Ctrlk
For the complete documentation index, see llms.txt. This page is also available as Markdown.

Copy of Embed an AI Chat

Self-service AI Chat lets embedded users explore data using natural language, without accessing Toucan AI platform.

The chat is embedded in your product and governed by your data, security, and configuration rules.

What the feature does

This feature allows you to:

  • Embed an AI-powered chat inside your product

  • Let Explorers ask questions in natural language

  • Return tables and charts generated from your data

  • Restrict answers using datasets, and RLS

The embedded AI chat is designed for data exploration, not dashboard building.

Prerequisites

Before enabling self-service AI chat, you must have:

  • At least one connected database

  • At least one generated API key

  • At least one generated access token

  • (Recommended) Dataset metadata analyzed or enriched

  • (Recommended) Row-Level Security configured

How to use it

Step 1 – Prepare your data

  • Analyze datasets with AI

  • Review and adjust column descriptions

  • Ensure metric definitions are explicit

Good metadata directly impacts answer quality.

Step 2 – Set up security

  • Define custom token attributes

  • Map attributes to dataset columns using RLS

Always validate with multiple attribute values.

Step 3 – Configure embed token

When generating a token:

  • Include required attributes

  • Set token expiration

  • Enable AI chat capability

Tokens fully define what the Explorer can see and do.

Step 4 – Embed the AI chat

Embed the chat using Web component. The chat runs fully inside your product UI.

Example

Use case: HR SaaS with multi-tenant customers

  • Attribute: customer_id

  • Dataset column: customer_id

  • Explorer question: &#xNAN;“How many hires last quarter?”

Result:

  • Query is filtered by customer_id

  • Explorer only sees their own data

  • No configuration or SQL required

Last updated

Was this helpful?