About the Governed Context Compiler (in preview)
The Governed Context Compiler gives AI agents, data platforms, and other tools the governed context that they need. The compiler delivers that context in the format that each tool expects. It reads an asset and its related assets in Collibra Platform and compiles them into one structured YAML file. You get the same trusted definitions, relations, and policies every time, without custom scripts.
With the Governed Context Compiler, you can:
- Define once how Collibra reads an asset type and how it shapes the output.
- Generate the output for any asset of that type.
- Produce output for Snowflake semantic views, Databricks metric views, Apache Ossie, or a structure that you define yourself.
- Copy the output into an AI conversation, or download it as a YAML file for a data platform.
Turn the Governed Context Compiler on or off
The Governed Context Compiler is on by default. An admin can switch it off on the In preview page in Settings, with the Governed Context Compiler option.
How the Governed Context Compiler works
A context specification is the configuration behind every output. The context specification defines which asset type to read from, which relations to follow, which values to extract, and how to arrange them. You create a context specification once for an asset type, and then you can run it on any asset of that type.
Collibra generates the output on demand, when you run a context specification on a specific asset. Collibra doesn't save the output to the asset or to an external system.
Two roles work with the Governed Context Compiler:
- Authors, typically data stewards or governance managers, create and maintain context specifications in the mapping builder.
- Consumers, such as data engineers and AI engineers, run existing context specifications on assets. They copy the output into another tool or an AI conversation, or they download it as a file.
Key concepts
| Concept | Definition |
|---|---|
| Context specification | A named configuration that combines a target schema, a target asset type, and a set of mappings. The context specification defines how one asset type becomes one output structure. |
| Target asset type | The asset type that a context specification reads from, such as Table, Data Product, or Column. Every mapping starts from an asset of this type. |
| Target schema | The structure of the output, either a predefined schema or a custom schema. |
| Field | One entry in the output structure. Each field has exactly one mapping. |
| Mapping type | The way that a field gets its value, for example from a property, from an attribute, or by following relations to other assets. |
| Output | The YAML document that Collibra generates when you run a context specification on an asset. |
Target schemas
The target schema controls the shape of the output and how the schema tree behaves when you configure a context specification.
| Target schema | Description |
|---|---|
| Snowflake Semantic View | A predefined field tree for a Snowflake semantic view. |
| Databricks Metric View | A predefined field tree for a Databricks metric view, version 1.1. |
| Apache Ossie | A predefined field tree for Apache Ossie, version 0.1.1. Apache Ossie is the new name of Open Semantic Interchange (OSI). |
| Custom Schema | An empty field tree. You define the full output structure yourself. |
A predefined schema gives you the fields, but not the mappings. You map each field to the information in Collibra that fits your operating model.
Context specifications as assets
Each context specification is a Context Specification asset. You can govern it like other assets, for example with responsibilities, permissions, and history. You configure and test the mapping on the asset page of the Context Specification asset.
Asset type inheritance
A context specification applies to its target asset type and to all subtypes of that asset type. For example, a context specification for Table also applies to Database View assets, because Database View is a subtype of Table.
The same principle applies when you configure a mapping. The attributes that you can select include the attributes assigned to the target asset type and to all of its subtypes.
Ways to get context
You can get the output of a context specification in the following ways:
- On the asset page of any asset, use Download context to copy the output or save it as a YAML file. For more information, go to Download context from an asset page.
- On the asset page of a Context Specification asset, use Test to check the output for any asset of the target asset type. For more information, go to Test a context specification.
- Use the Context Compiler API to compile the output for an asset from your own application or pipeline.
Helpful resources
- Example | Governed Context Compiler: Build a context specification for a data product (in preview)
- Create a context specification (in preview)
- Manage context specifications (in preview)
- Context specification mapping types (in preview)
- Test a context specification (in preview)
- Download context from an asset page (in preview)
- Context specification YAML reference (in preview)