Odysseus and Ollama: VPS setup tutorial

This tutorial sets up Odysseus as a self-hosted ChatGPT alternative with Ollama handling model inference on the same VPS. The presenter uses Hostinger's Docker manager to deploy the workspace and a separate Ollama container, then connects them so chats use a downloaded model rather than an external API.

The walkthrough covers server provisioning, the initial admin login, model installation in the terminal, and endpoint configuration in Odysseus. The presenter chooses Qwen 3 8B and describes its download as about 5 GB. He recommends a starting VPS plan with one CPU core, 4 GB of RAM, and 50 GB of storage, but the tutorial does not establish performance expectations or minimum hardware requirements for that model.

The main configuration step is Docker networking. Containers on the same VPS do not automatically discover each other in this setup. The presenter attaches them to a shared network, adjusts the endpoint to match the container identifier, and tests the connection. Odysseus then reports one available model and answers simple chat prompts.

The presenter describes the workspace as private and says this configuration avoids external API credits. Those claims concern the demonstrated server arrangement; the walkthrough also includes purchasing VPS hosting. It focuses on setup rather than testing the workspace's broader research, email, or memory features.