app.py

from dotenv import load_dotenv
import os
from docx import Document
from llama_index.llms.together import TogetherLLM
from llama_index.core.llms import ChatMessage, MessageRole
from Bio import Entrez
import ssl
from transformers import AutoModelForSequenceClassification, AutoTokenizer
import streamlit as st
from googleapiclient.discovery import build
from typing import List, Optional

load_dotenv()
# 995d5f1a8de125c5b39bb48c2613e85f57d53c0e498a87d1ff33f0ec89a26ec7
os.environ["TOGETHER_API"] = os.getenv("TOGETHER_API")
os.environ["GOOGLE_SEARCH_API_KEY"] = os.getenv("GOOGLE_SEARCH_API_KEY")

def search_pubmed(query: str) -> Optional[List[str]]:
    """
    Searches PubMed for a given query and returns a list of formatted results
    (or None if no results are found).
    """
    Entrez.email = "[email protected]"  # Replace with your email

    try:
        ssl._create_default_https_context = ssl._create_unverified_context

        handle = Entrez.esearch(db="pubmed", term=query, retmax=3)
        record = Entrez.read(handle)
        id_list = record["IdList"]

        if not id_list:
            return None

        handle = Entrez.efetch(db="pubmed", id=id_list, retmode="xml")
        articles = Entrez.read(handle)

        results = []
        for article in articles['PubmedArticle']:
            try:
                medline_citation = article['MedlineCitation']
                article_data = medline_citation['Article']
                title = article_data['ArticleTitle']
                abstract = article_data.get('Abstract', {}).get('AbstractText', [""])[0]

                result = f"**Title:** {title}\n**Abstract:** {abstract}\n"
                result += f"**Link:** https://pubmed.ncbi.nlm.nih.gov/{medline_citation['PMID']} \n\n"
                results.append(result)
            except KeyError as e:
                print(f"Error parsing article: {article}, Error: {e}")

        return results

    except Exception as e:
        print(f"Error accessing PubMed: {e}")
        return None

def chat_with_pubmed(article_text, article_link):
    """
    Engages in a chat-like interaction with a PubMed article using TogetherLLM.
    """
    try:
        llm = TogetherLLM(model="QWEN/QWEN1.5-14B-CHAT", api_key=os.environ['TOGETHER_API'])
        messages = [
            ChatMessage(role=MessageRole.SYSTEM, content="You are a helpful AI assistant summarizing and answering questions about the following medical research article: " + article_link),
            ChatMessage(role=MessageRole.USER, content=article_text)
        ]
        response = llm.chat(messages)
        return str(response) if response else "I'm sorry, I couldn't generate a summary for this article."
    except Exception as e:
        print(f"Error in chat_with_pubmed: {e}")
        return "An error occurred while generating a summary."

def search_web(query: str, num_results: int = 3) -> Optional[List[str]]:
    """
    Searches the web using the Google Search API and returns a list of formatted results
    (or None if no results are found).
    """
    try:
        service = build("customsearch", "v1", developerKey=os.environ["GOOGLE_SEARCH_API_KEY"])

        # Execute the search request
        res = service.cse().list(q=query, cx="e31a5857f45ef4d2a", num=num_results).execute()

        if "items" not in res:
            return None

        results = []
        for item in res["items"]:
            title = item["title"]
            link = item["link"]
            snippet = item["snippet"]
            result = f"**Title:** {title}\n**Link:** {link} \n**Snippet:** {snippet}\n\n"
            results.append(result)

        return results

    except Exception as e:
        print(f"Error performing web search: {e}")
        return None


from together import Together
def medmind_chatbot(user_input, chat_history=None):
    """
    Processes user input, interacts with various resources, and generates a response.
    Handles potential errors, maintains chat history,
    """
    if chat_history is None:
        chat_history = []

    response_parts = []  # Collect responses from different sources
    final_response = "";

    try:
        # PubMed Search and Chat
        pubmed_results = search_pubmed(user_input)
        if pubmed_results:
            for article_text in pubmed_results:
                title, abstract, link = article_text.split("\n")[:3]
                # print(article_text)
                response_parts.append(f"{title}\n{abstract}\n{link}\n")
        else:
            response_parts.append("No relevant PubMed articles found.")

        # Web Search
        web_results = search_web(user_input)
        if web_results:
            response_parts.append("\n\n**Web Search Results:**")
            response_parts.extend(web_results)
        else:
            response_parts.append("No relevant web search results found.")

        # Combine response parts into a single string
        response_text = "\n\n".join(response_parts)

        prompt = f"""You are a Health Assistant AI designed to provide detailed responses to health-related questions.
        Based on the information retrieved from the PubMed and Web Search below, answer the user's query appropriately.

        - If the user's query is health-related, provide a detailed and helpful response based on the retrieved information. Or if there is
        some previous conversation then answer the health by seeing the previous conversation also.
        - If the query is a general greeting (e.g., 'Hello', 'Hi'), respond as a friendly assistant.
        - If the query is irrelevant or unrelated to health, respond with: 'I am a health assistant. Please ask only health-related questions.'
        - Don't mention in response that where you reterived the information.
        Previous Conversation:
        {chat_history}

        User's Query: {user_input}

        Information retrieved from PubMed and Web Search:
        {response_text}

        Your response:"""

        client = Together(api_key=os.environ.get('TOGETHER_API'))

        response = client.chat.completions.create(
            model="meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo",
            messages=[{"role": "user", "content": prompt}],
        )

        final_response = response.choices[0].message.content

    except Exception as e:
        print(f"Error in chatbot: {e}")
        response_text = "An error occurred. Please try again later."

    chat_history.append((user_input, final_response))
    return final_response, chat_history

medmind_chatbot("What are the symptoms of COVID-19?")

import gradio as gr

def show_info_popup():
    info = """
    **HealthHive is an AI-powered chatbot designed to assist with medical information.**
    ...
    """
    return info


def main():
    # Initialize Gradio Interface
    with gr.Blocks() as demo:
        gr.Markdown("# HealthHive Chatbot")
        gr.Markdown("Ask your medical questions and get reliable information!")

        # Example Questions (Sidebar)
        gr.Markdown("### Example Questions")
        example_questions = [
            "What are the symptoms of COVID-19?",
            "How can I manage my diabetes?",
            "What are the potential side effects of ibuprofen?",
            "What lifestyle changes can help prevent heart disease?"
        ]
        for question in example_questions:
            gr.Markdown(f"- {question}")

        # Chat History and User Input
        with gr.Row():
            user_input = gr.Textbox(label="You:", placeholder="Type your medical question here...", lines=2)
            chat_history = gr.State([])

        # Output Container
        with gr.Row():
            response = gr.Textbox(label="HealthHive:", placeholder="Response will appear here...", interactive=False, lines=10)
        def clear_chat():
            return "", ""

        # Define function to update chat history and response
        def on_submit(user_input, chat_history):
            result, updated_history = medmind_chatbot(user_input, chat_history)
            info = show_info_popup()
            return result, updated_history, info

        # Link the submit button to the chatbot function
        gr.Button("Submit").click(on_submit, inputs=[user_input, chat_history], outputs=[response, chat_history])
        # gr.Button("Start New Chat").click(lambda: [], outputs=[chat_history])
        gr.Button("Start New Chat").click(clear_chat, outputs=[user_input, response])


    demo.launch()

if __name__ == "__main__":
    main()