kossakovsky/selfhost-ai

What it solves

Selfhost AI provides a streamlined way to deploy a comprehensive, private AI and automation ecosystem on your own hardware. It eliminates the manual effort of configuring dozens of separate open-source tools, managing reverse proxies, and setting up SSL certificates, allowing users to create a private "AI homelab" without vendor lock-in.

How it works

The project is a Docker Compose template and an interactive installation wizard. It automates the system preparation, Docker installation, and the deployment of a curated set of services. It uses Caddy as a reverse proxy to handle automatic HTTPS via Let's Encrypt and organizes services under a wildcard domain. Users can choose which specific tools they want to deploy from a list of over 30 available services, including LLM runners, vector databases, and workflow engines.

Who it’s for

It is designed for developers, AI enthusiasts, and privacy-conscious users who want to full control over their AI infrastructure, data, and automation workflows on a VPS or dedicated server.

Highlights

• One-Command Installation: An interactive wizard that handles secrets, DNS configuration, and service selection.
• Extensive Toolset: Includes support for Ollama (local LLMs), n8n (automation), Flowise/Dify (agent builders), and vector stores like Qdrant and Weaviate.
• RAG Ready: Pre-configured components for Retrieval-Augmented Generation, including document converters like Docling and graph-based RAG via LightRAG.
• Production-Grade Infrastructure: Features a scalable n8n queue-based architecture with Redis and a built-in monitoring stack using Grafana and Prometheus.
• Community Content: Optional import of over 300 pre-built community automation workflows.