The latest update adds a direct path for publishing an OKF bundle into Google Cloud's Knowledge Catalog, turning a static markdown collection into searchable, governed metadata that agents can consume at scale. Practitioners care because the previous git‑based distribution model did not integrate with existing data catalog assets, lacked unified access control, and required manual discovery of each bundle.
Why Knowledge Catalog solves the sharing problem
Knowledge Catalog already indexes assets from BigQuery, Cloud Storage, operational databases, and applications, attaching schema, lineage, ownership, and tags to each entry. By mapping OKF concepts onto the catalog’s native types, every piece of context becomes discoverable through a single, IAM‑protected index. This eliminates the need for separate repositories and aligns context data with the same governance policies applied to other data assets.
Registering an OKF bundle in Knowledge Catalog
Implementation starts with a one‑time setup using the sample code provided in the Knowledge Catalog repository. The workflow invokes gcloud dataplex to create three resources:
- An
EntryGroupthat groups all entries belonging to the bundle. - An
EntryTypenamedokf-bundleto represent each concept defined in the bundle. - An
AspectTypecalledokfthat stores the trust‑signal fields introduced in OKF v0.2 (provenance, verification, freshness, attestation).
After registration, the push step uses the kcmd CLI (Metadata‑as‑Code) to create an okf-bundle entry for every concept. Each entry receives two aspects: an overview aspect containing the markdown body, and the okf aspect holding the structured signals. Display name, description, and tags are stored directly on the entry. Navigation files such as index.md and the root log.md are also ingested as entries, with the overview aspect providing their content.
Operational flow and considerations
Once the resources are in place, publishing a new version of the bundle is a single push operation. The catalog automatically indexes the new entries, making them instantly searchable across projects. Because the catalog already supports cross‑project lookup, agents in any project can retrieve relevant context without additional configuration, provided they have the necessary IAM permissions.
Practitioners should monitor the following operational aspects:
- Consistency of the YAML front‑matter and required fields defined by OKF v0.1.
- Timeliness of the trust‑signal fields; stale freshness or missing attestation could reduce confidence in the bundle.
- IAM policy updates that may inadvertently restrict agent access to newly created entries.
Security and governance implications
Access to OKF bundle entries is enforced by Knowledge Catalog’s IAM model, meaning agents only see entries they are authorized to read. The inclusion of provenance and verification signals in the okf aspect provides an auditable trail for each piece of context, supporting compliance requirements without additional tooling.
While the catalog secures read access, write operations (setup and push) still rely on the credentials used to invoke gcloud dataplex and kcmd. Organizations should ensure those identities are protected and that role assignments follow the principle of least privilege.
Related CloudNinjas coverage: Google Cloud.
What This Means For Practitioners
Adopting the Knowledge Catalog integration turns OKF bundles from isolated markdown files into first‑class catalog assets, delivering searchable, governed context for any agent. The practical steps are limited to a one‑time resource registration and a repeatable push command, after which standard IAM controls and catalog search handle discovery and access. Teams should validate their trust‑signal fields, align IAM policies with agent requirements, and incorporate the push step into their CI/CD pipelines to keep context up to date.