DeepTutor overview: tutoring modes and self-hosted setup

DeepTutor is an open source learning assistant from the University of Hong Kong's Data Intelligence Lab. This overview describes its v1.0.0 architecture and explains how a self-hosted workspace combines tutoring with document retrieval. The speaker says persistent learner memory tracks progress and learning patterns across sessions, then adapts teaching to that profile.

The five modes cover chat, Deep Solve, quiz generation, Deep Research and Math Animator. Deep Solve separates investigation, solution and verification into distinct AI agent roles, with another solution pass when verification finds a problem. The speaker presents this as a reliability benefit; the transcript provides no benchmark of answer accuracy. Math Animator uses Manim for mathematical explanations.

Users can upload PDFs, Markdown and text documents. The overview describes direct retrieval through LlamaIndex, knowledge graph retrieval through LightRAG and multimodal graph retrieval through RAG-Anything. TutorBots have separate workspaces and persistent memory, while a Markdown co-writer helps revise notes.

The setup tour selects LLM and embedding providers and tests connections. The stated requirements are Python 3.11 or higher and at least one LLM API key. Docker images support AMD64 and ARM64, and a CLI supports chat and knowledge base management. The speaker mentions local models, but does not establish a fully offline setup.