Skip to content

Standard Skill: Improve the Semantic Layer

Improve the Semantic Layer is a Standard Skill: it adds the Verified Queries, Rules and Terms a Data Source’s Semantic Layer lacks, where they matter most. The agent loads it when you ask it to improve, extend or fill gaps in the Semantic Layer, to act on an assessment’s reasons, or to turn what Query Log Mining found into entries. To pick it yourself, type / and choose improve-the-semantic-layer.

The agent works on the first of these it has:

  1. The latest assessment. If you may accept Context Change Requests on the Data Source, the agent gets the reasons of its latest Quality score and starts from them, saying which reason each gap answers.
  2. A topic or gaps you give, such as revenue, or the gaps Query Log Mining reported.
  3. Gaps it suggests: from the entries, the tables of the Data Source and, if you may accept Context Change Requests on it, the last 90 days of its Query Log. It suggests up to five, most valuable first, and you choose which to work on.

It keeps to about ten entries per run, so that the request can be reviewed in one sitting.

  • It reads what exists first, and changes an entry that almost covers a gap rather than adding a second one. If what it would write disagrees with an entry, it asks you which is right.
  • Terms and Rules come from you, an entry or the data, never from its own idea of your business. When it inferred one from the data, it asks you to confirm it first.
  • Verified Queries are run and confirmed. The agent writes the query, runs it for one question, shows you the result as a Card or a short table, and asks whether it is right. Only on your yes does it record you as the one who confirmed it, with the date. A Verified Query never holds real rows.
  • It proposes everything in one Context Change Request, whose explanation says which gap or assessment reason each entry closes.