Solving the Fragmented Context Problem

Modern AI agents often struggle because the necessary context—table schemas, metric definitions, and runbooks—is trapped in incompatible silos like metadata catalogs, wikis, and docstrings. This fragmentation forces developers to rebuild context-assembly logic for every new agent or vendor. The Open Knowledge Format (OKF) addresses this by formalizing the 'LLM-wiki' pattern into a portable, interoperable standard that allows agents to read and update structured knowledge directly.

The OKF Specification

OKF v0.1 is a lightweight, file-based format that requires no new runtime, SDK, or compression scheme. It represents knowledge as a directory of markdown files containing YAML frontmatter.

• Structure: Each concept (e.g., a database table or metric) is a single markdown file. The file path serves as its unique identity.
• Metadata: Files use a reserved YAML frontmatter block for fields like type, title, description, resource, tags, and timestamp.
• Graph Capability: Concepts are linked using standard markdown syntax, allowing the directory to function as a rich, navigable knowledge graph.
• Interoperability: Because it is just markdown and YAML, OKF bundles are human-readable, version-control friendly, and can be consumed by any agent without translation or proprietary API calls.

OKF vs. Traditional RAG

Unlike Retrieval-Augmented Generation (RAG), which dynamically derives knowledge from raw, unstructured chunks at query time, OKF stores curated, cross-linked concepts. This approach provides agents with a stable, high-fidelity 'source of truth' that is easier to maintain and update. By keeping these files in version control alongside the code they describe, teams can ensure that documentation evolves in lockstep with the systems it represents, leveraging the ability of LLMs to manage cross-references and file updates effectively.