Player not loading? Watch on YouTube
This tutorial walks through Local Deep Research on Windows using the speaker's BAT installer. The script checks for Python, pip, Docker and Ollama, installs the research package, and starts the services. The speaker also describes a pip installation path for Windows, macOS and Linux.
The setup uses Ollama for the local LLM and runs SearXNG in a Docker container to aggregate search results. The demonstrated addresses use localhost port 8080 for search and port 5000 for the research interface. After creating an account, the speaker selects Qwen 3.6 27B BF16 through Ollama and submits a question about agent frameworks beyond LangChain.
The demonstration follows research subtopics, extracted search content and progress logs through to a report with source references. The speaker describes LangGraph as the workflow framework and SQLite as the local data store. Research uses SearXNG web searches and crawls external sources.
Reports can be exported as Markdown, saved in collections and indexed into a RAG knowledge base. PDF export needs extra setup: the speaker names Pango earlier and Pandoc later. He presents the project as an open source demonstration of agent workflows with local models and says it is not a complete enterprise tool.
