0G-OmniMemory Framework 🧠🚀

0G-OmniMemory is a modular agent framework designed for the 0G decentralized AI operating system (dAIOS). It provides a robust infrastructure for building autonomous agents with persistent long-term memory and self-evolving reflection loops powered by 0G Storage and 0G Compute.

🌟 Key Innovations

• Persistent On-Chain Memory: Leverages 0G Storage to create a "Decentralized Brain" for agents, ensuring that knowledge is never lost across sessions.
• Reflection Primitive: Implements a reflect() mechanism allowing agents to analyze past failures, extract lessons, and store them as evolved intelligence on-chain.
• Modular Architecture: A clean separation between the Memory Layer (0G SDK integration) and the Core Logic Layer (Agent brain), allowing developers to easily swap memory strategies.
• Zero-Config Deployment: Designed for rapid prototyping of on-chain AI agents.

🏗️ Architecture

The framework consists of three primary layers:

1. Storage Layer (src/memory): Interfaces with the 0G Storage SDK to handle data chunking, Merkle tree generation, and on-chain uploading/downloading.
2. Core Layer (src/core): Manages the OmniMemory class, handling the indexing of Root Hashes and the logic for remembering/recalling and reflecting.
3. Agent Layer (evolving-agent.js): The implementation layer where developers build specific agent behaviors using the framework primitives.

🚀 Quick Start

Prerequisites

• Node.js v18+
• 0G Testnet Private Key (stored in .env)

Installation

npm install

Running the Demo

To see the agent evolve in real-time:

node evolving-agent.js

🛠️ Technical Stack

• Blockchain: 0G Chain (Testnet)
• Storage: 0G Storage (Turbo Indexer)
• Language: JavaScript (Node.js)
• SDKs: @0gfoundation/0g-ts-sdk, ethers.js

🔮 Future Roadmap

• Integration with 0G Compute for verifiable reflection inference.
• Multi-agent memory sharing via shared 0G Storage streams.
• iNFT integration for agent identity and memory ownership.

---

Built with ❤️ for the 0G Hackathon.