Solving AI Context Decay with Google's Open Knowledge Format (OKF) For Deep State of Mind (DSOM) For My AI

In the world of Open Source and AI development, we are always learning and improving. Today, I want to share 

a major milestone in our Deep State of Mind (DSOM) project

 

 ( https://github.com/linuxmalaysia/deep-state-of-mind-for-my-ai ) : 

 

We have officially adopted Google Cloud's Open Knowledge Format (OKF) v0.1.

Why do we need OKF?

If you have been following the DSOM project, you know that our biggest challenge with autonomous AI agents

 is Context Decay. Over time, AI agents forget the architecture, the rules, and the operational history of a project.


Historically, organizational knowledge—like runbooks, architecture decisions, and API schemas—is scattered

 across different wikis, databases, and code comments. OKF solves this by creating a standardized 

"common language" for knowledge. It structures documentation so that it is equally readable by human engineers

 and autonomous AI agents.

 

Think of it like a cluster for your knowledge: instead of combining hardware to share workloads, we are combining

 structured metadata to share context!

The Migration Journey: What We Did Today

 

To achieve 100% compliance with the OKF v0.1 specification, we had to modernize our Sovereign AI Workspace.

 Here is what we accomplished today:

  1. Official Lexicon Integration: We updated our core documentation to use official OKF terminology.

     What we previously called "Markdown Files" are now Concepts, and the directories that hold them

     are Knowledge Bundles.

  2. Strict YAML Frontmatter: We executed a mass schema migration across our entire 

    Memory Palace. We programmatically updated 13 different Knowledge Bundles

     (including our closet.md and SKILL.md files) to replace legacy tags with the official title key,

     and we injected strict ISO 8601 timestamp metadata.

  3. Absolute Portability: We refactored our Bootstrap Blueprints. By converting hardcoded absolute

     paths into soft relative paths, our repository is now instantly portable for any team that clones it, 

    regardless of their operating system or drive letters.

What's Next for DSOM?

https://malaysia-open-source-community.gitbook.io/deep-state-of-mind-dsom-protocol-for-my-ai 

 

By enforcing YAML frontmatter across all our Markdown-based memory assets, we have transitioned from an

 ad-hoc AI memory structure into an enterprise-grade, OKF-compliant Knowledge Graph.


Future AI agents can now effortlessly traverse our .agents/ directory by querying structured YAML

 in milliseconds, without having to read the entire markdown body.


We must do our best, because we are the best, and we are going to be the best! Best of the Best.

Let’s keep building, learning, and pushing the boundaries of Open Source AI.

 

Stay tuned for more updates, and as always, keep experimenting!

Harisfazillah Jamel / LinuxMalaysia

Catatan popular daripada blog ini

Guide to Configuring PHP-FPM Slow Logs - Ubuntu Server or AlmaLinux

Modern SSL/TLS for Nginx: A Practical Guide to Balancing TLS 1.2 and 1.3 on AlmaLinux and Ubuntu

HOWTO: Install WSL2 and Move AlmaLinux 9 to Another Drive