An LLM-powered assistant to help UC Davis students access campus services, events, and academic resources using a retrieval-augmented generation (RAG) pipeline.
- Intelligent Campus Assistant: Powered by fine-tuned embedding models and LLMs
- Comprehensive Knowledge Base: Processes 8,000+ campus FAQs and webpages
- Fast Semantic Search: FAISS-powered retrieval for relevant information
- Student-Focused: Tailored responses for UC Davis campus services and resources
- Data Collection: Automated scraping of campus FAQs and webpages
- Embedding Models: Fine-tuned PyTorch models for domain-specific embeddings
- RAG Pipeline: Retrieval-augmented generation with FAISS vector search
- LLM Integration: OpenAI GPT models for natural language responses
- Web Interface: Streamlit-based chat interface
AggieConnect/
├── src/
│ ├── data/ # Data collection and processing
│ ├── models/ # Embedding model training and inference
│ ├── rag/ # RAG pipeline implementation
│ ├── api/ # FastAPI backend
│ └── web/ # Streamlit frontend
├── data/
│ ├── raw/ # Raw scraped data
│ └── processed/ # Processed and cleaned data
├── models/
│ ├── embeddings/ # Trained embedding models
│ └── checkpoints/ # Model checkpoints
├── tests/ # Unit and integration tests
├── docs/ # Documentation
└── notebooks/ # Jupyter notebooks for experimentation
- Clone the repository
- Install dependencies:
pip install -r requirements.txt - Set up environment variables (see
.env.example) - Run data collection:
python src/data/collector.py - Train embedding models:
python src/models/train_embeddings.py - Start the web interface:
streamlit run src/web/app.py
The system provides a conversational interface where students can ask questions about:
- Campus services and resources
- Academic policies and procedures
- Events and activities
- Housing and dining information
- Financial aid and registration
Developed by a 4-person team demonstrating how machine learning and LLMs can enhance student access to information at UC Davis.