The project trying to fix data’s oldest problem just joined the Apache Incubator. It’s called Ossie - and yes, the mascot is a kangaroo. 🦘
Every company has this problem: “Revenue” means one thing in Finance’s dashboard, another in Marketing’s, and a third in whatever your AI agent just hallucinated.
Same metric. Three definitions. Zero trust.
Apache Ossie (previously Open Semantic Interchange / OSI) is an open, vendor-neutral spec that defines metrics, dimensions, and relationships once, in declarative YAML, so every BI tool, query engine, and AI agent works from the same source of truth.
Why this one might actually stick:
- 50+ organizations on board, up from 17 launch partners
- Contributors from Snowflake, Databricks, dbt Labs, Salesforce, Dremio, GoodData and more
- Converters for dbt Semantic Layer and Apache Polaris already merged
- Now under ASF governance no single vendor controls the standard
The AI angle is the real driver: LLMs answering business questions need semantic context, not just table schemas. A portable, machine-readable definition of “what Monthly Active Users actually means” is exactly the grounding agents are missing today.
My honest take: semantic standards have a graveyard of predecessors, and the spec is still incubating - expression language and advanced metric logic are on the roadmap, not done. But the vendor lineup here is unusually broad, and “your metrics belong to you, not your BI tool” is a bet worth watching.
📚 Source: https://ossie.apache.org/
P.S. How many definitions of “Revenue” live in your stack right now? Be honest. 👇
