Standard Skill: Schema Bootstrap
Schema Bootstrap is a Standard Skill: it drafts a first Data Model of a Data Source that has none, so that it gets one in minutes instead of hours. The agent loads it when you ask it to model, describe or document a new database, or to get the Semantic Layer of a Data Source started. To pick it yourself, type / and choose schema-bootstrap.
What the agent does
Section titled “What the agent does”- It checks that there is no Data Model yet. A Data Source has at most one. If it has one, the agent updates it with Model a Data Source instead.
- It reads the schema: every table, with its columns and types. Every name in the model comes from there, never from a guess. With more than about 30 tables, it models the 30 that matter most and says which it left out. Technical tables, such as migrations and logs, stay out unless you name them.
- It checks keys and joins with queries. Before it declares a key, it counts whether it is unique; before it declares a join, it counts the rows that find no partner. These queries only count: no value from your data goes into the model.
- It writes the model as an Apache Ossie model: a dataset per table, the columns that matter, keys, relationships and the plain measures, such as a count of orders. Where it can’t tell what a column means, it notes an open question rather than a guess.
- It proposes it. You see a short summary and the open questions, then the Approval of a Context Change Request. Its explanation lists the open questions, so the reviewer sees them.
Afterwards it offers Interview to settle the open questions with you.
What you need
Section titled “What you need”You need access to the Data Source to have it inspected and queried, and the Right to create Context Change Requests on it to propose the model. Without that Right, the agent tells you who could propose it. Nothing changes in the Semantic Layer until someone who may accept Context Change Requests on the Data Source accepts the request.