How Ontos bridges the gap between technical metadata and business meaning
Here's a scenario that might sound familiar. You've got Unity Catalog humming alongโtables are registered, lineage is tracked, access controls are in place. Technically, everything's governed. But then someone from the business side asks: "What exactly does customer_ltv_score mean? Who owns the definition? Is this the canonical source or is there another table I should be using?"
And suddenly you realize: Unity Catalog tells you what exists. It doesn't tell you what it means.
That's the gap Ontos is designed to fill unless you are also using metric views with unity catalog.
https://learn.microsoft.com/en-us/azure/databricks/metric-views/
First, a Quick Word on Databricks Labs
If you're not familiar with Databricks Labs, it's essentially Databricks' experimental workshopโa GitHub org full of community-driven projects built by Databricks engineers and partners. These aren't official products. They don't come with SLAs or enterprise support contracts. They're tools built by practitioners to solve real problems, shared openly so others can benefit.
You've probably heard of some Labs projects already: UCX for Unity Catalog migrations, DLT-META for metadata-driven pipelines, Tempo for time series. Ontos is the newest addition, and it tackles something most of us have been duct-taping together with spreadsheets and Confluence pages: business semantics.
What "Labs" really means: These projects are provided AS-IS for exploration. No formal support, but active development and community contribution. Think of them as battle-tested tools from people who've felt the same pain you have.
What Ontos Actually Does
Ontos layers business context on top of Unity Catalog. It gives you a proper interface for managing the stuff that technical metadata doesn't capture:
Data Products โ Define and publish datasets as formal products with owners, SLAs, and documentation. Not just "here's a table," but "here's a curated, trustworthy asset you can build on."
Data Contracts โ Formalize agreements between producers and consumers. What's the schema? What's the freshness guarantee? What happens when something breaks?
Business Glossaries โ Finally, a place to define what terms actually mean. "Customer" might mean three different things across your orgโnow you can make that explicit.
Master Data Management โ Track canonical sources and golden records. When someone asks "which customer table is the source of truth?"โyou'll have an answer.
The architecture is straightforward: React frontend, FastAPI backend, deploys as a Databricks App. It talks directly to Unity Catalog APIs and stores its own metadata in PostgreSQL (or Lakebase in production).
Why This Matters
I've seen teams spend months building elaborate governance frameworks in wikis and spreadsheets, only to watch them go stale the moment someone changes a column name. The problem isn't disciplineโit's that the business context lives in a completely different system than the technical reality.
Ontos sits where it belongs: right next to Unity Catalog, pulling live metadata, letting you annotate it with business meaning. When a table changes, you see it. When a definition drifts, you catch it.
It's not magic. It's just putting the business semantics where they should have been all along.
Getting Started
The project is fully open source under Apache 2.0. You can run it locally for development or deploy it as a Databricks App. The README is thoroughโenvironment variables, database setup, the works.
If you're on the Free Edition, remember: you get one Databricks App. Ontos could be a strong candidate for that slot if data governance is your priority.
Quick start: Clone the repo, set up your .env with your Databricks workspace URL and a SQL warehouse ID, run yarn dev:frontend and yarn dev:backend. You'll have a local instance running in minutes.
Go Check It Out
If you've ever wished Unity Catalog understood your businessโnot just your tablesโOntos is worth a look. It's early days (46 stars as I write this), but the foundation is solid and the problem it solves is real.
Give it a spin. File issues if you hit snags. And if you build something cool with it, share it back with the community.
