app.py

import gradio as gr
from langchain.document_loaders import PyPDFLoader
from langchain.text_splitter import CharacterTextSplitter
from langchain.embeddings import SentenceTransformerEmbeddings
from langchain.vectorstores import FAISS
from langchain.memory import ConversationBufferMemory
from groq import Groq
import requests
from bs4 import BeautifulSoup

client = Groq(api_key="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")
embedding_model = SentenceTransformerEmbeddings(model_name="sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2")

memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)

def process_pdf_with_langchain(pdf_path):

    loader = PyPDFLoader(pdf_path)
    documents = loader.load()
    text_splitter = CharacterTextSplitter(chunk_size=500, chunk_overlap=50)
    split_documents = text_splitter.split_documents(documents)

    vectorstore = FAISS.from_documents(split_documents, embedding_model)
    retriever = vectorstore.as_retriever(search_kwargs={"k": 3})
    return retriever

SERPAPI_KEY = "8a20e83850a3be0a0b4e3aed98bd3addbad56e82d52e639e1a692a02d021bca1"

def scrape_google_search(query, num_results=3):
    params = {
        "q": query,
        "hl": "fa",
        "gl": "ir",
        "num": num_results,
        "api_key": SERPAPI_KEY,
    }
    search = GoogleSearch(params)
    results = search.get_dict()

    if "error" in results:
        return f"Error: {results['error']}"

    search_results = []
    for result in results.get("organic_results", []):
        title = result.get("title", "No Title")
        link = result.get("link", "No Link")
        search_results.append(f"{title}: {link}")
    return "\n".join(search_results) if search_results else "No results found"

def generate_response(query, retriever=None, use_web_search=False):

    knowledge = ""

    if retriever:
        relevant_docs = retriever.get_relevant_documents(query)
        knowledge += "\n".join([doc.page_content for doc in relevant_docs])

    if use_web_search:
        web_results = scrape_google_search(query)
        knowledge += f"\n\nWeb Search Results:\n{web_results}"

    chat_history = memory.load_memory_variables({}).get("chat_history", "")
    context = (
        f"This is a conversation with ParvizGPT, an AI model designed by Amir Mahdi Parviz from Kermanshah University of Technology (KUT), "
        f"to help with tasks like answering questions in Persian, providing recommendations, and decision-making."
    )
    if knowledge:
        context += f"\n\nRelevant Knowledge:\n{knowledge}"
    if chat_history:
        context += f"\n\nChat History:\n{chat_history}"

    context += f"\n\nYou: {query}\nParvizGPT:"

    chat_completion = client.chat.completions.create(
        messages=[{"role": "user", "content": context}],
        model="llama-3.3-70b-versatile",
    )
    response = chat_completion.choices[0].message.content.strip()

    memory.save_context({"input": query}, {"output": response})
    return response

def gradio_interface(user_message, chat_box, pdf_file=None, enable_web_search=False):
    global retriever
    if pdf_file is not None:
        try:
            retriever = process_pdf_with_langchain(pdf_file.name)
        except Exception as e:
            return chat_box + [("Error", f"Error processing PDF: {e}")]

    response = generate_response(user_message, retriever=retriever, use_web_search=enable_web_search)
    chat_box.append(("You", user_message))
    chat_box.append(("ParvizGPT", response))
    return chat_box

def clear_memory():
    memory.clear()
    return []

retriever = None
with gr.Blocks() as interface:
    gr.Markdown("## ParvizGPT")
    # with gr.Row():
    chat_box = gr.Chatbot(label="Chat History", value=[])

# with gr.Row():
    user_message = gr.Textbox(
        label="Your Message",
        placeholder="Type your message here and press Enter...",
        lines=1,
        interactive=True,
    )
    enable_web_search = gr.Checkbox(label="🌐Enable Web Search", value=False)

# with gr.Row():
    clear_memory_btn = gr.Button("Clear Memory", interactive=True)
    # enable_web_search = gr.Checkbox(label="🌐Enable Web Search", value=False, interactive=True)
    pdf_file = gr.File(label="Upload PDF for Context (Optional)", type="filepath", interactive=True , scale=1)

    submit_btn = gr.Button("Submit")
    submit_btn.click(gradio_interface, inputs=[user_message, chat_box, pdf_file, enable_web_search], outputs=chat_box)
    user_message.submit(gradio_interface, inputs=[user_message, chat_box, pdf_file, enable_web_search], outputs=chat_box)
    clear_memory_btn.click(clear_memory, inputs=[], outputs=chat_box)

interface.launch()