Ollama, Open WebUI and Immich: privacy stack overview

The speaker tours his personal self-hosted app stack, with Ollama and Open WebUI handling local LLM queries. He demonstrates code generation on a Framework laptop running Ollama in Docker and says that request stays on the laptop. Hardware selection falls outside the video's scope: he stresses that available hardware determines model size and response speed.

For web search, his Open WebUI installation sends queries through his own SearXNG instance. He had to enable JSON output in SearXNG for the connection to work. In his March 2026 setup, he reports that a separate query URL was unnecessary. He also routes search traffic through a privacy VPN using UniFi policy routing, which he explicitly distinguishes from a SearXNG feature. These searches still contact external search engines.

The broader tour covers RSS reading, home automation, browser workspaces and remote access. Immich provides phone photo backups, object search and facial recognition in a local database. The speaker says it lacks photo editing in March 2026 and cautions that large libraries can take considerable time to index on slower hardware. He also discusses the risks of exposing a photo server publicly.

This is an overview of tools he uses, with references to separate setup tutorials, rather than a complete installation guide or hardware buying guide.