Skip to main content

Overview

Pass arbitrary JSON metadata when initializing the widget to:
  • Personalize the AI assistant using Data-Driven Rules
  • Track users across your CRM systems via lead webhooks

Basic Usage

Hot Reload

Update metadata when users log in or state changes:
Chat history is preserved when updating metadata mid-session. The AI will adapt its behavior based on the new values.

Data-Driven Rules

Custom metadata powers Data-Driven Rules - a feature that lets you customize the AI assistant’s behavior based on visitor context.

How It Works

  1. Pass metadata fields from your website (e.g., isAuthenticated, customerType)
  2. Configure rules in the AI Playbook dashboard
  3. The AI automatically adapts its responses based on the metadata values

Predefined Fields

These fields have built-in example instructions in the dashboard:

Custom Fields

You can use any field name you like. Configure a “Custom field” rule in the AI Playbook to use it.

Example: Authentication Prompt

Rule configuration:
  • Field: isAuthenticated
  • Instructions: “If false, after they’ve sent 3 messages, politely suggest logging in to save their conversation. If true, don’t mention logging in.”
Website integration:

Webhook Integration

Metadata also appears in lead.captured.v1 webhooks for CRM integration:

Limitations

Sensitive keys are automatically stripped server-side. Never send actual passwords or tokens.