AI providers

The assistant uses an AI model from a provider you choose, with your own API key. This page explains how to pick one, connect it and keep the chat running if it fails.

How billing with a provider works

The plugin does not include AI usage. You create an account with a provider, generate an API key there, and the provider bills you for the messages your store sends. All of this is set up on AI Assistant → Settings → AI Provider.

Choose a provider

The provider cards on that screen are:

  • OpenAI: available on Free and Pro.
  • Anthropic, Google Gemini, DeepSeek, Mistral, Meta, xAI and Alibaba Pro
  • Custom Pro: any endpoint that speaks the OpenAI chat completions format. Enter its URL in Custom API Endpoint.

The free plugin runs on OpenAI, and you choose among all of OpenAI’s chat models. The other providers come with the Pro add-on; without it their cards are not shown. If you choose another provider while Pro is active and the subscription later ends, the chat goes back to OpenAI, and your choice returns when Pro is active again.

Enter your API key and test it

  1. Under Select Provider, click the provider you want. Provider cards have a Get API key link to that provider’s console.
  2. Paste the key in the API Key field and click Save Key. A ✓ Key saved note confirms it.
  3. Click Test Connection. The status at the top of the page changes to Connection OK, or to Last test failed with the provider’s error message.
  4. Under AI Model, click the refresh icon to load the models your key can use, pick one, then click Save Settings.

Keys are encrypted before they are saved to your WordPress database and are never printed in your store’s pages. The encryption uses your site’s WordPress secret keys, so if those change (for example after moving the site to a new server with a new wp-config.php), save the API key again.

Tune the answers

Open Advanced Settings on the same screen:

  • Temperature: lower values give more factual, repeatable answers; higher values are more creative. The screen recommends 0.5–0.9 for a support bot.
  • Max Tokens: the maximum length of a reply.
  • Request Timeout (seconds): how long to wait for the provider before giving up.
  • System Prompt: standing instructions for the assistant. You can use the variables {bot_name}, {site_name}, {store_url} and {current_date}.

The Cost Estimator at the bottom of the page projects a monthly cost from your conversation volume and compares models. Its prices are approximate; always check the provider’s own pricing page.

Set a fallback provider Pro

A fallback keeps the chat answering when your main provider has an outage. If the primary provider returns an error or an empty reply, the plugin automatically retries the same message with the fallback.

  1. Save and test an API key for the provider you want as a backup, the same way as above.
  2. Switch back to your main provider.
  3. In the Fallback Provider section, choose the backup provider, click refresh to load its models and pick a Fallback Model.
  4. Click Save Settings.

Without a fallback (and on the Free plan), a failed call shows the customer a short error message instead of an answer.

Check that it keeps working

The API Health & Usage card at the top of the screen shows the state of the primary and fallback providers, the time of the last test and how many messages the chat handled over the last 24 hours, 7 days and 30 days. For some providers it can also show your account balance.

If the assistant stops answering, start with the AI is not answering in the troubleshooting guide.

Still stuck?

Check the troubleshooting guide, or contact support and tell us what you tried.