An AI-powered academic mentoring system for Jeonbuk National University students, providing personalized curriculum recommendations, course information, and academic guidance.

The system consists of multiple microservices communicating via HTTP APIs:

• Open WebUI: Web interface (Port 8080)
• Pipeline: Request routing and processing (Port 9099)
• AI Modules: Specialized microservices for different tasks

AiMentor_edit/
├── open-webui/                  # Web UI application (Port 8080)
├── pipelines/                   # Request routing pipeline (Port 9099)
├── ai_modules/                  # AI microservices
│   ├── llm_agent-main/          # Main orchestration service with LangGraph (Port 8001)
│   ├── curriculum-main/         # Course recommendation and graph generation (Port 7996)
│   ├── faiss_search-main/       # Vector-based course search (Port 7997)
│   ├── department_mapping-main/ # Department mapping service (Port 8000)
│   └── tool_sql-main/           # Database query service (Port 7999)
├── docker-compose.yml           # Docker orchestration configuration
├── restart_all.sh               # Service restart script
└── Ai_mentor.png                # System interface screenshot

# Start all services
docker-compose up -d

# View logs
docker-compose logs -f

# Stop all services
docker-compose down

Access the application at http://localhost:8080

🌐 Production Service: http://210.117.181.110:8080

• Python 3.11+
• OpenAI API Key
• PostgreSQL database

Create .env file with:

OPENAI_API_KEY=your_openai_api_key
DB_HOST=your_database_host
DB_PASSWORD=your_database_password
VECTOR_DB_PASSWORD=your_vector_db_password

# Open WebUI (Port 8080)
cd open-webui
docker run -d -p 8080:8080 --name openwebui [image]

# Pipeline (Port 9099)
cd pipelines/
pip install -r requirements.txt
./start.sh

# LLM Agent (Port 8001)
cd ai_modules/llm_agent-main/
uvicorn main:app --host 0.0.0.0 --port 8001

# Curriculum Service (Port 7996)
cd ai_modules/curriculum-main/
python main.py

# FAISS Search (Port 7997)
cd ai_modules/faiss_search-main/
python main.py

# Department Mapping (Port 8000)
cd ai_modules/department_mapping-main/
python main.py

# SQL Tool (Port 7999)
cd ai_modules/tool_sql-main/
python main.py

The system employs four specialized AI agents, each designed for specific academic assistance tasks:

Port: 7996 | Service: curriculum-main

Provides personalized learning path recommendations based on student goals and interests.

Key Features:

• FAISS-based department and course retrieval
• LLM-powered query expansion for better search results
• Recursive course selection algorithm (up to 28 courses)
• NetworkX graph visualization of prerequisite/postrequisite relationships
• Semester-by-semester course layout (max 6 courses per semester)

Example Query: "I want to learn data science for startup entrepreneurship"

Port: 7999 | Service: tool_sql-main

Converts natural language questions into SQL queries for academic database searches.

Key Features:

• Natural language to SQL conversion using GPT-4o-mini
• Direct MySQL database query execution
• Professor, course, and schedule information retrieval
• Low temperature (0.05) for consistent SQL generation

Example Query: "What classes does Professor Oh Il-seok teach?"

Port: 7997 | Service: faiss_search-main

Hybrid search system combining LLM-based SQL filtering with vector similarity search.

Key Features:

• Two-stage search: LLM generates SQL filter → FAISS vector search
• OpenAI text-embedding-3-small for semantic search
• Pre-filtering to reduce unnecessary vector operations
• Dynamic indexing of filtered results

Example Query: "Machine learning courses in Computer Science department"

Port: 8000 | Service: department_mapping-main

Provides detailed department information, requirements, and curriculum structure.

Key Features:

• Department search and matching
• Graduation requirements information
• Major-specific curriculum details
• Course prerequisites and recommendations

Example Query: "Tell me about the Computer Science department requirements"

• Conversation context management across all agents
• Multi-agent query routing system via LangGraph
• Light query validation for academic-focused responses
• Real-time streaming responses

• Frontend: Svelte, Open WebUI
• Backend: Python, FastAPI, LangGraph
• AI: OpenAI GPT, LangChain
• Database: PostgreSQL
• Search: FAISS vector database
• Visualization: NetworkX, Matplotlib
• Containerization: Docker, Docker Compose

This project is developed for academic purposes at Jeonbuk National University.

All rights reserved by Natural Language Learning Lab (NLL Lab), Jeonbuk National University.

© 2025 NLL Lab, JBNU. All rights reserved.