Apache Ossie
The collab on this reads like a who’s-who: Snowflake, Databricks, dbt Labs, Salesforce, Collibra, Atlan, Mistral AI 👀 and fifty-plus others. Worth a serious look now that every vendor is racing to make their platform AI-ready. As a knowledge graph person I find it somewhat puzzling and it feels like yet another addition to the convoluted semantic space.
It’s a YAML specification for so query engines and AI agents stop disagreeing about stuff. That’s a real problem solved cleanly and the vendor-neutral governance under the ASF is the right home for it.
Now the graph question: is this RDF, LPG, ontology? Nope. The core model is relational: fact tables, dimension tables, joins, foreign keys. It’s a semantic layer in the BI/metrics sense, not a semantic layer in the knowledge representation sense. There is no “Ontology Representation” working group tasked with mapping Ossie’s classes onto formal ontology standards, which tells you the founders feel that gap too. Worth watching, but today Ossie’s “ontology” is a controlled vocabulary of metrics and dimensions, not a graph of typed entities, relationships and axioms you can reason over.
Does it help with provenance? Only in a narrow sense. Does it help with agent memory? No, different problem entirely. Is it data governance? No, and it’s honest about that boundary. Is it a good semantic layer for companies? No, it’s a letter of intention, not a tech stack.
It’s something valuable but the name “semantic layer” is the problem. Like so many other semantic layers out there 😔
Ossie 🦘: https://ossie.apache.org/