RAG-based Medical Chatbot

A Retrieval-Augmented Generation (RAG) based chatbot designed to provide accurate and personalized medical information to users.

🚀 Features

• Personalized Responses: Delivers tailored medical information based on user queries.
• Reliable Sources: Utilizes trusted medical databases to ensure information accuracy.
• Interactive Interface: Engages users through a conversational platform for seamless interaction.

🛠️ Installation

1. Clone the repository:

   git clone https://github.com/Raza-Aziz/RAG-based-Medical-Chatbot.git
   cd RAG-based-Medical-Chatbot

2. Set up a virtual environment:

   conda create -n medibot python=3.10 -y
   source env/bin/activate  # On Windows, use 'env\Scripts\activate'
   conda activate medibot   # Alternative to above line

3. Create a .env file in the root directory and add your Pinecone credentials as follows:

   PINECONE_API_KEY = "xxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
   GOOGLE_API_KEY = "xxxxxxxxxxxxxxxxxxxxxxxxxxxxx"

4. Install dependencies: All the dependencies are in requirements.txt, so you can install them all by simply running

   pip install -r requirements.txt

📌 Usage

1. Prepare the data:

   • Place your medical documents in the data/ directory. (PDF Files only!!)

2. Index the data: By running

   python store_index.py

3. Run the chatbot:

   python app.py

4. Access the chatbot interface:

   Open your web browser and navigate to http://localhost:8080.

🧐 Explanation

This chatbot follows the Retrieval-Augmented Generation (RAG) approach to provide accurate and context-aware medical responses:

1. Data Preparation:

   • Medical documents are stored in the data/ directory in PDF Format
   • These documents serve as the knowledge base for the chatbot.

2. Indexing Process (store_index.py):

   • Loads medical documents from the data/ folder using:
     • src.helper.load_pdf_file
   • Splits text into chunks of size 500 by using src.helper.text_split, which uses RecursiveCharacterTextSplitter
   • Downloads an embedding model from Hugging Face to convert text into vector representations:
     • sentence-transformers/all-MiniLM-L6-v2
   • Creates a Pinecone index to store embeddings.
   • Stores the document embeddings in Pinecone for efficient retrieval.

3. Chatbot Execution (app.py):

   • Loads the pre-built Pinecone index for retrieval:

     docsearch = PineconeVectorStore.from_existing_index(index_name, embeddings)

   • Sets up a retriever to fetch the most relevant medical document chunks:

     retriever = docsearch.as_retriever(search_type='similarity_score_threshold', search_kwargs={'k' : 3, 'score_threshold': 0.6})

   • Uses Google's Gemini AI model for response generation:

     llm = ChatGoogleGenerativeAI(model='gemini-2.0-flash', temperature=0.4, max_tokens=500)

   • Combines document retrieval and response generation into a pipeline:

   question_answer_chain = create_stuff_documents_chain(llm, prompt)
   rag_chain = create_retrieval_chain(retriever, question_answer_chain)

   • Defines an API endpoint to handle chat requests and return AI-generated responses.

4. User Interaction:

   • Users can interact with the chatbot via a web interface available at http://localhost:8080.
   • The chatbot retrieves relevant medical information and generates responses using RAG techniques.

🖼️ Screenshots

Feature	Screenshot
Chat Interface
Response Example

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

👨‍💻 Contact

For any questions, feel free to reach out:

• GitHub: @Raza-Aziz
• Email: razaaziz9191@gmail.com