Skip to content

dexloom/lloom

Repository files navigation

🌐 Lloom - Inference protocol. Peer-to-Peer

License: MIT Rust Built with libp2p

Lloom is a cutting-edge decentralized peer-to-peer network that creates a trustless marketplace for Large Language Model (LLM) services. Built with Rust for blazing-fast performance and rock-solid reliability, Lloom enables direct connections between AI consumers and providers without intermediaries.

✨ Key Features

🔐 Cryptographically Secure

  • EIP-712 structured data signing for tamper-proof messages
  • Ethereum-compatible identity system with wallet integration
  • Dual-signature verification ensuring accountability
  • Replay attack protection with nonce management

🌍 Truly Decentralized

  • No single point of failure or control
  • Peer-to-peer communication via libp2p networking stack
  • Kademlia DHT for distributed discovery
  • Self-organizing network topology

🤖 Universal LLM Support

  • Compatible with OpenAI, LMStudio, and custom backends
  • Dynamic model discovery and announcement protocol
  • Real-time capability matching
  • Automatic executor selection based on requirements

💰 Transparent Economics

  • On-chain settlement with smart contracts
  • Granular token usage tracking (input/output)
  • Market-driven pricing discovery
  • Fair compensation for compute providers

🚀 Quick Start

For quick start please try our demo that connects to LLOOM network and asks "gpt-oss:20b" for introduction.

# Clone the repository
git clone https://github.com/dexloom/lloom
cd lloom

cargo run -p lloom-client -- --demo

🏗️ Architecture

Lloom operates with three primary node types:

  • 🔹 Clients: Request LLM inference services and pay for usage
  • 🔸 Executors: Provide computational resources and earn tokens
  • 🔺 Validators: Maintain network integrity and facilitate discovery
graph TD
    C[Client] -->|Signed Request| E[Executor]
    E -->|Model Announcement| V[Validator]
    C -->|Query Models| V
    V -->|Model Registry| V
    E -->|Signed Response| C
    C & E -->|Settlement| B[Blockchain]
Loading

📦 Components

Component Description
lloom-core Core protocol definitions, networking, and cryptographic primitives
lloom-client Client library for requesting LLM services
lloom-executor Executor node for serving LLM requests
lloom-validator Validator node for network coordination
faucet-server Test token distribution service

🛠️ Technology Stack

  • Rust - Systems programming language for performance and safety
  • libp2p - Modular peer-to-peer networking stack
  • Ethereum - Identity management and on-chain settlements
  • EIP-712 - Structured data signing standard
  • Kademlia DHT - Distributed hash table for peer discovery
  • Docker - Containerization for easy deployment

📚 Documentation

Comprehensive documentation is available in the docs/ directory:

🤝 Contributing

We welcome contributions! Please see our Contributing Guide for details on:

  • Code style and standards
  • Development workflow
  • Testing requirements
  • Pull request process

🔮 Roadmap

  • Core P2P networking implementation
  • EIP-712 message signing
  • Model announcement protocol
  • Basic executor and client functionality
  • Enhanced smart contracts with dual signatures and payment channels
  • WebSocket support for real-time updates
  • Executors benchmarking
  • Reputation scoring

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🌟 Join the Revolution

Lloom is building the future of decentralized AI services. Join our community:


Empowering AI through decentralization

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published