HKUDS/DeepTutor
DeepTutor
DeepTutor is a framework for personalized AI tutoring and educational content generation. It functions as an agentic workflow system that executes reasoning loops to complete multi-step tasks, transforming raw sources into structured learning materials such as interactive books, quizzes, and concept graphs.
The platform distinguishes itself through an extensible skill architecture that allows the installation and auditing of third-party capability packages from community registries. It utilizes persona-driven tool policies to deploy persistent AI companions with unique behavioral profiles and specialized operational constraints.
The system integrates a versioned retrieval augmented generation knowledge management system to organize document collections and a context-aware markdown editor for evidence-grounded text expansion. It further maintains a personalized learning workspace by synchronizing memory across tools and utilizing hierarchical trace memory to make user personalization visible and editable.
A command line interface is provided for managing system configurations and triggering agent capabilities.
Features
- Agentic Workflow Frameworks - Functions as an agentic workflow framework executing reasoning loops with specialized tools and persistent memory.
- Interactive Learning Books - Compiles raw sources into structured interactive books containing quizzes, flashcards, animations, and concept graphs.
- Agent Skill Extensions - Implements a framework for installing and auditing modular third-party skill packages to expand agent capabilities.
- Agentic Reasoning Loops - Implements orchestration patterns for iterative execution where models call tools and refine answers to complete multi-step tasks.
- Agentic Workflow Automation - Uses intelligent agents to coordinate tasks and automate complex workflows by analyzing project context.
- Educational Tutors - Deploys persistent AI companions with unique personas and shared memory to provide tailored educational support.
- Content Generators - Transforms raw sources into structured learning books with interactive quizzes and concept graphs.
- Knowledge Management Systems - Provides a knowledge management system that organizes document collections into versioned indices for RAG.
- Persistent AI Companions - Creates specialized digital partners with unique personas and tool policies that connect to external messaging platforms.
- RAG Knowledge Management - Organizes and versions document collections to support retrieval augmented generation and prevent information loss.
- Educational Content Frameworks - Creates structured learning materials like quizzes and concept graphs from raw sources to improve student engagement.
- Agent Persona Frameworks - Assigns specialized roles and behavioral profiles to autonomous agents to define their operational constraints.
- Document Grounding - Integrates retrieved evidence with specific text selections in a markdown editor for targeted content expansion.
- User Personalization Traces - Implements a system for tracking user activity through multi-level traces to make personalization visible and editable.
- Multi-Level Memory Management - Records user interactions in a hierarchical structure to manage user, session, and agent state for personalization.
- Memory and Context - Shares knowledge bases and memory across different tools and workflows to prevent information silos.
- Selection-Aware Drafting - Provides a markdown editor capable of rewriting or expanding specific text selections grounded in retrieved evidence.
- Evidence-Grounded Editors - Provides a markdown editor that uses retrieved evidence to rewrite or expand specific text selections based on user context.
- Versioned Indices - Organizes document collections into distinct versioned snapshots to ensure stable retrieval augmented generation.
- Integrated Learning Workspaces - Integrates tutoring and research into a unified workspace where context persists across different learning modes.
- Plugin Architectures - Provides an architecture that allows extending core functionality through modular, dynamically loaded skill packages.
- Shared Knowledge Graph Memory - Synchronizes memory and state across different tools via a shared knowledge base to maintain consistent user context.
- End-User Applications - AI-powered personalized learning assistant with document Q&A, exercise generation, and deep research capabilities.
