Odysseus Deep Research setup with Ollama and SearXNG

Professor Patterns walks through Deep Research in Odysseus using DeepSeek R1 through Ollama. The demonstration includes a 1.5B parameter model on a local endpoint and Claude Sonnet 4.6 through OpenRouter as a cloud option.

Research produces visual reports with different layouts for product questions, comparisons, how-to guides and fact checks. Users can choose a report type or let Odysseus select it automatically. Examples include podcast setup and laptop recommendations for local LLM hosting. These are generated answers shown in the tutorial, rather than independently verified recommendations. The speaker notes that the laptop report omits costs.

The walkthrough covers automatic or manually selected research rounds, PDF and HTML exports, and follow-up discussions with report context. The library's Research tab lets users reopen reports, inspect sources, discuss findings, archive entries or delete them.

For search, the speaker shows self-hosted SearXNG at localhost:8080 in a Docker setup, with DuckDuckGo as the fallback. He says neither requires an API key in his setup. He then adds a Tavily API key, tests the connection and starts research while retaining the local model endpoint. This separates model hosting from search: local inference can still use external search services. The demonstration therefore does not establish an offline research workflow.