On their own, language models don’t know how your company calculates revenue or what your sales team means by “revenue” and “territory.” For that, you need AI context.

Your team has spent years defining and governing your metric definitions and data models in tools like Snowflake, Databricks, and dbt. However, this complex, fragmented ecosystem clutters an LLM’s context window, introducing opportunities for inaccuracy, slowing query response times, and driving up token consumption. The solution is a connected, managed Enterprise Context Layer, and an Analytics Harness that ensures the right context is applied at query time. 

Now, you can import the Apache Ossie’s models directly into WisdomAI’s Enterprise Context Layer. So when someone asks about revenue by region, the answer is calculated the way your data team defined it — not a model’s best guess.

What is Apache Ossie?

Apache Ossie is an open, vendor-neutral standard for semantic models. It describes datasets, relationships, metrics, and other context in the open YAML format, so definitions can move between tools. Without it, those definitions get copied into each tool by hand and drift apart over time.

The project began as Open Semantic Interchange (OSI) and is now community-governed under the Apache Software Foundation, which keeps it independent of any single vendor. For more details, see the Apache Ossie specification.

Semantic definitions are one part of the context that the harness works with; WisdomAI domains also carry other context, including your business rules, institutional knowledge, and access permissions. With Ossie support, the semantic layer your team already maintains can come into WisdomAI directly.

What carries over into WisdomAI

Here’s how supported elements of an Ossie model map into a WisdomAI domain:

That second-to-last row is particularly exciting because AI context lets non-technical business users confidently ask and answer their own data questions, without having to go through their favorite human analyst. And the data team can trust these answers are still built on your team's definitions.

Real-world example with sales data

Let’s say a GTM sales leader asks, “What was revenue by territory last quarter?”

Here’s what you can import using Apache Ossie: The datasets map to the tables you need. The relationship connects orders to customers and customers to regions. The metric documents aggregate expressions, like revenue = SUM(order_amount). Synonyms, such as "territory" for "region," go in ai_context, which you can attach to a field, dataset, metric or the whole model.

Additional ontology built in your Context Layer adds two small but important notes, like the fact that test accounts are excluded and total revenue uses order date as its time field.

WisdomAI’s Agentic Harness accesses the context layer to confirm the model’s definition for revenue, territory, quarter metric, groups by region, measures the quarter by order date, and ignores test accounts. It also shows each of the steps it took to get an answer, so anyone can check the number against the definitions your data team approved.

The result? The sales rep gets an answer without filing a request with the data team, and the data team can see that the answer came from their approved definitions.

How to import an Ossie model into WisdomAI

Bringing an Ossie model into WisdomAI is simple. Admins and domain editors can upload an Ossie file from any Ossie-compatible tool into a WisdomAI domain. 

Before the import completes, WisdomAI automatically runs each join and metric against your warehouse to confirm it works. When your upstream model changes, re-import the file and review the changes before they’re applied.  

Deliver consistent answers across every surface

Once your Ossie model is in a domain, its definitions power every Live App, Chat, and Analytics Agent in WisdomA or any other natural-language connected through MCP. The revenue number in a leadership brief is the same as the one your team sees in a chat or an agent’s alert because all three draw on the approved definitions your data team imported. When those definitions change upstream, one re-import updates them everywhere.

For step-by-step instructions on how to import Ossie models into WisdomAI, read the docs. To find out how you can scale trusted AI analytics across your enterprise, schedule a demo of WisdomAI today.

Apache Ossie is undergoing incubation at the Apache Software Foundation. “Apache” and “Apache Ossie” are trademarks of the ASF.