app.py

import gradio as gr
from huggingface_hub import InferenceClient
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS
from langchain.text_splitter import CharacterTextSplitter
from langchain_community.document_loaders import PyPDFLoader
import os

# Load the model client
client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")

# Initialize vector store
vector_store = None

# Preload and process the PDF document
PDF_PATH = "generalsymptoms.pdf"  # Path to the pre-defined PDF document

def preload_pdf():
    global vector_store

    # Load PDF and extract text
    loader = PyPDFLoader(PDF_PATH)
    documents = loader.load()

    # Split the text into smaller chunks for retrieval
    text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
    docs = text_splitter.split_documents(documents)

    # Compute embeddings for the chunks
    embeddings = HuggingFaceEmbeddings()
    vector_store = FAISS.from_documents(docs, embeddings)

    print(f"PDF '{PDF_PATH}' loaded and indexed successfully.")

# Response generation
def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    global vector_store

    if vector_store is None:
        return "The PDF document is not loaded. Please check the code setup."

    # Retrieve relevant chunks from the PDF
    relevant_docs = vector_store.similarity_search(message, k=3)
    context = "\n".join([doc.page_content for doc in relevant_docs])

    # Combine system message, context, and user message
    full_system_message = (
        f"{system_message}\n\nContext from the document:\n{context}\n\n"
    )

    messages = [{"role": "system", "content": full_system_message}]

    for val in history:
        if val[0]:
            messages.append({"role": "user", "content": val[0]})
        if val[1]:
            messages.append({"role": "assistant", "content": val[1]})

    messages.append({"role": "user", "content": message})

    response = ""

    for message in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = message.choices[0].delta.content
        response += token
        yield response

# Gradio interface

   

    demo = gr.ChatInterface(
        respond,
        additional_inputs=[
            gr.Textbox(
                value=(
                    "You are going to act like a medical practitioner. Hear the symptoms, "
                    "diagnose the disease, mention the disease name as heading, and suggest tips "
                    "to overcome the issue. Base your answers on the provided document. limit the response to 3-4 sentences. list out the response point by point" 
                ), visible=False,
                label="System message",
            ),
            gr.Slider(minimum=1, maximum=2048, value=512, step=1,visible=False, label="Max new tokens"),
            gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, visible=False,label="Temperature"),
            gr.Slider(
                minimum=0.1,
                maximum=1.0,
                value=0.95,
                step=0.05,visible=False,
                label="Top-p (nucleus sampling)",
            ),
        ],
        examples=[
            ["I feel stressed."],
            ["Can you guide me through quick health tips?"],
            ["How do I stop worrying about things I can't control?"]
        ],
        title = "Diagnify 🕊️"
    )

    if __name__ == "__main__":
        preload_pdf()
        demo.launch()