Knowledge Datasets

Datasets are the collection's tables — its knowledge in rows and columns beside the prose. Each one is a card, and opening it gives a spreadsheet grid: add rows, reshape the schema, import and export in one place.

At a glance

  • Cards with a health badge — Ready, No identifier or Empty
  • A spreadsheet grid that loads big tables a page at a time
  • A schema dialog that reorders, renames and retypes columns
  • A Score panel with Run score and AI fix on entry-shaped datasets
Knowledge Datasets

The dataset cards

The tab is a three-column grid of cards, each clickable whole: the name, a health badge, the description, the row count, column chips with identifier columns accented, and the updated date. Health is one of three words — Ready, or the dataset has no identifier column, or it is empty — so what needs attention reads from the landing grid.

The grid

The editor is a spreadsheet under a rows × columns badge: a frozen header, resizable sortable columns, inline cells and a per-row delete. The toolbar holds Add row, Add column, Import rows, Export CSV, Edit schema, undo and redo, an unsaved count and search, and it always says how much is loaded — 12,400 of 100,000. Data arrives a page at a time, 200 rows per request, and sort and search run over what is loaded so far.

Score and schema

Entry-shaped datasets carry a Score panel — the percentage, passed of tested, the timestamp and a reason per failure — with Run score and AI fix beside them. The schema dialog reorders, renames and retypes columns, takes synonyms, and toggles Identifier and Searchable per column; a schema save remounts the grid. Editing rights gate the grid and the schema dialog together, and every write goes through one route that rebuilds the search text on the server.

  1. Open a datasetEvery card is clickable whole — the name, a health badge, the description, the row count and its column chips, identifier columns accented — and opens the grid.
  2. Work the gridAdd row, Add column, Import rows and Export CSV sit in the toolbar with undo and redo; cells edit inline, every column sorts and resizes, and each row carries its own delete.
  3. Shape the schemaEdit schema opens the dialog: reorder, rename and retype the columns, add synonyms, and toggle Identifier and Searchable; saving remounts the grid.

Good to know — Sort and search cover what is loaded, not the whole table — the toolbar's count says exactly how much that is.