---

🧠 Medical RAG Chatbot using LangChain + Llama + PubMedBERT

This project is a Retrieval-Augmented Generation (RAG) pipeline that allows users to query medical PDFs and get context-aware, accurate answers using a local LLM. It uses LangChain, SentenceTransformers, Chroma vector store, and the BioMistral-7B model.

---

🚀 Features

• 📄 Upload and parse PDF documents
• 🔍 Perform semantic search over embedded medical content
• 🤖 Generate medically-aligned answers using Llama-based LLM
• 🧠 Uses PubMedBERT for biomedical embeddings

---

🛠️ Installation

Install required libraries (recommended to run in Google Colab):

pip install langchain sentence-transformers chromadb llama-cpp-python langchain_community pypdf

---

📁 File Structure

• /content/drive/MyDrive/Colab Notebooks/ → Your PDF files
• /content/drive/MyDrive/BioMistral-7B.Q2_K.gguf → Your LLM file

---

⚙️ Setup Steps

1. Mount Google Drive:

   from google.colab import drive
   drive.mount("/content/drive")

2. Load PDFs and Split Text:

   from langchain_community.document_loaders import PyPDFDirectoryLoader
   ...
   docs = loader.load()

3. Generate Embeddings using PubMedBERT:

   from langchain_community.embeddings import SentenceTransformerEmbeddings
   embedding = SentenceTransformerEmbeddings(model_name="NeuML/pubmedbert-base-embeddings")

4. Build Vector Store with Chroma:

   vectorstore = Chroma.from_documents(chunks, embedding)

5. Load the Language Model:

   from langchain_community.llms import LlamaCpp
   llm = LlamaCpp(model_path=".../BioMistral-7B.Q2_K.gguf", ...)

6. Start Chat Interface:

   while True:
       user_input = input("Input query: ")
       ...

---

🧬 Example Query

Input query: Who is at risk of heart disease?
Answer: [contextual answer based on the uploaded PDFs]

---

📦 LLM Model Download

You can download the BioMistral 7B GGUF model (compatible with llama-cpp-python) here:

• 🔗 BioMistral-7B.Q2_K.gguf (via Hugging Face)

Recommended model file: BioMistral-7B.Q2_K.gguf

Place the model in your Google Drive:
/content/drive/MyDrive/BioMistral-7B.Q2_K.gguf

---

🔐 Notes

• Don't forget to add your Hugging Face token if using private models or APIs:

  os.environ["HUGGINGFACEHUB_API_TOKEN"] = "your_token_here"

• Never upload private tokens to public repositories.

---

💡 Inspiration

This project combines:

• Biomedical language understanding (via PubMedBERT)
• Efficient document-based QA (via LangChain RAG pipeline)
• Local inference with LLMs (Llama.cpp)

---