SyllabusRAG

SyllabusRAG helps you ask questions about a university syllabus easily.
It reads a syllabus file, builds a knowledge base and uses an AI model to answer questions.

How It Works ⚡

• The syllabus data is stored in a file called syllabus.json.
• Each course is read and turned into a simple text summary.
• These summaries are saved in knowledge_base.txt.
• The summaries are converted into vector embeddings using sentence-transformers.
• The vectors are stored in a FAISS index for fast similarity search.
• A small AI model (google/gemma-7b-it) is loaded from Hugging Face.
• When you ask a question:
  • If you include a course code, the system retrieves that course directly.
  • If not, it finds the closest course by comparing your question with the vector index.
• The AI generates an answer based only on the matched course information.
• If nothing relevant is found, it clearly responds that it cannot answer.

Folder Structure 📂

.
├── README.md
├── data
│   └── syllabus.json        # Course syllabus data in JSON format
├── main.ipynb               # Main implementation notebook
├── report/
│   └── report.pdf           # Project report
├── requirements.txt         # Python dependencies
└── training_data/
    ├── knowledge_base.txt   # Processed text summaries
    └── faiss_index/
        ├── index.faiss      # FAISS vector index
        └── index.pkl        # Index metadata

Requirements 🛠️

To run the project, create a requirements.txt file with the following content:

transformers
torch
accelerate
bitsandbytes
langchain
langchain-community
sentence-transformers
faiss-cpu
langchain-huggingface

Then, run this command in your notebook or terminal:

!pip install -r requirements.txt

Setup 🚀

1. Open Google Colab (recommended) or Jupyter Notebook with GPU support.
2. Clone or download this repository and upload main.ipynb along with syllabus.json.
3. Create a free Hugging Face account and copy your access token.
4. In Colab, click the 🔑 icon on the left, add a new secret named HF_TOKEN and paste your token.
5. Open main.ipynb and run all cells.

This will:

• Install required dependencies ⚙️
• Process the syllabus data 📑
• Build a FAISS index 📊
• Load the model 🤖
• Enable direct Q&A from the syllabus 💬

Example 💡

ask_question("What is the title for BECE309L?")
ask_question("What are the prerequisites for BECE309L?")
ask_question("What are the objectives of the Artifical Intelligence and Machine Learning course?")

✅ After completing these steps, your setup will be ready to query the syllabus!

Project Report 📊

For detailed analysis and evaluation results, check out the comprehensive project report:

PDF: report/report.pdf

The report includes qualitative and quantitative analysis, performance metrics and experimental results.

Limitations ⚠️

• The model is very basic and only lightly fine-tuned on a small dataset.
• Data pre-processing is raw and automated, not carefully cleaned or standardized.
• Because of this, the model can sometimes give inaccurate or incomplete answers.
• It works best only for specific, direct questions related to the syllabus.
• This project should be seen as a beginner-friendly starting point, not a production-ready system.