Introducing the Open Knowledge Format (OKF)

As foundation models advance, the key factor determining the performance of agentic systems is securing relevant Context. However, knowledge within enterprises today is scattered across fragmented systems such as metadata catalogs, wikis, and code comments, making it very difficult for AI agents to integrate and utilize it.

To address this, we are introducing the Open Knowledge Format (OKF) v0.1. OKF is an open specification that formalizes the LLM-wiki pattern proposed by Andrej Karpathy, aiming to be a universal knowledge representation that is easy for both agents and humans to read.

The key features of OKF are as follows:

• Markdown-based: Readable in any editor and on GitHub without a separate runtime or SDK.
• File-centric structure: Has a simple file structure that can be distributed as a tarball or hosted in a Git repository.
• Uses YAML Frontmatter: Structured data that needs to be queried, such as type, title, description, and tags, is managed via YAML.

Through this standard, developers can manage knowledge like code, and AI agents can autonomously perform the tedious work of reading and updating documentation, building an increasingly useful knowledge library over time.