Google Cloud: the Markdown OKF Revolutionizes Data

Google Cloud recently unveiled the Open Knowledge Format (OKF), an initiative aimed at transforming data management by standardizing knowledge in the form of Markdown files. This new format allows information to be portable and interoperable across various systems, drawing inspiration from the "wiki LLM" model popularized by Andrej Karpathy.

The first version, OKF v0.1, presents knowledge in the form of directories of Markdown files enriched with YAML metadata. These files include a mandatory "type" field and several optional fields such as title, description, and tags. Concepts are interconnected through Markdown links, thus forming a knowledge graph accessible via any editor or platform like GitHub.

The OKF addresses a common issue in organizations where knowledge is often scattered across various catalogs, wikis, and code comments. This fragmentation complicates the work of AI agents, who must gather disparate information to perform tasks such as writing SQL queries. Google Cloud emphasizes that each agent developer and catalog provider currently has to solve this problem in isolation, which limits efficiency and interoperability.

Designed to be minimalist and portable, the OKF requires only a "type" field, leaving the rest to the discretion of the producer. This allows for great flexibility, where a package created by humans can be used by AI agents and vice versa. Google Cloud also offers several reference implementations, including an enrichment agent for BigQuery and examples for datasets like GA4 e-commerce, Stack Overflow, and Bitcoin.

In parallel, Google Cloud has updated its Knowledge Catalog to ingest and distribute OKF packages to agents. The specification and code are available on GitHub, providing a valuable resource for developers looking to adopt this new standard.