Odysseus setup with Docker, Ollama and Llama 3.2

This tutorial walks through a self-hosted Odysseus installation on a Mac, using Docker Desktop and Ollama to run models locally. The presenter starts with the supporting applications, then uses the terminal to download a model and retrieve Odysseus code from its official website and GitHub repository. The transcript does not include the literal commands, so viewers need the referenced installation instructions to follow those steps.

The model example is Llama 3.2, with a download the presenter describes as about 2 GB. He checks that it appears in Ollama and can answer messages before building and starting Odysseus. Docker Desktop must remain running for the demonstrated container setup; its interface shows the stack's status and CPU and memory use.

After startup, the presenter opens a local HTTP address in a browser and signs in with a username and password. He then connects the local LLM through the model selector: click the plus button, use quick search and confirm Add. API connections are also offered as optional alternatives. Typing the suggested setup command produces no response in his demonstration, so he uses the interface instead.

The final chat test initially invokes web search unintentionally. The presenter adjusts the settings and retries with a greeting and a joke request. This detail matters for anyone seeking offline use: the demonstration includes a web-search option, rather than establishing that every feature works without a network connection.