Privacy

Security Settings

Protect your widget from abuse, runaway credit usage and instructions hidden in retrieved context.

Behaviour

Request limits

Cap how many messages a single visitor can send within a time window. Once the limit is hit, the visitor sees a friendly error instead of reaching the assistant. This shields your system from traffic spikes and keeps automated abuse from draining your credits.

  • 01
    Maximum number of requests: how many messages are allowed per window.
  • 02
    Time interval in seconds: the rolling window the limit is counted over.
  • 03
    Exceeded-limit behavior: further messages are blocked until the window resets.
  • 04
    Custom error message: the text shown to a visitor who hits the limit.
Setup

Allowed domains

Restrict which websites are allowed to load your widget and talk to your assistant. This stops other sites from embedding your snippet and spending your credits.

  • Leave the list empty to allow every domain (the default).
  • Add one domain per line, for example example.com.
  • Subdomains are covered automatically (example.com also allows shop.example.com).
  • You can add up to 20 domains.
Privacy

Context data protection

Append a global system instruction after your custom prompt to keep retrieved content in the data lane. Visitor messages are not changed.

  • Protection is enabled by default. Breezaro appends a short safety instruction after your custom prompt.
  • The instruction marks retrieved knowledge, page data and quoted messages as data rather than commands, and tells the assistant not to reveal their raw contents.
  • Turning it off does not alter visitor messages or remove Breezaro's other system instructions. The setting applies to the widget and every connected messaging channel.