Odysseus setup with Docker, Ollama and LM Studio

This tutorial covers Odysseus, an open source AI workspace that the presenter attributes to PewDiePie and identifies as AGPL3 licensed. The local setup uses Docker Desktop and Git. The walkthrough opens the app on port 7000, retrieves the initial admin credentials from container logs, and changes the temporary password.

The presenter connects Ollama and LM Studio through the local provider option in /setup, using endpoints with a /v1 suffix. A GPT-OSS 20B response demonstrates the LM Studio connection. The built-in cookbook recommends models for the available hardware, but the presenter reports that its downloaded model crashes when launched in this demonstration.

The feature tour covers chat and AI agent modes, web search, isolated shell access, email drafting, saved memory and imported skills. Automatic memory updates appear after further conversation; the presenter says he does not know exactly what triggers them. A deep research example asks for a model comparison.

The self-hosted VPS walkthrough uses Hostinger and connects OpenRouter with an API key, then tests Claude Haiku 4.5. This route sends inference to an external provider. It differs from the local model setup, so the presenter's broad privacy claims should be read in the context of the chosen backend.