Local AI runs on macOS, Linux, and Windows. Choose your preferred method below — all options pull the same pre-built Docker images and produce an identical result.
Local AI is fully containerized, so the only real requirement is Docker. Make sure the following are in place on the machine where you plan to run it.
Install Docker Desktop for your platform. On Windows, Docker Desktop includes WSL 2 integration — enable it during setup. On macOS and Linux, Docker Engine works too. Verify your installation by running the command below.
Docker Compose V2 is included with Docker Desktop (all platforms). If you installed Docker Engine on Linux without Desktop, add the Compose plugin separately. The installer will alert you if it is missing.
Minimum 8 GB RAM and 50 GB of free disk space. Local AI runs on Apple Silicon (M1/M2/M3), NVIDIA GPUs, and standard x86 CPUs. No dedicated GPU is required — all models run on CPU if no GPU is available.
macOS 12+, any modern Linux distribution (Ubuntu 20.04+, Debian 11+, Fedora 37+), or Windows 10/11 (64-bit). On Windows, Docker Desktop must have WSL 2 backend enabled.
The fastest way to get started. A single command checks your system requirements, downloads all Docker images, generates secure secret keys, and starts every service automatically. No manual configuration needed for a standard setup. Supported on macOS, Linux, and Windows.
Open a terminal and paste the command for your operating system. The installer will verify that Docker is running, check available disk space and RAM, download all required images from Docker Hub, generate unique secret keys for your instance, and start all services automatically.
Apple Silicon (M1 / M2 / M3) — in Machine mode Ollama uses the Apple Metal GPU, so inference is significantly faster.
On first run, the installer pulls all Docker images from Docker Hub — approximately 3–6 GB in total depending on your connection speed. This is a one-time download; subsequent starts take only a few seconds. The installer will display a progress indicator and confirm when all services are healthy.
Once installation is complete, open your browser and navigate to http://local-ai.localhost. You will be guided through a brief first-time setup to create your admin account. After that, go to Settings → Models to pull your first AI model from Ollama.
For users who want full control. Clone the repository, review and customize the environment file, then bring the stack up yourself. This method lets you adjust any configuration option before the first launch.
Copy .env.example to .env and open it in any text editor. At minimum, replace the placeholder secret keys with secure random values. You can also adjust the Ollama endpoint, Whisper model size, and compose profiles to match your setup.
Deploy Local AI to a Kubernetes cluster for team or enterprise use. The Helm chart includes configurable resource limits, ingress settings, persistent volume claims, and horizontal pod autoscaling out of the box.
These commands work from the directory where Local AI is installed (default: ~/local-ai). Use the -f docker-compose.release.yml flag to target the correct compose file.
Another application — such as Apache, Nginx, or XAMPP — is listening on port 80. Stop that service before running the installer. On macOS you can check with sudo lsof -i :80. On Windows, open Task Manager and look for the conflicting process under the Services tab.
Ollama automatically uses Apple Silicon GPU on macOS and NVIDIA CUDA on Linux when available. On Windows, make sure Docker Desktop has WSL 2 GPU support enabled (Settings → Resources → WSL Integration). Also ensure Docker Desktop is allocated sufficient RAM — at least 6 GB is recommended for running models.
AI model files are large — typically 4–8 GB each. You can remove models you no longer need from Settings → Models inside the app. To free space from Docker itself, run docker system prune to remove unused images and build cache.
Allow 1–2 minutes for all services to become healthy on first launch. If the app still does not load, check the logs for errors: docker compose -f ~/local-ai/docker-compose.release.yml logs -f. Common causes are insufficient RAM causing a container to crash, or a port conflict that was not caught by the installer.
Join the community or file an issue on GitHub.