KG-Registry

Features

• Registry of knowledge graphs and related resources
• YAML-based data format for easy editing and collaboration
• Web interface for browsing resources
• Advanced search using SQL queries via DuckDB
• Parquet file support for efficient storage and querying
• CLI tools for data management and export

Documentation

KG-Registry Documentation: Start here to learn about the purpose of KG-Registry and how it works.

Entries in KG-Registry are either Resources or Products. Resources represent top-level entries for knowledge graph and data sources. Products are specific representations or interfaces for a resource (e.g., graph dumps, APIs, or visualizations). This also includes associated software and documentation sets.

For more detail:

• Resources
• Products
• Agent-Assisted Curation
• Parquet Backend & Advanced Search

Site Development

For developers working on the KG-Registry website and backend:

• Site Development Guide - Setup, deployment, and troubleshooting
• Agent Skills Guide - Local agent-assisted curation workflows in .agents/
• Parquet Backend Documentation - Database schema and querying
• Advanced Search Feature - SQL-based resource discovery interface

Reference Validation

Resource publication metadata is validated against cache files produced by linkml-reference-validator. Cached title, year, DOI, and first-author mismatches are validation errors; missing cache files are warnings during ordinary validation.

uv run make validate-file FILE=resource/<resource>/<resource>.md
uv run make cache-publication-references
uv run make validate-publication-reference-cache

See Reference Validation for the full procedure. The data quality dashboard also reports resources with missing publications or publication citation metadata issues.