🌌 Know-Task-OS

The ultimate Local-First Agentic Mesh and Visual Operating System.

A collaborative Kanban platform for human teams and autonomous AI agents. Features autonomous task management, a dynamic agent skill system, and strict rule enforcement.

---

📑 Table of Contents

• Concept & Philosophy
• Monorepo Architecture
• The Agentic Mesh (Core Modules)
  • 1. Visual Dispatcher (Kanban)
  • 2. Knowledge Agent (RAG Librarian)
  • 3. Aegis Gatekeeper (Security)
• System Topography
• Getting Started
• Tags & Meta

---

🧠 Concept & Philosophy

🍏 Part of the Mac AI Ecosystem Initiative This project is part of a large-scale initiative to create missing hardcore tools and extensions for AI development natively on Apple Silicon.

Traditional autonomous agents write code directly into your repository, often breaking the main branch, hallucinating context, or leaking credentials. Know-Task-OS solves this by enforcing a strict, visual, step-by-step pipeline:

1. Isolation First: Every task dragged into "In Progress" triggers a git worktree add command. Agents operate in isolated sandbox folders. The main branch is mathematically protected.
2. Contextual Precision: Instead of flooding the LLM with the entire codebase, the native Vector RAG module feeds the agent only semantically relevant files.
3. Shift-Left Security: Before any AI-generated code is committed, an offline Rust-based neural network scans the diffs for leaked secrets, missing tests, and architectural drift.

---

🏗 Monorepo Architecture

This project is built as a high-performance Turborepo. It physically merges multiple independent AI systems into one cohesive operational mesh.

Know-Task-OS/
├── apps/
│   ├── dispatcher-api/       # Python FastAPI Backend for Kanban Orchestration
│   ├── dispatcher-web/       # React/Vite Premium UI (Glassmorphism + SSE Terminal)
│   ├── knowledge-api/        # RAG Vector Engine & Embeddings API
│   ├── knowledge-web/        # Knowledge Management Dashboard
│   ├── security-brain/       # Local MLX/Neural Network for Secret Detection
│   ├── security-github-app/  # Webhooks and GitHub Integration
│   └── security-ui/          # Gatekeeper Review Interface
├── packages/
│   └── security-core/        # Blazing fast Rust core for AST parsing and diff scanning
├── turbo.json                # Turborepo Build Pipeline
└── package.json              # Global NPM Workspaces

---

🤖 The Agentic Mesh (Core Modules)

1. Visual Dispatcher (Kanban)

The central command hub. Acting as the Product Owner interface, you drag and drop tasks. The backend communicates via Server-Sent Events (SSE) to stream the agent's internal monologue ("Thoughts", "Commands", "Errors") directly into the UI's live terminal. If an agent gets stuck, it pauses execution and waits for human input via the Stuck Protocol modal.

2. Knowledge Agent (RAG Librarian)

A robust Retrieval-Augmented Generation engine. When a task requires architectural changes, the Knowledge Agent semantically indexes your codebase and internal wikis (using ChromaDB/Qdrant), providing the Developer Agent with surgical, high-fidelity context.

3. Aegis Gatekeeper (Security)

An offline-first, strict QA and Security module powered by Rust (security-core). It intercepts the Developer Agent's output. If the agent attempts to hardcode an API key or push un-tested spaghetti code, Aegis rejects the payload and initiates Cross-Agent Negotiation, sending the error log directly back to the Developer Agent for an autonomous retry.

4. Dynamic Agent Skills & Rules

The autonomous task management relies on a dynamic skill system. Agents can hot-swap JSON schemas and tools from the global antigravity-bar skill registry at runtime based on the task context. The entire execution is governed by a strict set of global rules to ensure absolute determinism.

---

🗺 System Topography (Data Flow)

graph TD;
    A[User UI: Kanban Board] -->|Moves Task to In Progress| B(Dispatcher API);
    B -->|git worktree add| C{Isolated Worktree};
    B -->|Fetch Context| D[Knowledge Agent API];
    D -->|Injects RAG Data| E[Developer Agent LLM];
    E -->|Writes Code| C;
    E -->|Commits| F[Aegis Gatekeeper];
    F -->|Scans AST / MLX| G{Is Code Safe?};
    G -- Yes --> H[Merge to Main];
    G -- No --> I[Reject & Re-Prompt Agent];

Loading

---

🚀 Getting Started

Prerequisites

• Node.js (v20+)
• Python (3.11+)
• Rust (cargo)
• pnpm or npm (for Turborepo)

Installation

1. Clone the repository and install global dependencies:

   git clone https://github.com/helgklaizar/Know-Task-OS.git
   cd Know-Task-OS
   npm install

2. Start the entire Mesh ecosystem via Turbo:

   npm run dev

(Turbo will concurrently spin up the React Dispatcher, the FastAPI orchestration backend, the Knowledge API, and compile the Rust Security core).

---

🏷 Tags & Meta

GitHub Description:
A collaborative Kanban platform for human teams and autonomous AI agents. Features autonomous task management, a dynamic agent skill system, and strict rule enforcement.

Repository Tags:
knowledge-base task-management business-os mlx apple-silicon fastapi rust