Google Cloud Introduces Open Knowledge Format for AI Systems

Image credit: Search Engine Journal

Google Cloud on Tuesday introduced the Open Knowledge Format (OKF), an open specification designed to standardize how artificial intelligence systems access and exchange essential knowledge.

The new format aims to address a significant limitation in current foundation models, which often lack relevant context, particularly for advanced agentic AI systems, Google Cloud said.

OKF seeks to standardize disparate knowledge sources, such as internal catalogs, wikis, and documentation, into a common format that both humans and AI systems can readily interpret. The specification organizes various concepts, including datasets, metrics, application programming interfaces (APIs), tables, and runbooks, into structured documents.

Google has provided reference implementations, including an enrichment agent and an HTML visualizer, along with sample bundles available on GitHub to assist developers in adopting the format. The company stated that OKF utilizes markdown files and YAML frontmatter, making it human-readable and easily editable.

The format operates on a producer-and-consumer model, with AI agents and large language models (LLMs) identified as the primary consumers of the structured knowledge. Google Cloud emphasized that OKF is not a new tool or platform but rather an agreement on how knowledge should be structured.

The need for such a standard was previously highlighted by Andrej Karpathy, who discussed the concept of ‘LLM Wikis’ to provide contextual information to AI. OKF v0.1 represents an initial release, with Google intending for the specification to evolve through community input and collaboration.

Source: Search Engine Journal