Odysseus setup with Ollama on Apple silicon

This tutorial installs Odysseus, a self-hosted AI workspace, on an Apple silicon Mac and connects it to Ollama. The speaker also explains the native Windows installation option for users who want local GPU acceleration.

The Mac walkthrough starts with Homebrew and Python 3.11 or later, then clones the repository and runs its macOS start script. The script creates a virtual environment and prompts for login credentials. The speaker uses the dev branch, described as the default at recording time, while suggesting main for greater stability. Although the guide recommends Docker generally, the presenter chooses native installation because Docker cannot access Metal for GPU acceleration on this Mac.

Odysseus opens at localhost on port 7860. To run models locally, the presenter installs Ollama, sets its context length to 16K, and tests the connection before adding it in Odysseus. LM Studio is mentioned as another supported endpoint, but its setup is not demonstrated.

The final example uses an API key for OpenCodeGo and selects MiniMax M3 in agent mode to generate a to-do list application. The speaker reports that the result works; this is a single demonstration rather than a general performance test. The tutorial ends with Control+C to stop the server and the same start script to restart it.