Getting Started

Follow this tutorial to understand how the Medical AI Chatbot processes medical documents, retrieves relevant information, and generates responses using a Retrieval-Augmented Generation (RAG) pipeline.

← Back to Home

Prepare the Medical Documents

Place the medical PDF documents that will form the chatbot's knowledge base in the project's document directory.

Use reliable, authoritative medical sources and keep the document collection focused on the intended use case.

Extract and Split Text

The application reads the PDF files and extracts their text. Large documents are then divided into smaller chunks so that relevant sections can be retrieved efficiently.

PDF
 ↓
Text extraction
 ↓
Text splitting
 ↓
Document chunks

Create Embeddings

Each document chunk is converted into an embedding — a numerical representation that captures the semantic meaning of the text.

These vector representations allow the system to identify information that is semantically similar to a user's question.

Store the Vectors

The generated embeddings are stored in Pinecone together with the associated document information.

Document chunk
      ↓
Embedding model
      ↓
Vector
      ↓
Pinecone index

Ask a Question

The user enters a medical-related question through the chatbot interface.

For example:

What are the common symptoms of iron deficiency?

Retrieve Relevant Information

The user's question is converted into an embedding and compared with the vectors stored in Pinecone.

The most relevant document chunks are retrieved and supplied as context for the language model.

Generate the Answer

LangChain coordinates the retrieval and generation process. The retrieved context is provided to GPT-4o, which generates the final natural-language response.

User question
      ↓
Retriever
      ↓
Relevant documents
      ↓
Prompt + context
      ↓
GPT-4o
      ↓
Answer

Run the Application

The Flask application provides the web interface and backend API used by the chatbot.

python app.py

Once the Flask server is running, open the application in a browser and submit a question through the chat interface.

Deployment

For deployment, the application can be packaged using Docker and deployed to an AWS environment. GitHub Actions can automate the build and deployment workflow.

git push
   ↓
GitHub Actions
   ↓
Build Docker image
   ↓
Deploy
   ↓
Running application

This chatbot is intended to provide information based on its configured knowledge base. It should not be treated as a substitute for professional medical advice, diagnosis, or emergency medical care.

← Back to Home