self-hosted-ai-starter-kit

Why we included this project

Teams that want to prototype AI agents on their own hardware often spend days wiring together a workflow engine, a model runner, and a vector store. This Docker Compose template skips most of that: it pairs n8n's low-code workflow editor with Ollama for local LLM inference, Qdrant as the vector store, and PostgreSQL for storage, so you can build things like appointment-scheduling agents, PDF summarizers, or Slack bots that keep data off third-party servers. The repo ships an example workflow plus AI Agent, text classifier, and information extractor nodes, giving you a concrete starting point rather than a blank canvas. It is honest about being a proof-of-concept foundation rather than a production deployment, which suits teams that want to validate a self-hosted idea quickly. Nvidia and AMD GPU profiles are included, and Apple Silicon users can fall back to CPU.