DeepTutor: AI tutor that tracks accuracy rates to recommend only weak points

DeepTutor is not a simple Q&A chatbot but a system that remembers users' learning levels and preferences to provide personalized guidance. It utilizes various search engines, including LlamaIndex-based RAG, GraphRAG, and LightRAG, to extract accurate information from documents and generate explanations based on that data. It analyzes PDFs or web pages uploaded by users to build a knowledge base, enabling natural question-and-answer flows based on that content.

The core feature, 'Guided Learning', evaluates user comprehension in real time and suggests the next learning step. It tracks accuracy rates and response times to identify weaknesses and automatically recommends additional materials to address them. Additionally, the 'Deep Research' mode generates reports that analyze complex topics from multiple perspectives, explicitly citing sources for each claim. This enables a deep learning experience that goes beyond the fragmented answers provided by existing chatbots.

This platform supports major LLM providers such as OpenAI, Anthropic, and Gemini, and also allows for local model execution. Through MCP (Model Context Protocol), it integrates with external tools to perform various tasks, including code execution and web search. In particular, the 'Partners' feature allows users to converse with AI agents they have defined, enabling interaction with virtual tutors that act as experts in specific fields. This is useful for educators and developers building their own AI tutoring environments.

It is suitable for a wide range of users, from individual learners to educational institutions. It is especially useful for students who need to understand complex concepts step-by-step or developers who need to quickly master new tech stacks. Released under the Apache-2.0 license, it is customizable and can be easily deployed via Docker. Unlike existing chatbot services, it features the ability to be used securely through self-hosting, even in environments where data privacy is a priority.