Player not loading? Watch on YouTube
This tutorial sets up Local Deep Research as a Docker application on Unraid, with Ollama providing the local LLM and SearXNG handling web searches. The presenter uses Qwen 2.5 7B and manually selects the provider and model in the container settings. The setup uses application port 5000; the search service has a separately configured endpoint.
After creating an encrypted account, the presenter adjusts the context size and runs a question about AI in North America. The walkthrough covers research iterations, source links and PDF reports, then creates a daily news subscription with a 1,440-minute update interval. A Vancouver news query demonstrates location-based research.
The troubleshooting is a useful part of the demonstration. The presenter discovers an incorrect SearXNG IP address and says the initial reports relied on model data, including news from 2023. Continued communication problems lead him to switch to an existing instance on his archive server. Later searches return reports with sources, though the Vancouver PDF export fails.
The final section shows how to save research into collections for the application's RAG library. The presenter explains that later queries can reference saved documents while still searching the web.
