Import an Apache Ossie model

Apache Ossie is an open, vendor-neutral YAML format for describing datasets, relationships, metrics, and AI context. You can import an Ossie semantic model into a WisdomAI domain to reuse definitions your data team already governs in Snowflake, Databricks, dbt, or another Ossie-compatible tool. WisdomAI then answers questions using the same logic those tools use. This release supports import only. The Ossie file remains the source of truth. To pick up upstream changes, re-import the file.

​

Before you begin

Make sure you have the following:

• An Ossie model file (ossie-model.yaml or osi_document.json) that uses spec version 0.1.0 or 0.1.1.
• A WisdomAI connection to the warehouse that holds the tables referenced in the model (Snowflake, Databricks, or ClickHouse).
• Admin or Domain Editor permission in WisdomAI.
• Tables referenced in the model’s source fields that are visible to the WisdomAI connection.

​

Get an Ossie file

How you get an Ossie file depends on where your semantic model lives, export it from Snowflake, Databricks, or dbt, write it by hand for ClickHouse, or author your own.

• Snowflake Run the following query against an existing Semantic View and save the output to a .yaml file:
  SELECT SYSTEM$READ_OSSIE_YAML_FROM_SEMANTIC_VIEW('database.schema.semantic_view_name');
  
• Databricks Export the Unity Catalog Metric View YAML, then convert it with the Ossie UC Metric Views converter in the converters/ directory of the apache/ossie repository.
• dbt dbt 1.12 and later writes target/osi_document.json every time you run dbt parse. Use that file directly.
• ClickHouse ClickHouse has no native semantic layer or Ossie export. Author the Ossie file by hand, or generate it from a dbt project that runs on the dbt-clickhouse adapter. Set the expression dialect to CLICKHOUSE to use ClickHouse-specific functions.
• Hand-authored files Any file that validates against the Ossie 0.1.x JSON schema imports. See the Apache Ossie specification for details.

​

What WisdomAI imports

The table below shows how each Ossie object maps to a WisdomAI object in your domain.

​

Import a model

Complete the following steps to import your Ossie file into a domain:

1. Open the domain you want to import into, or create a new domain.
2. Go to Domain Settings and select Import from Apache Ossie.
3. Select the warehouse connection that holds the tables referenced in the model.
4. Upload your ossie-model.yaml or osi_document.json file.
5. Review the preview. WisdomAI lists every table, join, metric, and context item it found, and flags anything it could not resolve.
6. If WisdomAI could not find a table, use the dropdown next to it to map it to the correct table in your connection.
7. Click Import.
8. Wait for validation to complete. WisdomAI runs each join and metric against the warehouse to confirm that it executes.
9. Open the domain and ask a test question that uses one of the imported metrics. Check the step-by-step reasoning panel to confirm that WisdomAI used the imported definition.

​

After the import

Keep the following in mind when you work with imported objects in your domain:

• WisdomAI labels imported objects Source: Ossie so you can tell them apart from objects created in WisdomAI.
• You can add tables, joins, and metrics alongside imported ones.
• You can edit an imported object in WisdomAI, but the next re-import overwrites those edits. Make durable changes in the Ossie source and re-import.

​

Update the model

To pick up upstream changes, repeat the import with the new file. WisdomAI shows a diff of added, changed, and removed objects before it applies anything. WisdomAI deletes removed objects from the domain unless you clear them in the diff view.

​

Limitations

The Apache Ossie import has the following limitations in this release:

• Import only. Export from WisdomAI to Ossie is not available.
• Spec versions 0.1.0 and 0.1.1 only. WisdomAI rejects files marked 0.2.0 or later with a version error.
• Manual upload only. Connector-native import from Snowflake Semantic Views and Databricks Metric Views, and API-based import for CI pipelines, are planned.
• Metrics without an executable expression for your dialect (for example, some dbt MetricFlow derived or cumulative metrics) import as documented metrics. They are not queryable until you add an expression.
• Custom extensions are preserved but not used.

​

Troubleshoot import errors

​

Next steps