Upload 3 files

main.py

@@ -0,0 +1,122 @@
+from util import *
+from streamlit_option_menu import option_menu
+from langchain_core.prompts import ChatPromptTemplate
+from langchain_groq import ChatGroq
+from dotenv import load_dotenv
+
+# --- PAGE CONFIGURATION ---
+st.set_page_config(page_title="Doc Chat", page_icon=":robot_face:", layout="centered")
+
+# --- SETUP SESSION STATE VARIABLES ---
+if "vector_store" not in st.session_state:
+    st.session_state.vector_store = False
+if "response" not in st.session_state:
+    st.session_state.response = None
+if "prompt_activation" not in st.session_state:
+    st.session_state.prompt_activation = False
+if "conversation" not in st.session_state:
+    st.session_state.conversation = None
+if "chat_history" not in st.session_state:
+    st.session_state.chat_history = None
+if "prompt" not in st.session_state:
+    st.session_state.prompt = False
+
+load_dotenv()
+
+# --- SIDEBAR CONFIGURATION ---
+st.sidebar.header('Configuration')
+groq_api_key = sidebar_api_key_configuration()
+model = sidebar_groq_model_selection()
+
+# --- MAIN PAGE CONFIGURATION ---
+st.title("Doc Chat :robot_face:")
+st.write("*Interrogate Documents :books:, Ignite Insights: AI at Your Service*")
+st.write(':blue[***Powered by Groq AI Inference Technology***]')
+
+# ---- NAVIGATION MENU -----
+selected = option_menu(
+    menu_title=None,
+    options=["Doc Chat", "Reference", "About"],
+    icons=["robot", "bi-file-text-fill", "app"],  # https://icons.getbootstrap.com
+    orientation="horizontal",
+)
+
+llm = ChatGroq(groq_api_key=groq_api_key, model_name=model)
+
+prompt = ChatPromptTemplate.from_template(
+    """
+    Answer the question based on the provided context only. If question is not within the context, do not try to answer
+    and respond that the asked question is out of context or something similar.
+    Please provide the most accurate response based on the question.
+    <context>
+    {context}
+    Questions: {input}
+    """
+)
+# ----- SETUP Doc Chat MENU ------
+if selected == "Doc Chat":
+    st.subheader("Upload PDF(s)")
+    pdf_docs = st.file_uploader("Upload your PDFs", type=['pdf'], accept_multiple_files=True,
+                                disabled=not st.session_state.prompt_activation, label_visibility='collapsed')
+    process = st.button("Process", type="primary", key="process", disabled=not pdf_docs)
+
+    if process:
+        with st.spinner("Processing ..."):
+            st.session_state.vector_store = create_vectorstore(pdf_docs)
+            st.session_state.prompt = True
+            st.success('Database is ready')
+
+    st.divider()
+
+    if "messages" not in st.session_state:
+        st.session_state["messages"] = [{"role": "assistant", "content": "How can I help you?"}]
+
+    for msg in st.session_state.messages:
+        st.chat_message(msg["role"]).write(msg["content"])
+
+    container = st.container(border=True)
+    if question := st.chat_input(placeholder='Enter your question related to uploaded document',
+                                 disabled=not st.session_state.prompt):
+        st.session_state.messages.append({"role": "user", "content": question})
+        st.chat_message("user").write(question)
+
+        with st.spinner('Processing...'):
+            st.session_state.response = get_llm_response(llm, prompt, question)
+            st.session_state.messages.append({"role": "assistant", "content": st.session_state.response['answer']})
+            st.chat_message("assistant").write(st.session_state.response['answer'])
+
+# ----- SETUP REFERENCE MENU ------
+if selected == "Reference":
+    st.title("Reference & Context")
+    if st.session_state.response is not None:
+        for i, doc in enumerate(st.session_state.response["context"]):
+            with st.expander(f'Reference # {i + 1}'):
+                st.write(doc.page_content)
+
+# ----- SETUP ABOUT MENU ------
+if selected == "About":
+    with st.expander("About this App"):
+        st.markdown(''' This app allows you to chat with your PDF documents. It has following functionality:
+
+    - Allows to chat with multiple PDF documents
+    - Support of Groq AI inference technology 
+    - Display the response context and document reference
+
+        ''')
+    with st.expander("Which Large Language models are supported by this App?"):
+        st.markdown(''' This app supports the following LLMs as supported by Groq:
+
+    - Chat Models -- Groq
+        - Llama3-8b-8192 
+        - Llama3-70b-8192 
+        - Mixtral-8x7b-32768
+        - Gemma-7b-it
+        ''')
+
+    with st.expander("Which library is used for vectorstore?"):
+        st.markdown(''' This app supports the FAISS for AI similarity search and vectorstore:
+        ''')
+
+    with st.expander("Whom to contact regarding this app?"):
+        st.markdown(''' Contact [Sree Narayanan]([email protected])
+        ''')

requirements.txt

@@ -0,0 +1,11 @@
+pypdf
+langchain
+langchain-core
+langchain-groq
+langchain-community
+streamlit
+streamlit-option-menu
+python-dotenv
+boto3
+faiss-cpu
+gpt4all

util.py

@@ -0,0 +1,101 @@
+from pypdf import PdfReader
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain_community.embeddings import HuggingFaceInferenceAPIEmbeddings
+from langchain_community.embeddings.ollama import OllamaEmbeddings
+from langchain_community.embeddings.bedrock import BedrockEmbeddings
+from langchain_community.vectorstores import FAISS
+from langchain.chains.combine_documents import create_stuff_documents_chain
+from langchain.chains import create_retrieval_chain
+from dotenv import load_dotenv
+import streamlit as st
+import os
+
+load_dotenv()
+
+
+# Function to get the API key
+def get_api_key():
+    # Try to get the API key from st.secrets first
+    try:
+        groq_api_key = os.getenv("GROQ_API_KEY", "")
+        
+        return groq_api_key
+    except Exception as e:
+        print(e)
+
+def get_inference_api_key():
+    try:
+        inference_api_key = os.getenv("INFERENCE_API_KEY", "")
+
+        return inference_api_key
+    except Exception as e:
+        print(e)
+
+
+# Function for API configuration at sidebar
+def sidebar_api_key_configuration():
+    groq_api_key = get_api_key()
+    if groq_api_key == '':
+        st.sidebar.warning('Enter the API Key(s) 🗝️')
+        st.session_state.prompt_activation = False
+    elif (groq_api_key.startswith('gsk_') and (len(groq_api_key) == 56)):
+        st.sidebar.success('Lets Proceed!', icon='️👉')
+        st.session_state.prompt_activation = True
+    else:
+        st.sidebar.warning('Please enter the correct API Key 🗝️!', icon='⚠️')
+        st.session_state.prompt_activation = False
+    return groq_api_key
+
+
+def sidebar_groq_model_selection():
+    st.sidebar.subheader("Model Selection")
+    model = st.sidebar.selectbox('Select the Model', ('Llama3-8b-8192', 'Llama3-70b-8192', 'Mixtral-8x7b-32768',
+                                                      'Gemma-7b-it'), label_visibility="collapsed")
+    return model
+
+
+# Read PDF data
+def read_pdf_data(pdf_docs):
+    text = ""
+    for pdf in pdf_docs:
+        pdf_reader = PdfReader(pdf)
+        for page in pdf_reader.pages:
+            text += page.extract_text()
+    return text
+
+
+# Split data into chunks
+def split_data(text):
+    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
+    text_chunks = text_splitter.split_text(text)
+    return text_chunks
+
+
+def get_embedding_function():
+    # embeddings = BedrockEmbeddings(
+    #     credentials_profile_name="default", region_name="us-east-1"
+    # )
+    #embeddings = OllamaEmbeddings(model="nomic-embed-text")
+    inference_api_key = get_inference_api_key()
+
+    embeddings = HuggingFaceInferenceAPIEmbeddings(
+        api_key=inference_api_key, model_name="sentence-transformers/all-MiniLM-l6-v2"
+    )
+    return embeddings
+
+
+# Create vectorstore
+def create_vectorstore(pdf_docs):
+    raw_text = read_pdf_data(pdf_docs)  # Get PDF text
+    text_chunks = split_data(raw_text)  # Get the text chunks
+    embeddings = get_embedding_function()  # Get the embedding function
+    vectorstore = FAISS.from_texts(texts=text_chunks, embedding=embeddings)
+    return vectorstore
+
+
+# Get response from llm of user asked question
+def get_llm_response(llm, prompt, question):
+    document_chain = create_stuff_documents_chain(llm, prompt)
+    retrieval_chain = create_retrieval_chain(st.session_state.vector_store.as_retriever(), document_chain)
+    response = retrieval_chain.invoke({'input': question})
+    return response