Skip to main content

Project Overview

CLOVA Tutor is an open-source demo project that explores how to design a conversational AI tutor for education, grounded in the context of K–12 learning environments in Korea.

Rather than delivering a finished product or validating learning outcomes, this project focuses on how to structure model architecture and dialogue flow when building an education-oriented AI tutor. Its primary goal is to serve as a reference for designing learning-centered conversational systems.

System Overview​

CLOVA Tutor is an AI tutor system built on Naver Cloud’s HCX-005 model. It is designed around the idea that learning should happen through continuous dialogue, where problem solving, explanation, questioning, and review are connected within a single conversational flow.

Through conversation with a Math or English tutor, learners can solve problems, receive explanations, ask follow-up questions, and revisit past learning contexts. These interactions are preserved and reused, allowing learning dialogue to continue over time within a single interface.

Design Focus​

CLOVA Tutor treats large language models not as simple response generators, but as systems responsible for managing the flow of learning dialogue.

Instead of optimizing for one-off answers, the project focuses on how an AI tutor can guide learners through a structured learning process while preserving context and intent.

The full model structure and tool orchestration are described in detail in the
Model Architecture documentation.

Project Goals​

CLOVA Tutor starts from the observation that many learning services focus primarily on providing correct answers, while far fewer systems are designed around structured learning dialogue and continuity.

This project is organized around the following goals.

1. Dialogue structures for self-directed learning​

  • Enable learners to discover and solve problems on their own
  • Build a learning cycle that flows from
    learning Q&A → recommended problem solving → related questions → review

2. Improved understanding through natural-language dialogue​

  • Provide problem-solving experiences through real-time conversation rather than static content
  • Adapt explanations and difficulty dynamically based on learner responses

3. Learning continuity​

  • Preserve learning context by linking chat history with learning notes
  • Enable re-solving and review to happen naturally without breaking flow

4. AI as a supportive learning partner​

  • Explicitly limit the AI’s role to supporting the learner’s thinking process, not replacing parents or teachers
  • Reduce the burden of studying alone while ensuring learning agency remains with the learner

Service Architecture​

CLOVA Tutor is organized into four conceptual layers.

1. Access Layer​

Responsible for service entry points and basic navigation.

  • Account creation and login
  • Home screen and sidebar UI
  • New chat creation
  • Recent and full chat lists
  • Settings (theme, user guide)

2. Conversation Layer​

The core layer where learning happens through real-time dialogue with the AI tutor.

  • Text input for problems and questions
  • Learning goal setting and goal-aware feedback
  • Tutor responses (solutions, explanations, Q&A)
  • Automatic suggestion of example questions

3. Learning Content Layer​

Handles learning-related content and problem interactions.

  • Problem solving and evaluation
  • Step-by-step explanations and concept clarification
  • Viewing previously solved problems
  • Follow-up questions based on problems
  • Saving problems and explanations as learning notes

4. Learning Note Layer​

Supports review and repeated learning based on stored learning content.

  • Viewing saved problems and explanations
  • Deleting learning notes
  • Re-solving saved problems
  • Navigating back to the original chat context

Who Is This Project For?​

This project is intended as a reference for:

  • Developers designing or researching educational AI tutors
  • Product managers and planners exploring dialogue structures for LLM-based learning services
  • Teams looking to apply ReAct or tool-based LLM architectures in a real service context
  • Anyone seeking educational AI system design examples for research or portfolio purposes

Disclaimer​

CLOVA Tutor is a demo and research-oriented project, not a production-ready service.

  • All code in this repository is provided for educational and experimental purposes.
  • Learning outcomes, correctness, and system stability are not guaranteed.
  • The architecture does not account for production-scale traffic, full privacy compliance, or operational robustness.
  • Learning content is generated using an AI-backed temporary database; completeness and correctness are not guaranteed.

Recommended use cases include:

  • Studying AI tutor architecture and UX design
  • Prototyping conversational learning systems
  • Research, learning, and portfolio projects

If you intend to build a production or commercial service based on this project, additional validation, security design, and legal review are strictly required.