Wagtail Localize AI Translator

A machine translator for Wagtail Localize that uses Any-LLM for translation

Prerequisites

• Python 3.11+
• a Wagtail project with Wagtail Localize correctly configured

Dependencies

• Wagtail Localize
• any-llm-sdk

Installation

Install the package using pip:

pip install wagtail-localize-ai

In your settings.py file,

• Add wagtail_localize_ai to your INSTALLED_APPS
• Add wagtail.contrib.settings to your INSTALLED_APPS (used to setup model and prompt)
• Setup WAGTAILLOCALIZE_MACHINE_TRANSLATOR like this:

  WAGTAILLOCALIZE_MACHINE_TRANSLATOR = {
      "CLASS": "wagtail_localize_ai.translator.AITranslator",
  }
  

Then, run python manage.py migrate wagtail_localize_ai to create the required database tables

Setting up providers

To set up providers add to your settings.py file a dict called AI_PROVIDERS.
Each entry is an arbitrary key identifying the provider instance, with a dict of configuration.

Reserved keys (all start with _):

• _name: display name shown in the Wagtail admin
• _provider: any-llm provider name (e.g. openai, anthropic). Defaults to the entry key, so you can omit it when the key matches the provider name.

All other keys are passed as kwargs to AnyLLM.create(provider, **kwargs).

You can find supported providers and their kwargs in the any-llm documentation.

Having multiple instances of the same provider (e.g. different API keys or endpoints) is supported:

AI_PROVIDERS = {
    "openai_main": {
        "_name": "OpenAI (main)",
        "_provider": "openai",
        "api_key": "sk-...",
    },
    "openai_secondary": {
        "_name": "OpenAI (secondary)",
        "_provider": "openai",
        "api_key": "sk-other-...",
        "api_base": "https://my-proxy.example.com/v1",
    },
}

Here are some examples for the most popular providers:

OpenAI

AI_PROVIDERS = {
    "openai": {
        "_name": "OpenAI",
        "api_key": "sk-...",
        "organization": "org-...",  # Optional
        "api_base": "https://api.openai.com/v1",  # Optional
    }
}

Anthropic

AI_PROVIDERS = {
    "anthropic": {
        "_name": "Anthropic",
        "api_key": "sk-ant-...",
    }
}

Azure OpenAI

Provider key: azureopenai (OpenAI-compatible Azure endpoint).

AI_PROVIDERS = {
    "azureopenai": {
        "_name": "Azure OpenAI",
        "api_key": "...",
        "api_base": "https://<resource>.openai.azure.com",
        "api_version": "preview",  # Optional, defaults to "preview"
    }
}

Azure AI

Provider key: azure (Azure AI Inference SDK, for models deployed on Azure AI Foundry).

AI_PROVIDERS = {
    "azure": {
        "_name": "Azure AI",
        "api_key": "...",
        "api_base": "https://<model-deployment-name>.<region>.models.ai.azure.com",
    }
}

Gemini

AI_PROVIDERS = {
    "gemini": {
        "_name": "Gemini",
        "api_key": "...",
    }
}

Vertex AI

Provider key: vertexai. project, location and credentials are passed directly to the Google GenAI client.

Using Application Default Credentials (e.g. gcloud auth application-default login):

AI_PROVIDERS = {
    "vertexai": {
        "_name": "Vertex AI",
        "project": "my-gcp-project",
        "location": "europe-west8",
    }
}

Using a service account JSON file:

from google.oauth2 import service_account

AI_PROVIDERS = {
    "vertexai": {
        "_name": "Vertex AI",
        "project": "my-gcp-project",
        "location": "europe-west8",
        "credentials": service_account.Credentials.from_service_account_file(
            "/path/to/service_account.json",
            scopes=["https://www.googleapis.com/auth/cloud-platform"],
        ),
    }
}

Usage

You can set the provider, model and prompt from the AI Translator page reachable from the Settings menu entry.
You can also see the token usage from the Logs pagea in the Reports menu entry.

License

This project is released under the BSD license.

Contributors