Update app.py

app.py

@@ -2,11 +2,11 @@ import streamlit as st
 from dotenv import load_dotenv
 from PyPDF2 import PdfReader
 from langchain.text_splitter import CharacterTextSplitter
-from langchain_huggingface import HuggingFaceEmbeddings
+from langchain_huggingface import HuggingFaceEmbeddings, HuggingFacePipeline
 from langchain_community.vectorstores import FAISS
-from langchain_huggingface import HuggingFaceHub
 from langchain.memory import ConversationBufferMemory
 from langchain.chains import ConversationalRetrievalChain
+from transformers import pipeline  # Hugging Face pipeline for using T5 model
 import os
 
 
@@ -44,21 +44,18 @@ def get_vectorstore(text_chunks):
 
 # Function to create the conversation chain using T5 from Hugging Face API
 def get_conversation_chain(vectorstore):
-    llm = HuggingFaceHub(
-        repo_id="google/t5-large",  # Replace with your Hugging Face model ID
-        temperature=0.7,            # Adjust for creativity
-        max_new_tokens=100,         # Control response length
-        top_p=0.9,                  # Nucleus sampling for better variety
-        top_k=50,                   # Top-k filtering
-        repetition_penalty=1.0      # Default value to prevent repetitive outputs
-    )
+    # Use HuggingFacePipeline with a Hugging Face T5 model
+    t5_model = pipeline("text2text-generation", model="google/t5-large", tokenizer="google/t5-large")
+    
+    llm = HuggingFacePipeline(pipeline=t5_model)
+    
     memory = ConversationBufferMemory(
         memory_key="chat_history", return_messages=True
     )
     
-    # Create a conversation chain using the correct LLM (llm, not t5_model)
+    # Create a conversation chain using the T5 model
     conversation_chain = ConversationalRetrievalChain.from_llm(
-        llm=llm,  # Pass 'llm' instead of 't5_model'
+        llm=llm,
         retriever=vectorstore.as_retriever(),
         memory=memory,
     )