offdata ai — agentic AI for data modelers and data engineers

Model Builder

Drag-and-drop data modeling on a full-screen canvas

Build Kimball, Data Vault 2.0 or 3NF models by hand. Drop building blocks onto the canvas, define typed columns, draw relationships and let auto-arrange tidy the layout. Free on every plan.

model-builder · export.jsonjson
{
  "schema_version": "builder-0.1",
  "entities": [
    {
      "id": "n1", "name": "dim_customer", "kind": "dimension",
      "columns": [
        { "id": "c1", "name": "customer_sk", "type": "bigint",
          "is_pk": true, "nullable": false },
        { "id": "c2", "name": "customer_id", "type": "varchar",
          "is_business_key": true },
        { "id": "c3", "name": "email", "type": "varchar",
          "pii": true }
      ]
    },
    { "id": "n2", "name": "fct_invoice", "kind": "fact" }
  ],
  "relationships": [
    { "id": "r1", "from_entity_id": "n1",
      "to_entity_id": "n2", "cardinality": "one_to_many",
      "fk_column": "customer_sk" }
  ]
}

Not every model should start with a conversation. When you know exactly what you want, drawing it is faster than describing it. The Model Builder gives you a canvas with the building blocks data modelers actually use: facts and dimensions, hubs, links and satellites, bridges, lookups and reference tables.

Open the canvas in full screen for large models and press Arrange to lay everything out cleanly. Every change saves automatically. Export the model as JSON or a PNG for documentation, or use it as the source of an ontology so your business vocabulary starts from the design you drew.

What the Model Builder gives you

  • Building blocks for three methodologies

    Dimension, fact, factless fact, junk and date dimensions; hub, link, satellite, effectivity satellite and PIT; tables, lookups and associative tables.

  • Typed columns and flags

    13 column types, with primary key, business key, PII and nullable flags on each column.

  • Relationships with cardinality

    Draw edges between tables and set 1:1, 1:N or M:N cardinality and the foreign key column.

  • Full screen and auto-arrange

    Work on large models in full screen, collapse the sidebar, and lay out the whole diagram in one click.

  • Autosave and export

    Changes save as you go. Export JSON for version control or a PNG for docs and reviews.

  • Feeds your ontology

    Use a saved model as an ontology source, so concepts and relations come from the design you drew.

Frequently asked questions

What is the OffDataAI Model Builder?
A visual, drag-and-drop canvas for designing a data model by hand. You add building blocks, define columns and draw relationships, then use the model in the rest of OffDataAI, for example as the source of an ontology.
Which modeling styles does it support?
Three groups of building blocks. Dimensional (Kimball): dimension, date dimension, junk dimension, fact, factless fact and bridge. Data Vault 2.0: hub, link, satellite, effectivity satellite, point-in-time and reference tables. Relational (3NF): table, associative table, lookup and reference table.
What can I set on each column?
A name and one of 13 types (varchar, text, integer, bigint, smallint, numeric, decimal, float, boolean, date, timestamp, json, uuid), plus primary key, business key, PII and nullable flags.
Is the Model Builder free?
Yes. It is included on every plan, including Free, and it uses no AI credits.
Does it save my work?
Yes, automatically. There is no save button to forget; models appear in your list where you can reopen, rename or delete them. You can also export a model as JSON or as a PNG image.
When should I use the Model Builder instead of the AI interview?
Use the Model Builder when you already know the design and want to draw it precisely. Use the AI interview when you want OffDataAI to propose the design from a description of your business, or start from one of 174 templates.

Your data warehouse is one conversation away.

Describe your domain, or open one of 170+ production-grade templates. ERDs, DDL, and a complete dbt project — generated in under a minute.