offdata ai — agentic AI for data modelers and data engineers

Free erwin converter

Convert an erwin model for Snowflake, Databricks or BigQuery

Upload an erwin Data Modeler XML export. In seconds you get a health report an architect would write and your model converted for a modern platform, with keys, relationships, subject areas and definitions intact.

Analyse a modelBook a demo
erwin-converter · reportsnowflake
erwin model   Retail Orders · erwin XML r9
read          8 tables · 40 columns · 6 relationships

health        52 / 100 · grade D
  high    personal data with no classification     × 5
  high    columns that look like FKs, no relation  × 2
  medium  customer_id: bigint vs varchar(20)        × 1
  medium  ORDER is a reserved word → orders

convert       snowflake
  NUMBER(12,2)  → NUMBER(12,2)
  VARCHAR2(30)  → VARCHAR(30)
  MONEY         → NUMBER(19,4)
  supplier ↔ product  → supplier_product (bridge)

Analyse an erwin model

In erwin Data Modeler: File → Save As → XML. Upload the XML here. It is read in memory and never stored.

Drop your erwin XML here, or click to chooseerwin r7 and later, and ERwin 4 XML · up to 10 MB

Years of design work sit in erwin models: entities, keys, relationships, definitions. Moving that work to a cloud warehouse usually means re-keying it by hand or forward-engineering Oracle-flavoured DDL and fixing it table by table.

This converter reads erwin's XML export directly. It keeps the model's structure and documentation, converts every physical type for the platform you choose, resolves logical many-to-many relationships, renames reserved words, and reviews the model the way a principal data architect would, so you know what to fix before it reaches production.

What the converter does

  • Reads erwin XML, both dialects

    erwin r7 and later, and ERwin 4 XML. Entities, attributes, physical types, null options, PK and AK key groups, relationships, subject areas, domains and definitions.

  • Converts types, keeps sizes

    Oracle, SQL Server, DB2 and Teradata types become their Snowflake, Databricks, BigQuery, Postgres, Redshift, Synapse or Fabric equivalents, with lengths and precision kept.

  • Architect review, free

    Missing keys, undeclared joins, type drift between tables, unclassified personal data, undocumented tables and reserved words, ranked by severity with a score.

  • Fixes what would break

    Logical many-to-many relationships become bridge tables, reserved words are renamed, and foreign keys are ordered so the DDL runs first time.

  • Private by design

    The file is parsed in memory with hardened XML handling and discarded. No AI model sees it, and nothing is stored unless you import it.

  • Then modernise it

    Import into OffDataAI for full DDL, dbt, docs, governance and an ontology, or have AI remodel it as a star schema or Data Vault.

Frequently asked questions

How do I export a model from erwin Data Modeler?
Open the model in erwin Data Modeler and choose File → Save As, then pick XML as the file type. Upload that .xml file here. The native .erwin file is a proprietary binary format that can only be read by erwin itself.
Which erwin versions are supported?
The XML export from erwin Data Modeler r7 and later (the erwin.com repository format), and the older ERwin 4 XML format that many other tools still export. Logical and physical names, physical data types, null options, primary and alternate keys, relationships and the columns they migrate, subject areas, domains and definitions are read.
Is my model stored or used to train AI?
No. The converter reads the file in memory, returns the report and discards it. Nothing is saved, and no AI model sees it: the conversion and the health report are deterministic. If you choose to import the model into OffDataAI, it is stored in your private workspace.
What does the health report check?
Tables without a primary key, surrogate keys without a business key, columns that look like foreign keys but have no relationship, the same column typed differently in different tables, columns with no data type, types that don't exist on your target platform, personal data with no classification, undocumented tables and columns, SQL reserved words, orphan tables and mixed naming conventions.
How are data types converted?
Each physical type is normalised and then rendered for the platform you choose, keeping lengths and precision: VARCHAR2(30) becomes VARCHAR(30) on Snowflake and STRING on BigQuery, NUMBER(12,2) becomes NUMBER(12,2) or NUMERIC(12,2), DATETIME2 becomes a timestamp, UNIQUEIDENTIFIER a UUID-sized string. Types with no equivalent, such as TIME on some warehouses or MONEY, are converted and listed in the report.
What happens to many-to-many relationships and reserved words?
A logical many-to-many in erwin becomes an associative (bridge) table with both keys. A table or column named with a SQL reserved word, like ORDER, is renamed (to orders) so the generated DDL runs. Every change made during conversion is listed.
What do I get if I import the model into OffDataAI?
A project with the full model: the complete DDL for your platform, an ERD, a dbt project, documentation, governance and lineage, and a business ontology your AI agents can query over MCP. You can then refine it in plain English, or let AI remodel it as a Kimball star schema or Data Vault 2.0. Importing works on the free plan.

Explore OffDataAI

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.

Get startedTalk to usBrowse templates

erwin is a trademark of Quest Software Inc. OffDataAI is not affiliated with or endorsed by Quest Software.