feat(kb-tags): natural language pre-filter tag system for knowledge base searches

[icecrasher321]

Description

Can set natural language tag names for knowledge base records to pre-filter before vector search.

System design:

• Natural language tags latch onto an actual column "tag slot" in the docs / embeddings table.
• Caps at 7 per KB for now but can add more columns
• Works through KB Block + Doc page
• Can only search 1 KB at once as a result. For multi KB search use -- multiple KB blocks with search tool

Type of Change

• New feature (non-breaking change which adds functionality)

How Has This Been Tested?

Test adding, editing tags in KB Block [Create Document Tool]. Search using filters in search tool.

Checklist:

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have added tests that prove my fix is effective or that my feature works
• All tests pass locally and in CI (bun run test)
• My changes generate no new warnings
• Any dependent changes have been merged and published in downstream modules
• I have updated version numbers as needed (if needed)
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

Security Considerations:

• My changes do not introduce any new security vulnerabilities
• I have considered the security implications of my changes

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
sim	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Jul 29, 2025 1:51am

1 Skipped Deployment

Name	Status	Preview	Comments	Updated (UTC)
docs	⬜️ Skipped (Inspect)			Jul 29, 2025 1:51am

[greptile-apps]

Greptile Summary

This PR implements a comprehensive natural language tagging system for knowledge base documents that enables pre-filtering before vector search operations. The system allows users to create custom, human-readable tag names (like 'Department' or 'Priority') instead of generic numbered labels, while maintaining efficient search performance through actual database columns.

The core architecture maps natural language tag names to fixed database columns (tag1-tag7) through a new knowledge_base_tag_definitions table. This design choice provides good performance for filtering operations while capping the system at 7 tags per knowledge base. The implementation includes:

• Database Layer: New migration adds knowledge_base_tag_definitions table and indexes tag columns on both document and embedding tables for efficient filtering
• API Layer: New endpoints for managing tag definitions at the knowledge base level, with proper CRUD operations and cleanup utilities
• Frontend Components: New React components (KnowledgeTagFilters, DocumentTagEntry, KnowledgeTagFilter) that integrate with the existing sub-block workflow system
• Block Integration: Updated Knowledge block to use the new tag system, replacing individual tag1-tag7 inputs with dynamic tag management
• Search Enhancement: Modified search functionality to support OR logic within tag groups and restrict to single knowledge base searches

The system maintains backward compatibility with existing tag workflows while providing a more intuitive user experience. Users can now define meaningful tag categories and filter knowledge base content using natural language terms before expensive vector similarity searches.

Confidence score: 4/5

• This PR introduces significant new functionality with proper architectural design and maintains backward compatibility
• The implementation follows established patterns and includes comprehensive error handling and validation
• Potential concerns include some type assertions, magic number usage for ID generation, and the restriction to single knowledge base searches which may impact existing workflows

_{27 files reviewed, 15 comments}

_{Edit Code Review Bot Settings | Greptile}