Semantic layer
A semantic model turns database tables into something your people and your AI understand: friendly names, clear descriptions, the right data types and only the columns that matter. Define it once, and every dashboard, extension and conversation uses it.
In your database
In your semantic model
_etl_loaded_at hidden
One model, used everywhere
The context you add to a model goes wherever the data goes, so a field means the same thing in every dashboard, every extension and every answer.
Dashboards, KPI cards and spreadsheets show your friendly field names and respect the data types you set.
SuperTables, FinanceTables and the other extensions in Analytics read the same fields, with the same names.
Vizzy picks the right model for a question and reads your comments to understand what every field means.
Two ways to start
Model a table straight from one of your database connections, or reuse the models your analysts already maintain in Tableau.
Connection-backed model
Pick a connection, drag a table onto the model canvas and give it the context it needs. Run the semantic profile and Trulyp suggests what every field means.
Semantic provider
Connect Tableau Cloud or Tableau Server once, then import published data sources with all their tables, joins and relationships. Trulyp only reads; nothing is copied.
From Tableau to Trulyp
Pick a published data source and Trulyp brings in its tables, relationships and calculated fields as a semantic model. Nothing is rebuilt and nothing is copied.
In Tableau
In Trulyp
4 tables · 3 relationships · read-only
Context for AI
Vizzy doesn't guess what a column means. It reads your semantic model and turns your question into a query on it, and AI Trace records every step.
Vizzy works out what you are asking.
It picks the semantic model that fits the question.
Your field names and comments tell it what each column means.
It runs a query on the model, not a guess.
You get the answer, with every step recorded in AI Trace.
Semantic modeling
Everything you add to a model is used by Analytics, by SuperTables and other extensions, and by Vizzy AI.
Give tables and columns an alias and a description. Analytics shows the names, Vizzy AI reads the comments.
Trulyp samples your data, tells dimensions from measures, explains what each field likely means and fills in missing comments.
Treat postcodes, IDs and codes as Text so nobody adds them up, without changing anything in your database.
Hide technical and load columns from Analytics, extensions and AI, and keep the model clean.
Check the first rows of every table while you model, so you know exactly what people will see.
Export a model with its aliases, comments and settings, and import it in another environment. Credentials never travel along.