What the Open Semantic Interchange (OSI) spec means for metrics, semantics, and AI

The pace of AI adoption across the data ecosystem is accelerating, and semantic context is quickly becoming critical infrastructure for trustworthy AI-enabled analytics.

At Coalesce 2025, we shared our commitment to open, interoperable semantics: we introduced MetricFlow as an open-source project and announced dbt Labs’ participation in the Open Semantic Interchange (OSI), alongside partners like Snowflake, Databricks, and Salesforce, working toward a shared specification.

Since then, dbt Labs has worked closely with the OSI partners to publish an initial OSI specification, so businesses can standardize semantic context before semantics get defined differently across tools.

Today, the first version of the OSI specification is available in an open-source, Apache 2.0–licensed repository.

OSI defines a vendor-neutral, extensible model for representing semantic layer constructs, including datasets, metrics, dimensions, relationships, and context, so they can be interpreted consistently across tools, platforms, and AI-enabled applications.

Why OSI matters to analytics engineering teams

Organizations often define business metrics in multiple places. If you've ever rebuilt the same metrics across tools, debated which "active users" definition is correct, or watched dashboards drift after silent changes, you understand the problem OSI is designed to address.

OSI provides a vendor-neutral interchange format for semantic definitions—metrics, dimensions, datasets, relationships, and the context needed to interpret them—so those definitions can be transferred between tools without being re-authored or re-interpreted.

dbt Labs' focus: Operationalizing semantics

dbt complements the OSI spec by making semantic definitions operational: teams can define and govern metrics in the dbt Semantic Layer, and execute them consistently with MetricFlow. OSI provides the interchange format to move those definitions across tools (like Snowflake and Tableau). Together, they let teams author once and adopt interoperable semantics incrementally.

Looking ahead, dbt is committed to using the new OSI spec to power semantic interoperability from the dbt Semantic Layer across the broader analytics and AI ecosystem. Keep an eye out for updates!

A shared goal is semantics you can trust anywhere

We're proud to be collaborating on OSI and committed to building in the open with the OSI community and partners. The specification will only get better with real-world input from teams defining metrics today.

If you want to participate in OSI, the best ways to engage include:

• Visit the OSI project site to learn more.
• Review the OSI repository and the current specification.
• Start a discussion in the OSI community forums if you have a use case or proposal.
• Contribute directly via pull requests, including tooling and converters.