Odysseus and Ollama: Windows setup without Docker

Professor Patterns walks through a manual Windows installation of Odysseus, a chat interface for a local LLM. The tutorial uses his repository fork because he encountered Windows compatibility problems in the original. He says the fork removes Unix-only dependencies and recommends returning to the main repository once those issues are fixed.

The walkthrough assumes Git and Python are installed. It covers cloning the fork, creating and activating a Python virtual environment in PowerShell, installing requirements.txt, and running setup.py. Setup creates an admin account with a temporary password. The presenter changes the launch port from 7000 to 8000 after encountering a conflict, then signs in.

Odysseus detects his existing Ollama installation. For viewers who want to run models locally, he discusses download sizes and suggests starting small. His examples include DeepSeek R1 at 1.5B and 7B parameters. He then downloads Qwen 3.5 0.8B through Ollama, refreshes Odysseus, selects the model, and sends a test message.

The interface also has OpenAI and Anthropic API settings. The presenter describes a cookbook option that selects model downloads based on hardware. These cloud connections are optional paths alongside the local setup; the tutorial does not establish a minimum hardware requirement.