CrowdLlama is a distributed system that leverages the open-source Ollama project to run LLM inference tasks across multiple nodes using peer-to-peer (P2P) networking, enabling collaborative large language model inference workloads.

CrowdLlama enables distributed AI computing by allowing users to share their computational resources and access distributed AI capabilities through a peer-to-peer network. The system uses a DHT (Distributed Hash Table) for peer discovery and coordination.

The core distributed system implementation featuring:

• DHT-based peer discovery for network coordination
• Worker nodes that advertise GPU capabilities and supported models
• Simple metadata protocol for querying worker information
• Consumer components (planned) for distributed task execution

Built with Go, this repository contains the foundational P2P networking infrastructure that powers the entire CrowdLlama ecosystem.

Terraform-based infrastructure automation for deploying CrowdLlama DHT servers on Linode cloud infrastructure:

• Automated deployment with Docker and Docker Compose
• Auto-updating containers via Watchtower integration
• Production-ready configuration with firewall rules and systemd services
• Multi-region support across Linode's global infrastructure
• GitHub Actions integration for CI/CD pipeline

This repository enables scalable, reliable deployment of CrowdLlama infrastructure components.

Cross-platform desktop application built with Electron and React, providing:

• User-friendly interface for sharing computational resources
• Chat interface for interacting with AI models
• Network status monitoring with real-time peer count
• Model selection for contributing different AI models (llava, mistral, codellama, phi, gemma)
• Modern UI with Tailwind CSS and responsive design

The desktop app makes it easy for users to participate in the CrowdLlama network and access distributed AI capabilities.

┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
│   Desktop App   │    │   Worker Node   │    │   DHT Server    │
│   (Electron)    │◄──►│   (Go/Ollama)   │◄──►│   (Go)         │
│                 │    │                 │    │                 │
│ • Chat UI       │    │ • GPU Resources │    │ • Peer Discovery│
│ • Network Status│    │ • Model Serving │    │ • Coordination  │
│ • Resource Share│    │ • P2P Network   │    │ • Metadata      │
└─────────────────┘    └─────────────────┘    └─────────────────┘

1. Deploy Infrastructure: Use the infra repository to set up DHT servers
2. Run Workers: Build and run worker nodes using the crowdllama repository
3. Use Desktop App: Download and run the desktop application

We welcome contributions! Each repository has its own contribution guidelines:

• crowdllama - Core distributed system
• infra - Infrastructure automation
• desktop - Desktop application

All CrowdLlama projects are licensed under the MIT License. See individual repositories for license details.

• Main Repository: crowdllama/crowdllama
• Infrastructure: crowdllama/infra
• Desktop App: crowdllama/desktop
• Ollama Project: ollama/ollama

CrowdLlama - Distributed AI Computing for Everyone