Skip to main content
Apache Ossie (Incubating) — formerly the Open Semantic Interchange (OSI) — is an open, vendor-neutral specification for exchanging semantic models between tools. It defines a machine-readable format for describing datasets, relationships, fields, and metrics, along with the business context around them, so that any tool in a data stack can produce or consume semantic definitions without losing meaning. The project is incubating at the Apache Software Foundation, developed in the open on GitHub, and backed by a coalition of data and analytics vendors. Its goal is to make a business concept like “Monthly Active Users” resolvable consistently across warehouses, BI tools, and AI agents, regardless of where it was originally defined.

​
What Ossie covers

Ossie standardizes the shape of a semantic layer as an interchange format:
  • Semantic models — the top-level unit exchanged between tools.
  • Datasets — the queryable tables in a model.
  • Relationships — how datasets connect to each other.
  • Fields — the dimensions on each dataset.
  • Metrics — the measures defined against fields.
  • AI context — the descriptions and business meaning consumers (including AI agents) rely on.
  • Custom extensions — a custom_extensions slot for tool-specific detail that would otherwise be lost in a round-trip.

​
How Ossie maps to the Lightdash semantic layer

The Lightdash semantic layer already models the same building blocks Ossie standardizes: A Lightdash project doesn’t read or write Ossie files itself. Define your semantic layer in Lightdash YAML or dbt, where its tables, joins, dimensions, and metrics map to Ossie as shown above.

​
Where to learn more