Article Overview

This support article explains how to train a Conversation AI bot using web URLs. The guide covers three crawling modes and URL management practices.

Key Training Methods

Exact URL: The recommended approach where the bot will crawl the exact URL provided and train itself. This method offers precision for targeted training.

All URLs in this Domain: This broader option allows the bot to crawl an entire domain and presents users with a list to select relevant pages for training.

All URLs with this Path: Similar to the domain option, but filters results based on specific path patterns found in page URLs.

Important Constraints

The system supports training on up to 4,000 web URLs in a single Knowledge Base. Users should wait for all URLs to complete processing before deploying the bot, as URLs are not instantly added to the Uploaded Links table.

URL Status Categories

Trained URLs display one of three statuses:

• Getting Data - Currently retraining on updated information
• Trained - Successfully learned with refresh timestamp visible
• Failed - Training unsuccessful; can be refreshed or removed

Performance Recommendation

The article advises that maintaining concise & relevant data significantly boosts the bot's performance, suggesting regular reviews and removal of outdated URLs.