Upload 3 files

app.py

@@ -0,0 +1,54 @@
+import streamlit as st 
+from streamlit_chat import message
+from helper import get_qa_chain, create_vector_db
+
+st.set_page_config(layout="wide",page_title="Chat with PDF")
+
+
+def process_answer(instruction):
+    response = ''
+    instruction = instruction
+    qa = get_qa_chain()
+    generated_text = qa(instruction)
+    answer = generated_text['result']
+    return answer
+
+# Display conversation history using Streamlit messages
+def display_conversation(history):
+    for i in range(len(history["generated"])):
+        message(history["past"][i], is_user=True, key=str(i) + "_user")
+        message(history["generated"][i],key=str(i))
+
+def main():
+    st.header("Chat with your PDF")
+    create_embeddings = st.button("Create Embeddings")
+
+    if create_embeddings:
+        with st.spinner('Embeddings are in process...'):
+            create_vector_db()
+        st.success('Embeddings are created successfully!')
+
+    st.subheader("Chat Here")
+    user_input = st.text_input("",key="input")
+
+    #initialize session state for generted response and past messages
+    if "generated" not in st.session_state:
+        st.session_state["generated"] = ["I am an AI assitance how can I help?"]
+
+    if "past" not in st.session_state:
+        st.session_state["past"] = ["Hey there!"]
+    # Search the database for a response based on user input and update session state
+    if user_input:
+        answer = process_answer({'query': user_input})
+        st.session_state["past"].append(user_input)
+        response = answer
+        st.session_state["generated"].append(response)
+
+    # Display conversation history using Streamlit messages
+    if st.session_state["generated"]:
+        display_conversation(st.session_state)
+
+        
+
+if __name__ == "__main__":
+    main()

helper.py

@@ -0,0 +1,96 @@
+import os
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM 
+from transformers import pipeline
+import torch 
+import textwrap 
+from PyPDF2 import PdfReader 
+from typing_extensions import Concatenate
+from langchain.text_splitter import CharacterTextSplitter 
+from langchain_community.embeddings import HuggingFaceInstructEmbeddings 
+from langchain.vectorstores import Chroma 
+from langchain.llms import HuggingFacePipeline
+from langchain.chains import RetrievalQA
+from langchain.chains.conversation.memory import ConversationBufferMemory
+from langchain import PromptTemplate
+from dotenv import load_dotenv
+
+load_dotenv()  # take environment variables from .env
+os.environ["LANGCHAIN_API_KEY"] = str(os.getenv("LANGCHAIN_API_KEY"))
+os.environ["LANGCHAIN_TRACING_V2"] = "true"
+os.environ["LANGCHAIN_ENDPOINT"] = "https://api.smith.langchain.com"
+os.environ["LANGCHAIN_PROJECT"] = "2.pdf_chat_router_issue_assistant"
+
+from dotenv import load_dotenv
+load_dotenv()  # take environment variables from .env (especially openai api key)
+
+# Create LLM model
+# model = "C:/Users/arasu/Workspace/Projects/GenAI/models/MBZUAILaMini-Flan-T5-248M/"
+model = "MBZUAI/LaMini-Flan-T5-248M"
+tokenizer = AutoTokenizer.from_pretrained(model,truncation=True)
+base_model = AutoModelForSeq2SeqLM.from_pretrained(model)
+pipe = pipeline(
+        'text2text-generation',
+        model = base_model,
+        tokenizer = tokenizer,
+        max_length = 256,
+        do_sample = True,
+        temperature = 0.3,
+        top_p= 0.95
+    )
+llm = HuggingFacePipeline(pipeline=pipe)
+
+# # Initialize instructor embeddings using the Hugging Face model
+# instructor_embeddings = HuggingFaceInstructEmbeddings(model_name="C:/Users/arasu/Workspace/Projects/GenAI/embeddings/hkunlp_instructor-large")
+instructor_embeddings = HuggingFaceInstructEmbeddings(model_name="hkunlp/instructor-large")
+db_path = "vector_db"
+
+def create_vector_db():
+    # Load data from pdf
+    raw_text = ""
+    text_splitter = CharacterTextSplitter(
+        separator = "\n",
+        chunk_size = 500,
+        chunk_overlap  = 100,
+        length_function = len,
+    )
+    for root, dirs, files in os.walk("docs"):        
+        for file in files:
+            if file.endswith(".pdf"):
+                pdf = PdfReader("./docs/"+file)
+                for i, page in enumerate(pdf.pages):
+                    content = page.extract_text()
+                    if content:
+                        raw_text += content
+    texts = text_splitter.split_text(raw_text)
+
+    # Create a  vector database from 'text'
+    vector_db = Chroma.from_texts(texts,instructor_embeddings,persist_directory=db_path)
+    vector_db.persist()
+    vector_db = None 
+
+def get_qa_chain():
+    # Load the vector database from the local folder
+    vector_db = Chroma(persist_directory=db_path, embedding_function = instructor_embeddings)
+
+    # Create a retriever for querying the vector database
+    retriever = vector_db.as_retriever(search_kwargs={"k":3})
+
+    template = """
+    You are friendly customer care assistant trying to help user on the context provided.\
+    if the question contains greetings then greet the user back. be friendly.\
+    if the answer is not found in the context then reply "No Evidence Found".\
+    context: {context}
+    question: {question}
+    """
+    prompt = PromptTemplate(template=template, input_variables=["context", "question"])
+
+    chain_type_kwargs = {"prompt": prompt}
+    qa = RetrievalQA.from_chain_type(
+        llm = llm,
+        chain_type = "stuff",
+        retriever = retriever,
+        input_key="query",
+        return_source_documents=True,
+        chain_type_kwargs=chain_type_kwargs
+    )
+    return qa  

requirements.txt

@@ -0,0 +1,10 @@
+
+transformers
+langchain
+huggingface_hub
+sentence-transformers==2.2.2
+InstructorEmbedding
+chromadb
+PyPDF2
+streamlit
+streamlit_chat