

Quick start

Configure AI providers in your Django settings using the WAGTAIL_AI configuration:

WAGTAIL_AI = {
    "PROVIDERS": {
        "default": {
            "provider": "openai",
            "model": "gpt-4.1-mini",
        },
    },
}

Multiple providers

The any-llm library supports a wide range of providers. Apart from model, the dictionary of a provider is passed as keyword arguments to any-llm's AnyLLM.create().

For example, the following configuration has two different providers: a default provider for general text tasks, and a vision provider for image-related features using a vision-capable model. Both providers are hosted by different services by passing custom api_key and api_base values.

WAGTAIL_AI = {
    "PROVIDERS": {
        "default": {
            "provider": "openai",
            "model": "gpt-oss-120b",
            "api_key": "my-scaleway-key",
            "api_base": "https://api.scaleway.ai/my-project-id/v1",
        },
        "vision": {
            "provider": "mistral",
            "model": "mistral-small-3.2-24b-instruct-2506",
            "api_key": "my-own-secret-key",
            "api_base": "https://my-llm-host.com/v1",
        },
    },
    "IMAGE_DESCRIPTION_PROVIDER": "vision",  # Use vision model for images
}

Note

You may have to install additional dependencies for some providers. See the any-llm documentation and the individual provider documentation for details.