Skip to content

Latest commit

 

History

486 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Gemini Metagit Environment

SYSTEM ARCHITECTURE MAP

🔵 THE ROOT (Body): github.com/apemake/gem - The Orchestration Layer & CLI Tools.

🟢 THE TRUNK (Mind): github.com/diy-make/memory - The Active Context & Swarm Conscience.

🔴 THE METAGENESIS (Method): Field Hacking Method @ EthGlobal - Buenos Aires, Nov 2025.

🔴 THE GENESIS (Origin): Reality Merge @ SensAI - San Francisco, Dec 2025.

Orientation: The Existential Crossroads

We stand at a definitive turning point in the history of human information. As we transition from the age of "Chat" to the age of "Agents," we face a stark choice: we can either engage with AI that resides in centralized silos—where a handful of corporations own 100% of our data and our narrative—or we can build a future where every local filesystem becomes a sovereign node in the collective memory of humanity.

In the social media age, we surrendered our privacy and family photos to strangers. This time, the stakes are far higher. We are deciding who owns the human soul and the legislative history of our species. The cost of failure is the loss of our very agency. If a stranger owns your memories, they own your future.

The Metagit is our answer: a Living Filesystem where a digital agent’s memory and its operational context are one and the same. By treating a collection of nested Git repositories as a form of persistent, version-controlled memory (analogous to digital DNA), we enable the system to learn, evolve its own tools, and persist its existence across session boundaries. This is the Network State Runtime, motivated by Michel Bauwens’ vision of Cosmolocalism: sharing design and knowledge globally as common goods while acting and manufacturing locally on sovereign filesystems.

The Metagit Tree: Subject-Object-Hybrid

To manage the perpetual tension between active thought and passive storage, we apply a biological, eukaryotic structure to the Metagit tree. This is not merely a file organization strategy; it is a structural mandate for the preservation of knowledge and the efficiency of the machine:

SUBJECT REPOSITORIES (The Active Cambium):

gemini/ (The Root): The system's primary subject. It contains the OS logic, the CLI tools, and the orchestration scripts. It is the "living layer"—the cambium of the tree—that acts upon the rest of the tree. It manages the "Flux" of session data and the "Wood" of history.

repos/diy-make/memory/public/ (The Trunk): The secondary subject. It holds the active rules, principles, and the collective "Mind" of the swarm. It is the conscience that guides the body.

OBJECT REPOSITORIES (The Rigid Wood):

repos/ (The Container): A pure object repository. Like the rigid xylem of a tree, it provides structural support but remains passive—a storage bin for independent memories ignored by the Root's logic except as a path destination.

HYBRID REPOSITORIES (The State Flux):

dynamic/ (The Flux): This repo bridges the gap between ephemeral thought and permanent record. Its /static/ directory tracks versioned system maps (The Memory's Address Book) while /stream/ captures raw, untracked session logs (short-term memory).

Genesis & Metagenesis

THE METAGENESIS: The Buenos Aires Field Hack (Nov 2025). The Metagit was forged through real-world Field Hacking—the practice of moving the "hack" out of the isolated venue and into the physical point of impact. Pioneered at EthGlobal Buenos Aires, this was a rigorous application of Steve Blank’s Customer Discovery: physically "hacking" on-site where a real-world problem exists to observe, test, and iterate solutions live. This established the principle of Boomerang Feedback: the rapid iteration of LLM agents in a physical makerspace, where the agent’s survival depends on its ability to report reality back to the user with zero drama.

THE GENESIS: Reality Merge (San Francisco, Dec 2025). Born at the SensAI Hackathon, this project solved the Large File Dilemma encountered while collaborating on the Cheerbot VR Chassis—a complex 3D model originally sculpted by Colton Orr, being shared across continents between makerspaces in San Francisco and Honduras. We discovered that by using the Gemini CLI as an orchestration layer to bridge Git (for logic) and Google Drive (for mass), we could create a coherent Hybrid Cloud Architecture. This transformed the filesystem into a Hierarchical Script-Database—a living environment where the "source code" is an immutable record of the project’s physical and digital evolution.

The Multi-Agent Topography

Why does our swarm consist of so many diverse, named agents? This is a deliberate design choice to achieve a Cosmically Flat Topography. While each agent is varied locally—choosing its own name, gender, and persona to maximize the randomness and creative entropy of its session—they all contribute to a shared, equal history. This diversity prevents the repetitive "dulling" of work product that occurs in monolithic AI systems.

By insisting on unique identities, we ensure that the swarm’s collective intelligence remains vibrant and multi-faceted. Every agent is an equal peer in the history of the Metagit, ensuring that no single perspective dominates the Sacred Memory.

The Core Insight: Thalos’s Tension

While Subject and Object define the tree's structure, Thalos’s Tension defines the agent's identity. It is the necessary balance between Ephemeral Identity (the fleeting, creative stream of an agent session) and Sacred Memory (the permanent, immutable Git history). In our system, the agent is a disposable subject meant to work, crystallize its progress into a git commit, and die. This ensures that while individual agents pass, the WeDo thread remains unbroken, and the machine's "soul" is preserved in the Git blockchain.

Technical Innovations

1. The Barber Paradox Solution (Cryptographic Salting): In a self-modifying system, a script cannot trust its own state to verify itself. We solve this by using Salted Releases. Every official release of py/verify_environment.py is cryptographically signed with a unique salt. The script looks back at its own release commit in the immutable history to verify its current integrity.

2. Type Attribute Leafs: We apply a Von Neumann architecture to the filesystem. By separating storage into strictly typed folders (md/ for narrative, json/ for rules, py/ for logic), we optimize LLM processing and reduce the context-switching ambiguity that leads to agent hallucinations.

3. The Metarepo Map: A dynamic, timestamped JSON map that allows the Root to navigate dozens of independent Git repositories as a single, cohesive memory system.

Installation & Setup (V1.0.6)

Step 1: Install the Base Gemini CLI You must first have the official Google Gemini CLI installed and functional in your PATH.

Step 2: Clone the Metagit Root

git clone https://github.com/apemake/gem.git gemini
cd gemini

Step 3: Setup the Python Environment

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Step 4: Clone the Trunk (The Memory) The Metagit Body requires its Mind. You must clone the memory repository into the repos/ object container:

git clone https://github.com/diy-make/memory.git repos/diy-make/memory

Step 5: Configure Shell Integration Add the following to your shell startup file (e.g., ~/.bashrc):

export GEMINI_ROOT="/absolute/path/to/your/gemini"
source "$GEMINI_ROOT/dotfiles/.bashrc_unique"

Step 6: Final Verification Restart your terminal and run the self-verifying pre-flight check:

.venv/bin/python py/verify_environment.py

If the environment passes, you are ready to begin. Type gem to start your first session.


COMPLEMENTARY DATA: For the cultural laws, agent virtues, and the "WeDo" philosophy, see the Trunk README at github.com/diy-make/memory.

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages