Open Knowledge Format (OKF) Architecture & Implementation

Overview

The Open Knowledge Format (OKF) is an open, vendor-neutral specification published by Google Cloud to formalize the "LLM-wiki" knowledge pattern. It standardizes how curated organizational and domain knowledge is authored, structured, and consumed by AI agents and RAG (Retrieval-Augmented Generation) systems.

In Mission Control, OKF serves as the structured knowledge and offline failover layer for the in-game AI overlay and decision-making engine.

Why OKF in Mission Control?

Zero Binary & C++ Dependency Overhead:
- Eliminates the need for heavy, compilation-prone vector database binaries (e.g., ChromaDB, FAISS C++ wheels) during PyInstaller packaging.
- Avoids dynamic runtime linking issues on Windows and Linux release distributions.
Human & Agent Co-Authoring:
- Knowledge files are clean, readable Markdown documents with YAML frontmatter.
- Developers, gamers, and AI agents can create, edit, or patch game intelligence directly via text editors or Git.
Dual-Tier Resilient Retrieval:
- Tier 1 (Distributed Cloud Sync): Fetches live catalog intelligence, game features, and summaries from the central Distributed Server (/api/catalog).
- Tier 2 (Local OKF Markdown & SQLite/BM25): Loads and indexes structured .md files directly from backend/rag_data/ and backend/data/ for 100% offline, zero-latency in-game guidance.

File Structure & Specification

All OKF documents live in Gaming/backend/rag_data/ (or Gaming/backend/data/knowledge/) using the .md or .okf extension.

Example OKF Document (`rag_data/cyberpunk_2077.md`)

MARKDOWN
---
type: game_intel
title: Cyberpunk 2077 Optimization and Night City Guide
game_id: cp2077
tags: [fps, dlss, ray-tracing, settings, night-city]
version: 1.0.0
last_updated: "2026-08-24"
---

# Cyberpunk 2077 Intel

Night City is divided into six main districts: City Center, Heywood, Santo Domingo, Pacifica, Watson, and Westbrook.

## Performance & Optimization Guidelines

- **Crowd Density**: Reduce to Medium on 6-core CPUs to alleviate draw-call bottlenecks in dense downtown areas.
- **DLSS & Ray Tracing**: Enable DLSS Super Resolution in Quality mode for 1440p / 4K. Combine with Frame Generation for smooth 100+ FPS output.
- **Path Tracing**: Recommended only on NVIDIA RTX 4070 Ti / 5070 and above.

Supported Metadata Schema (YAML Frontmatter)

Field	Type	Description
`type`	string	Document categorization (e.g., `game_intel`, `hardware_profile`, `patch_notes`).
`title`	string	Human-readable title of the knowledge document.
`game_id`	string	Unique identifier for game-scoped RAG filtering (e.g., `cp2077`, `witcher3`, `general`).
`tags`	list	List of indexing keywords used to assist lexical search matching.
`version`	string	Semantic version of the knowledge asset.
`last_updated`	string	Timestamp of last modification.

System Architecture: OKF & Resilient RAG

The RAG architecture in Mission Control connects local OKF authoring, automated distributed catalog enrichment, SQLite persistence, and in-memory BM25 retrieval to fuel in-game real-time AI decisions.

1. High-Level Multi-Tier Architecture Diagram


SYSTEM ARCHITECTURE DIAGRAMMERMAID SVG ENGINE
Generating visual flowchart...

2. Runtime Query & Retrieval Sequence Workflow

The sequence diagram below demonstrates how an in-game query is resolved with zero network latency and localized isolation:


SYSTEM ARCHITECTURE DIAGRAMMERMAID SVG ENGINE
Generating visual flowchart...

3. Failover & Self-Healing Lifecycle

If SQLite is corrupted, missing, or if network connectivity drops, the system self-heals automatically:


SYSTEM ARCHITECTURE DIAGRAMMERMAID SVG ENGINE
Generating visual flowchart...

Build & Deployment Compatibility

PyInstaller (MissionControl.spec): The rag_data folder is bundled as a physical data asset:

Python
datas = [
    ('data', 'data'),
    ('rag_data', 'rag_data'),
    ...
]

Packaging: No third-party C++ libraries or binary wheels are required. PyInstaller bundles the pure Python parser cleanly with zero build warnings.
Electron Builder (package.json): Packed directly into extraResources as part of the backend bundle.

Open Knowledge Format (OKF)