Overview
Filters
Multi-type filter controls for refining AI output or input
filterssearchrefinement
Docs
Filters let users narrow AI-generated results or refine input parameters through structured controls like checkboxes, radio buttons, sliders, and dropdowns. They bring precision to broad queries by letting users specify exactly what they want without rewriting their prompt. Filters are especially valuable when AI produces large result sets or when users need to apply domain-specific constraints.
Variants
- Sidebar filters — A vertical panel of filter groups, common in search and browse interfaces where screen space allows a persistent filter panel.
- Horizontal filters — Compact filter controls arranged in a row, suitable for simple filtering needs or toolbar integration.
- Popover filters — Filters hidden behind a button that reveals a popover panel, conserving space while providing full filter capabilities on demand.
Use Cases
- Search result refinement by content type, quality, or date
- AI output filtering by confidence score or source
- Dataset exploration with multi-dimensional constraints
- Content moderation tools filtering by category or severity
- Model output comparison filtered by performance metrics
Best Practices
- Show result counts next to filter options so users can gauge the impact of each filter.
- Provide a 'Clear all' action for quick filter reset.
- Preserve filter state across pagination and sorting changes.
- Use the right control type for each filter: checkboxes for multi-select, radios for single-select, sliders for ranges.
- Update results immediately as filters change rather than requiring a separate 'Apply' action.