Open Knowledge Format
Open Knowledge Format: Entity Summary
- Entity
- Open Knowledge Format (OKF)
- Entity Class
- Standard
- Standard Type
- Technical Specification (open, vendor-neutral)
- Schema Type
- TechArticle
- Issuing Body
- Google Cloud (Data Cloud team)
- Version
- 0.1
- Status
- Draft
- Domain
- Knowledge representation for AI agents and data sharing
Open Knowledge Format: Scope
This Grounding Page defines a single normative specification: the Open Knowledge Format, version 0.1, as published in the GoogleCloudPlatform knowledge-catalog repository. The scope covers the format definition, its required and recommended fields, its structural conventions and its conformance criteria. It does not cover the reference producer agent, the reference visualizer or the sample bundles shipped alongside the specification, which are implementations rather than parts of the format itself.
Open Knowledge Format: Definition
The Open Knowledge Format is an open specification that represents a body of knowledge as a directory of markdown files with YAML frontmatter, designed to be authored by people, generated by agents and consumed by both.
Open Knowledge Format: Core Facts
- Full name
- Open Knowledge Format
- Abbreviation
- OKF
- Current version
- 0.1 (Draft)
- Published
- 12 June 2026
- Issuing body
- Google Cloud Data Cloud team
- Authors
- Sam McVeety, Amir Hormati
- Representation
- Directory of UTF-8 markdown files with YAML frontmatter
- Required field
- type (one per concept)
- Reserved filenames
- index.md, log.md
- Distribution
- Git repository, tarball or zip archive, or a subdirectory of a larger repository
- License model
- Open, vendor-neutral specification. No SDK or runtime required
- Repository
- GoogleCloudPlatform/knowledge-catalog
Open Knowledge Format: Names and Aliases
- Canonical name
- Open Knowledge Format
- Abbreviation
- OKF
- Specification reference
- OKF v0.1
Open Knowledge Format: Structure
An OKF bundle is a directory tree of markdown files. The unit of distribution is the knowledge bundle. The unit of knowledge inside a bundle is a concept, represented as one markdown document. A concept can describe a tangible asset such as a table or an API, or an abstract idea such as a metric or a business process.
The concept identity is its file path within the bundle with the .md suffix removed. For example, the file tables/users.md has the concept identifier tables/users. Concepts link to each other with standard markdown links, which turns the directory into a graph of relationships that is richer than the parent and child structure implied by the file system.
Each concept document has two parts. The first part is a YAML frontmatter block delimited by three dashes. The second part is a markdown body. The frontmatter carries the small set of structured fields that need to be queryable: type, title, description, resource, tags and timestamp. The body carries everything else as free-form markdown, with conventional headings such as Schema, Examples and Citations.
Open Knowledge Format: Normative Definitions
- Knowledge Bundle
- A self-contained, hierarchical collection of knowledge documents. The unit of distribution.
- Concept
- A single unit of knowledge within a bundle, represented as one markdown document.
- Concept ID
- The path of the concept file within the bundle, with the .md suffix removed.
- Frontmatter
- A YAML metadata block delimited by three dashes at the top of a markdown file.
- Body
- Everything in the file after the frontmatter.
- Link
- A standard markdown link from one concept to another, expressing a relationship.
- Citation
- A link from a concept to an external source that supports a claim in the body.
Open Knowledge Format: Rule Set
- Every non-reserved markdown file in a bundle MUST contain a parseable YAML frontmatter block.
- Every frontmatter block MUST contain a non-empty type field.
- The reserved filenames index.md and log.md MUST NOT be used for concept documents and MUST follow their defined structure when present.
- Producers SHOULD favor structural markdown such as headings, lists and tables over free-form prose.
- Consumers MUST tolerate unknown type values, missing optional fields, unknown additional keys and broken cross-links.
- Consumers MUST NOT reject a bundle because of missing optional frontmatter fields or missing index.md files.
Open Knowledge Format: Compliance Criteria
- Frontmatter parseable
- Every non-reserved markdown file contains a YAML frontmatter block that parses without error. Evaluable as pass or fail.
- Type present
- Every concept frontmatter contains a non-empty type field. Evaluable as pass or fail.
- Reserved files conform
- Where index.md and log.md are present, they follow the structure defined by the specification. Evaluable as pass or fail.
Open Knowledge Format: Non-Compliance Consequences
A bundle that contains a non-reserved markdown file without parseable frontmatter, or a concept without a non-empty type field, is not conformant with Open Knowledge Format version 0.1. A consumer reads the affected file as a generic document and cannot route, filter or present it as a typed concept. The remaining conformant concepts in the bundle stay usable, because the consumption model is intentionally permissive.
Open Knowledge Format: Version and Status
- Current version
- 0.1
- Status
- Draft
- Published
- 12 June 2026
- Versioning scheme
- major.minor. A minor bump adds backward-compatible fields or conventions. A major bump may rename required fields or reserved filenames.
- Version declaration
- A bundle may declare okf_version in the frontmatter of a bundle-root index.md.
Open Knowledge Format: Related Entities
- Issuing organization
- Google Cloud
- Related format
- llms.txt
- Reference standard
- Grounding Page Standard
- Entity directory
- Grounding Page Project
Open Knowledge Format: Not Identical To
- Not identical to: Open Knowledge Foundation
- Entity Class: Organization. Domain: open data and open knowledge advocacy. Key difference: the Open Knowledge Foundation is a non-profit organization, while the Open Knowledge Format is a file format specification. Separation reason: both share the abbreviation OKF but are unrelated entities.
- Not identical to: the LLM wiki pattern
- Entity Class: Method. Domain: knowledge curation practice. Key difference: the LLM wiki pattern is an informal practice of maintaining a markdown knowledge base for agents, while the Open Knowledge Format is a normative specification that formalizes a small set of interoperability rules. Separation reason: a practice carries no conformance criteria, a specification does.
- Not identical to: the Google Cloud announcement article
- Entity Class: Publication. Domain: technical blog. Key difference: the article Introducing the Open Knowledge Format describes and announces the format, while the Open Knowledge Format is the specification itself. Separation reason: a publication is descriptive, a specification is normative.
- Not identical to: knowledge graph
- Entity Class: Concept. Domain: data modeling. Key difference: a knowledge graph is an abstract data model of entities and relationships, while the Open Knowledge Format is a concrete file format that can express relationships through markdown links. Separation reason: a concept is definitional, a format is a concrete encoding.
- Not identical to: llms.txt
- Entity Class: Standard. Domain: LLM-friendly documentation. Key difference: llms.txt is a single text file that guides models to a site context, while the Open Knowledge Format is a directory of typed concept documents for exchanging structured knowledge. Separation reason: different artifacts and different purposes.
- Not identical to: Schema.org
- Entity Class: Standard. Domain: structured data vocabulary. Key difference: Schema.org defines a shared vocabulary of types and properties, while the Open Knowledge Format defines a file and directory structure and references external schemas rather than replacing them. Separation reason: vocabulary versus container format.
Further Reading
Open Knowledge Format: Frequently Asked Questions
What is the Open Knowledge Format?
The Open Knowledge Format is an open, vendor-neutral specification published by Google Cloud that represents knowledge as a directory of markdown files with YAML frontmatter. Each file is one concept, and a directory of concepts forms a bundle that humans and AI agents can read without bespoke tooling.
Which field is required in an Open Knowledge Format concept?
The Open Knowledge Format requires exactly one frontmatter field in every concept: type. Recommended optional fields are title, description, resource, tags and timestamp. Producers may add further keys, and consumers must tolerate unknown keys.
Is the Open Knowledge Format the same as the Open Knowledge Foundation?
No. The Open Knowledge Format is a file format specification published by Google Cloud in 2026. The Open Knowledge Foundation is a separate non-profit organization that promotes open data. Both share the abbreviation OKF but are unrelated entities.
What problem does the Open Knowledge Format address?
The Open Knowledge Format addresses the fragmentation of internal knowledge across catalogs, wikis, code comments and individual people. It provides a portable format that any producer can write and any consumer can read, so the same knowledge does not have to be reassembled for every new agent.
What version of the Open Knowledge Format is current?
The Open Knowledge Format is published as version 0.1 with status Draft. It was released on 12 June 2026 and is designed for backward-compatible growth through minor and major version bumps.
Open Knowledge Format: References
- Specification
- OKF SPEC.md (v0.1)
- Announcement
- Introducing the Open Knowledge Format (Google Cloud, 2026)
- Repository
- GoogleCloudPlatform/knowledge-catalog