local-ai-packaged
Run all your local AI together in one package - Ollama, Supabase, n8n, Open WebUI, and more!
Local AI Packaged is a Docker Compose setup that runs a collection of AI and automation tools together on your own machine or server, without sending data to external cloud services. You download the repository, fill in a configuration file with passwords and secret keys, run one Python script, and all the services start up together.
The bundle includes Ollama, which downloads and runs large language models locally so you can chat with them privately. Open WebUI provides the chat interface for those models and also lets you connect to AI agent workflows. n8n is a visual workflow automation tool with over 400 pre-built integrations, used here to build the AI agents themselves. Flowise is a second, simpler tool for building AI agent flows with a drag-and-drop interface. Supabase provides a PostgreSQL database, authentication, and file storage as a self-hosted alternative to cloud database services. Qdrant and Neo4j offer two different ways to store and query AI-relevant data: Qdrant is a vector database for similarity search (used in retrieval-augmented generation workflows), and Neo4j is a graph database for storing relationships between pieces of information. SearXNG is a privacy-focused search engine that aggregates results from many sources, letting agents search the web without going through Google. Langfuse tracks what your AI agents are doing over time so you can observe and debug them. Caddy handles HTTPS certificates automatically if you want to expose the services on a public domain.
Startup profiles select the GPU configuration: Nvidia, AMD, CPU-only, or a mode for Mac users who run Ollama outside of Docker since Apple Silicon cannot directly share the GPU with Docker containers. Pre-built n8n workflow files are included in the repository to give you working examples without starting from scratch.
The README notes that recent Supabase updates have added new required environment variables, and it calls out the specific values to add if you have an existing installation from before mid-2025.