Skip to content

Public access and rate limits

Control who can access your AI Search features and set rate limits to manage usage. These settings determine how the chat widget and search page behave for your visitors.

Configure public access to your AI Search features from the Search settings page in the dashboard.

Setting Default Description
enabled false Enable public access to AI Search
allowedDomains [] Domains permitted to embed search widgets
searchEnabled true Allow search queries
chatEnabled true Allow chat conversations
anonymousSessions true Allow unauthenticated users
sessionTTLMinutes 30 Session expiry time in minutes (5 to 1440)

Public access is off for a new site. Until you enable it, the chat widget, the search page and the public API return 403.

Rate limits protect your AI Search usage and prevent abuse. Configure these values to balance availability with cost control.

Setting Default Description
searchPerMinute 100 Maximum search queries per minute (up to 1000)
chatPerMinute 20 Maximum chat messages per minute (up to 100)
perIpPerMinute 30 Maximum requests per IP address per minute (up to 100)

When a rate limit is exceeded, the widget or search page displays a message asking the visitor to try again shortly. Rate limits reset every minute.

Restrict which domains can embed the chat widget and search page by adding your production domain(s) to the allowedDomains list. This prevents unauthorised usage of your AI Search on third-party sites.

For example, to allow only your production site:

{
"allowedDomains": ["www.example.com", "example.com"]
}

If the allowedDomains list is empty, requests from any domain are accepted when public access is enabled. In the dashboard, enter one domain per line; * is a wildcard, for example *.example.com.

The Display URL setting rewrites the domain of result and source links. Use it when content was indexed from a different host, such as a preview or internal URL, so that links go to your real domain. Leave it blank to use the indexed URLs unchanged.

Every question is screened for prompt attacks and harmful content, and personal details in a question are masked before the AI sees them. Changing the level needs the AI governance permission.