📄 Syllabus Q&A Bot using LangChain, FAISS, and Ollama

This is a simple Retrieval-Augmented Generation (RAG) pipeline that allows users to ask context-specific questions about a syllabus document (PDF). It leverages the LangChain framework with FAISS for vector storage, HuggingFace for embeddings, and Ollama to run the LLaMA3.2 model locally.

---

🔧 Features

• Loads a syllabus.pdf using PyMuPDF
• Splits the text into manageable chunks for embedding
• Generates dense vector embeddings using HuggingFace's all-MiniLM-L6-v2
• Stores and retrieves chunks via FAISS similarity search
• Uses Ollama's LLaMA3.2 model for answering questions
• Continuously answers syllabus-related questions in a loop

---

📁 File Structure

├── syllabus.pdf             # The syllabus document (you need to place it here)
├── main.py                  # The script described below
└── README.md                # You're reading this file

---

🚀 Getting Started

1. Install Dependencies

Make sure you have Python 3.8+ installed, then install the required libraries:

pip install langchain langchain-community faiss-cpu PyMuPDF langchain-huggingface langchain-ollama

2. Set Up Ollama

Download and install Ollama. Then pull the required model:

ollama pull llama3

Ensure ollama is running in the background.

---

🧠 How It Works

1. Loads and reads syllabus.pdf
2. Splits the document into 500-character chunks (with 100-character overlap)
3. Generates vector embeddings using all-MiniLM-L6-v2
4. Stores them in a FAISS vector database
5. Uses LLaMA3.2 model from Ollama to answer user queries
6. Retrieves top 3 most relevant chunks as context

---

▶️ Running the Script

python main.py

You'll see a prompt like:

Question:

Ask your syllabus-related question and get a crisp, context-aware answer.

Example:

Question: Do we study about drones in the course?

Answer: Yes, we do learn about drones...

---

🛠 Customization

• Change PDF: Replace syllabus.pdf with any document.
• Switch Models: Modify the embedding or LLM model to suit your performance needs.
• Tweak Retriever: Adjust the number of returned documents (k) or switch to mmr search for diversity.

---

✅ Requirements

• Python 3.8+
• Ollama installed and running
• syllabus.pdf placed in the working directory

---

📌 Notes

• The prompt in the script is wrapped with: "And only Crisp to the Question asked only with respect to the context:\n\nQuestion: ..." This ensures responses are concise and context-bound.
• The script runs in an infinite loop; press Ctrl+C to exit.

---

📜 License

This project is free to use under the MIT License.