DeepTutor is an innovative, agent-native personalized learning assistant designed to revolutionize the way individuals learn and interact with educational content. Developed by HKUDS, this repository offers a comprehensive suite of tools and features aimed at creating a dynamic and adaptive learning experience. The project is built using Python 3.11+ for the backend and Next.js 16 for the frontend, ensuring a robust and modern technological foundation.
At its core, DeepTutor provides a unified chat workspace that integrates six distinct modes: Chat, Deep Solve, Quiz Generation, Deep Research, Math Animator, and Visualize. This integrated approach allows users to seamlessly transition between different learning activities within a single context, fostering a more cohesive and efficient learning process. Users can initiate a conversation, then leverage multi-agent problem-solving capabilities, generate quizzes, visualize complex concepts, and delve into in-depth research, all without losing the thread of their learning journey.
A key feature is the AI Co-Writer, a multi-document Markdown workspace that empowers users with AI-driven collaboration. Users can select text, and the AI can rewrite, expand, or summarize content, drawing upon the user's knowledge base and web resources. This feature promotes active learning and content creation, allowing users to build upon their understanding and create personalized learning materials.
The Book Engine is another standout feature, enabling users to transform their learning materials into interactive "living books." This multi-agent pipeline designs outlines, retrieves relevant sources, and compiles rich pages with 14 different block types, including quizzes, flashcards, timelines, concept graphs, and interactive demos. This feature allows for the creation of engaging and dynamic learning experiences.
The Knowledge Hub is designed to facilitate the organization and utilization of learning resources. Users can upload PDFs, Markdown, and text files to build RAG-ready knowledge bases. Insights can be organized within color-coded notebooks, quiz questions can be revisited in the Question Bank, and custom Skills can be created to tailor DeepTutor's teaching approach. This feature ensures that user-provided documents actively contribute to every conversation, enhancing the learning experience.
DeepTutor also incorporates Persistent Memory, which builds a dynamic profile of the user's learning journey, including what they've studied, how they learn, and their learning goals. This profile is shared across all features and TutorBots, allowing the system to adapt and personalize the learning experience over time.
Personal TutorBots are a unique aspect of DeepTutor. These are not simply chatbots, but autonomous tutors that reside in their own workspaces, each with its own memory, personality, and skill set. These TutorBots can set reminders, learn new abilities, and evolve as the user progresses.
The Agent-Native CLI provides a command-line interface for interacting with all capabilities, knowledge bases, sessions, and TutorBots. This CLI offers rich terminal output for human users and structured JSON output for AI agents and pipelines. This feature allows for seamless integration with other AI tools and workflows.
The repository provides multiple methods for getting started, including a guided Setup Tour, manual local installation, and Docker deployment. The Setup Tour simplifies the installation and configuration process, while the manual installation offers more control. Docker deployment provides a streamlined way to run DeepTutor without requiring local Python or Node.js installations. The project also supports multiple languages, including Chinese, Japanese, Spanish, French, Arabic, Russian, Hindi, Portuguese, and Thai, making it accessible to a global audience. The project is actively developed, with frequent releases that introduce new features and improvements.
