Knowledge Unified Katalog And Network
A Data Catalog for Everyone
The modern, open-source alternative to CKAN
Group a long resource list under headings. In the dashboard a heading is a divider you drag into place. A "/" in the name nests headings up to three levels, and section names count in search and over MCP.
Replacing a file keeps the old content as a version. Download any past version; for CSV/TSV, see added and removed rows — or, with a primary key, how each row changed (enhanced in v0.15.0). Individual versions can be deleted when content itself has to go.
AI drafts titles, descriptions, and tags from what is actually in your files — CSV columns and sample rows, even the text inside PDF and Office documents. Review and adopt field by field; nothing is overwritten automatically.
Run multiple data catalog sites on one shared infrastructure. Databases, search indices, and storage stay isolated per site while the costly backbone is shared — keeping each additional site inexpensive.
Event-driven architecture handles large datasets with ease — file processing and indexing run in the background. A unified TypeScript stack from API to frontend.
Deploy on AWS or with Docker Compose in minutes, with AWS CDK IaC and an automated deploy pipeline included. Hybrid design supports on-prem and air-gapped networks.
Semantic search with natural-language queries, plus full-text search across metadata, CSV files, and text files. Native Japanese morphological analysis via kuromoji.
Preview CSV, JSON, GeoJSON, PDF, XLSX, and more. GeoJSON is rendered on maps with GSI and OpenStreetMap; CSV/TSV tables sort, filter and search in place, with SQL queries via the Data API.
OIDC / SAML integration with existing identity providers. Role-based access control across organizations and groups keeps public and private data under control.
Drop-in compatibility with major CKAN read APIs — call package_search, package_show, and more as-is. Works with existing CKAN tools and connects catalogs together.
Broken-link detection and metadata validation built in to keep your catalog healthy over time. Reduces the operational burden of reliable data publishing.
Built-in MCP server lets AI agents search and browse your catalog — semantic search finds datasets from natural language, and SQL queries reach the data itself.
AGPL-3.0 licensed with the full source open. Japanese and English UI out of the box, and brand customization tailors colors, logo, and wording to your organization.
