RAG-Enabled Q&A Chatbot

This project is a Retrieval-Augmented Generation (RAG) powered chatbot that provides accurate, contextually relevant answers based on a private knowledge base. Instead of relying solely on pretrained knowledge, the chatbot retrieves information from a custom textbook on biochemistry and cell biology genetics, making it a powerful study tool for students and researchers in these fields.

The system uses cutting-edge AI tools and libraries including LangChain, Pinecone, Hugging Face models, and Streamlit to deliver authoritative, domain-specific responses through a simple and intuitive web interface.

qa-rag-chatbot.mp4

Features

• Domain-Specific Q&A: Answers are based on a private collection of scientific documents, not just general LLM knowledge.

• RAG Pipeline: Combines retrieval from an external vector database with generative AI for enhanced accuracy and relevance.

• Streamlit Frontend: Clean, interactive interface for submitting queries and viewing detailed responses and source references.

• Modular Design: Separate components for document processing, embedding generation, vector storage, retrieval, and response generation.

• Efficient Information Retrieval: Quick similarity search powered by Pinecone's cloud-native vector database.

How It Works

1. Document Processing: A textbook PDF is loaded, segmented into coherent chunks (~500 characters each), and slightly overlapped for better context preservation.

2. Embedding Creation: Each chunk is transformed into a dense vector representation using Hugging Face's all-MiniLM-L6-v2 model.

3. Vector Storage: The embeddings are stored in a Pinecone index, enabling fast similarity searches.

4. Query Handling:

   • When a user submits a question, it is embedded into a vector.
   • The system retrieves the most semantically similar chunks from Pinecone.
   • Retrieved context is combined with the user’s question.

5. Response Generation:

   • A system prompt and the retrieved context guide the LLM (Microsoft's Phi-4 model via Hugging Face API) to generate a targeted answer.
   • The chatbot outputs the answer along with the source documents used.

Tech Stack

• LangChain: Build the RAG pipeline and retrieval chains
• Pinecon: Vector database for fast document retrieval
• Hugging Face (Phi-4, all-MiniLM-L6-v2): Embedding generation and language model responses
• Streamlit: Frontend for user interaction
• Python: Core application logic
• Jupyter Notebook: For experimentation and prototyping

Setup Instructions

1. Create a Virtual Environment:

   Open your terminal or command prompt and navigate to the project directory. Run the following command to create a virtual environment named chatbot:

   python -m venv chatbot

2. Select Python Interpreter in VS Code:

   • Open the project in VS Code.
   • Press Ctrl+Shift+P (or Cmd+Shift+P on macOS) to open the command palette.
   • Type "Python: Select Interpreter" and press Enter.
   • Select the Python interpreter from the chatbot virtual environment. It will typically be located at .\chatbot\Scripts\python.exe (Windows) or ./chatbot/bin/python (macOS/Linux).

3. Bypass Execution Policy (Windows Only):

   If you are using Windows, you might need to bypass the execution policy to run the activation script. Open PowerShell as an administrator and run:

   Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass

   Note: This command temporarily bypasses the execution policy for the current process. It is generally safe for setting up virtual environments.

4. Activate the Virtual Environment:

   • Windows:

     .\chatbot\Scripts\activate

   • macOS/Linux:

     source chatbot/bin/activate

   Your terminal prompt should now indicate that the virtual environment is active (e.g., (chatbot) C:\path\to\project>).

5. Install Dependencies:

   Install the required Python packages using pip from the requirements.txt file:

   pip install -r requirements.txt

6. (Optional) Recreate Pinecone Index:

   If you want to create new Pinecone index (it is already created), run the following command:

   python store_index.py

   Important: This step requires that you have your Pinecone API key and environment variables properly set up. Also, make sure to change the name of the variable "index_name" for your new index.

7. Run the Application:

   Start the Streamlit application by running:

   streamlit run app.py

   This will open the application in your default web browser.

Acknowledgments

• Textbook sourced from the Open Textbook Library.
• This project was developed as part of a software development course project exploring modern AI technologies and RAG architectures. The goal was to demonstrate practical skills in building intelligent, modular, and production-ready AI applications using Python ecosystems. Please see the report in the assets folder for more details.

---

Author: Nadiia Geras
Date: March 26, 2025