Categories
Overview
Categories let you automatically classify incoming traces and spans based on rules you define in your project settings. Once configured, every new trace is evaluated against your categories and assigned a match — giving you a structured way to filter, analyze, and monitor different types of interactions.
If a trace doesn’t match any defined category, it remains uncategorized.
Define Categories
Categories are managed in your project settings on Confident AI:
- Navigate to Project Settings > Categories
- Click New Category
- Provide a Name (e.g. “Customer Support”, “Billing Inquiry”) and a Description of what the category represents
The description is used to determine whether an incoming trace matches the category, so be specific. For example:
Once saved, all incoming traces will be automatically classified against your defined categories.
Filtering by Category
After categories are applied, you can filter traces by category in the Observatory to quickly surface specific types of interactions. This is useful for:
- Monitoring the volume of different interaction types over time
- Running targeted evaluations on specific categories
- Identifying trends in user behavior across categories
