Upload 3 files

app.py

@@ -0,0 +1,131 @@
+import streamlit as st
+from pdf_reader import *
+
+# Creating Session State Variable
+if 'API_Key' not in st.session_state:
+    st.session_state['API_Key'] = ''
+if 'Pinecone_API_Key' not in st.session_state:
+    st.session_state['Pinecone_API_Key'] =''
+if 'summary' not in st.session_state:
+    st.session_state.summary = ''
+if 'history' not in st.session_state:
+    st.session_state.history = {}
+if 'chat' not in st.session_state:
+    st.session_state.chat = ''
+if 'counter' not in st.session_state:
+    st.session_state.counter = 1
+
+st.title('PDF Chat Bot') 
+
+#********SIDE BAR Funtionality started*******
+
+# Sidebar to capture the API keys
+st.session_state['API_Key'] = st.sidebar.text_input("What's your OPENAI API key?",type="password")
+# File uploader widget
+uploaded_file = st.sidebar.file_uploader("Choose a PDF file", type=["pdf"])
+
+load_button = st.sidebar.button("UPLOAD", key="load_button")
+
+#If the bove button is clicked, pushing the data to Pinecone...
+if load_button:
+    #Proceed only if API keys are provided
+    if st.session_state['API_Key'] != '' and uploaded_file is not None:
+        file = save_pdf(uploaded_file)
+        file = "uploaded.pdf"
+        st.session_state.summary = load_db_sum(file, st.session_state['API_Key'])
+        st.session_state.chat = load_db(file, st.session_state['API_Key'])
+        st.session_state.history = {}
+
+    elif st.session_state['API_Key'] == '':
+        st.sidebar.error("Please enter your OpenAI API key.")
+    elif uploaded_file is None:
+        st.sidebar.error("Please attach a PDF file.")
+
+#********SIDE BAR Funtionality ended*****
+        
+
+if st.session_state['API_Key'] != '' and uploaded_file is not None:
+    file = "uploaded.pdf"
+    st.markdown('<br>', unsafe_allow_html=True)
+    st.markdown("#### **Summary**")
+    st.markdown('<hr style="margin: -10px 0; border-top: 1px solid black;">', unsafe_allow_html=True)
+    st.write(st.session_state.summary)
+
+    # create a variable for the chat
+    conversation = {}
+
+    #Captures User Inputs
+    user_input = st.text_input('Ask about the PDF',key="prompt")  # The box for the text prompt
+    # document_count = st.slider('No.Of links to return 🔗 - (0 LOW || 5 HIGH)', 0, 5, 2,step=1)
+
+    submit = st.button("SUBMIT") 
+
+    if submit:
+        #Proceed only if API keys are provided
+        if st.session_state.summary == '':
+            st.error("Please upload the PDF file.")
+        
+        # user_input = request.form['user_input']
+        else:
+            result = st.session_state.chat({"question": user_input})
+            answer_text = str(result['answer'])
+            question_text = str(result['question'])
+            user = "User"
+            chatbot = "Chat Bot"
+            conversation.update({user: question_text, chatbot: answer_text})
+
+            user_hist = f"[{st.session_state.counter}] {user}"
+            chat_hist = f"[{st.session_state.counter}] {chatbot}"
+            st.session_state.history.update({user_hist : question_text})
+            st.session_state.history.update({chat_hist : answer_text})
+            st.session_state.counter += 1
+
+
+            st.markdown('<br>', unsafe_allow_html=True)
+            st.markdown("#### **Conversation**")
+            st.markdown('<hr style="margin: -10px 0; border-top: 1px solid black;">', unsafe_allow_html=True)
+
+            table_data = list(conversation.items())
+
+            # Display the table with keys bolded using HTML
+            html_table = """
+                <style>
+                    table, tr {border:hidden;}
+                    table, td {border:hidden;}
+                </style>
+                <table><tr><th><strong></strong></th><th></th></tr>
+                """
+            for key, value in table_data:
+                html_table += f"<tr><td style='width: 90px;'><strong>{key}:</strong></td><td>{value}</td></tr>"
+            html_table += "</table>"
+            st.markdown(html_table, unsafe_allow_html=True)
+
+
+            st.markdown('<br>', unsafe_allow_html=True)
+            st.markdown("#### **Chat History**")
+            st.markdown('<hr style="margin: -10px 0; border-top: 1px solid black;">', unsafe_allow_html=True)
+            
+            table_data2 = list(st.session_state.history.items())
+
+            # Display the table with keys bolded using HTML
+            html_table = """
+                <style>
+                    table, tr {border:hidden;}
+                    table, td {border:hidden;}
+                </style>
+                <table><tr><th><strong></strong></th><th></th></tr>
+                """
+            for key, value in table_data2:
+                key = key[4:]
+                html_table += f"<tr><td style='width: 90px;'><strong>{key}:</strong></td><td>{value}</td></tr>"
+            html_table += "</table>"
+            st.markdown(html_table, unsafe_allow_html=True)
+
+elif st.session_state['API_Key'] == '':
+    st.error("Please enter your OpenAI API key.")
+elif uploaded_file is None:
+    st.session_state.summary = ''
+    st.error("Please upload the PDF file.")
+
+
+   

pdf_reader.py

@@ -0,0 +1,121 @@
+# Imports
+from langchain.embeddings.openai import OpenAIEmbeddings
+from langchain.text_splitter import CharacterTextSplitter, RecursiveCharacterTextSplitter
+from langchain.vectorstores import DocArrayInMemorySearch
+from langchain.chains.summarize import load_summarize_chain
+from langchain.chains import RetrievalQA,  ConversationalRetrievalChain
+from langchain.memory import ConversationBufferMemory
+from langchain.chat_models import ChatOpenAI
+from langchain.document_loaders import TextLoader
+from langchain.document_loaders import PyPDFLoader
+from langchain.prompts import PromptTemplate
+from langchain.llms import OpenAI
+import tiktoken 
+import os
+import sys
+sys.path.append('../..')
+
+import datetime
+current_date = datetime.datetime.now().date()
+if current_date < datetime.date(2023, 9, 2):
+    llm_name = "gpt-3.5-turbo-0301"
+else:
+    llm_name = "gpt-3.5-turbo"
+
+def load_db(file, api_key):
+    os.environ['OPENAI_API_KEY'] = api_key
+    # load documents
+    loader = PyPDFLoader(file)
+    # loader = file
+    documents = loader.load()
+    # documents = loader.read()
+    # split documents
+    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=150)
+    docs = text_splitter.split_documents(documents)
+    # define embedding
+    embeddings = OpenAIEmbeddings()
+    # create vector database from data
+    db = DocArrayInMemorySearch.from_documents(docs, embeddings)
+    
+    # add in the prompt
+    prompt_template_doc = """
+
+    Use the following pieces of context to answer the question at the end. {context}
+    You can also look into chat history. {chat_history}
+    If you still can't find the answer, please respond: "Please ask a question related to the document."
+
+    Question: {question}
+    Answer:
+    """
+    prompt_doc = PromptTemplate(
+        template=prompt_template_doc,
+        input_variables=["context", "question", "chat_history"],
+    )
+    # define retriever
+    retriever = db.as_retriever(search_type="similarity", search_kwargs={"k": 4})
+    # keeps a buffer of history and process it
+    memory = ConversationBufferMemory(
+        memory_key="chat_history",
+        output_key="answer",
+        return_messages=True
+    )
+    # create a chatbot chain
+    qa = ConversationalRetrievalChain.from_llm(
+        llm=ChatOpenAI(model_name=llm_name, temperature=0), 
+        chain_type="stuff", 
+        retriever=retriever,
+        combine_docs_chain_kwargs={"prompt": prompt_doc},
+        memory=memory
+    )
+    return qa 
+
+def load_db_sum(file, api_key):
+    os.environ['OPENAI_API_KEY'] = api_key
+    # load documents
+    loader = PyPDFLoader(file)
+    # loader = file
+    documents = loader.load()
+    # documents = loader.read()
+    # split documents
+    text_splitter = RecursiveCharacterTextSplitter(chunk_size=4000, chunk_overlap=150)
+    docs = text_splitter.split_documents(documents)
+    # create string of documents
+    str_docs = str(documents)
+
+    # define number of tokens from text
+    def num_tokens_from_string(string: str, encoding_name: str) -> int:    
+        encoding = tiktoken.encoding_for_model(encoding_name)
+        num_tokens = len(encoding.encode(string))
+        return num_tokens
+    
+    # get tokens
+    num_tokens = num_tokens_from_string(str_docs, llm_name)
+    model_max_tokens = 4097
+    # define embedding
+    embeddings = OpenAIEmbeddings()
+    # create vector database from data
+    db = DocArrayInMemorySearch.from_documents(docs, embeddings)
+    # define retriever
+    retriever = db.as_retriever(search_type="similarity", search_kwargs={"k": 4})
+    #Keeps a buffer of history and process it
+    memory = ConversationBufferMemory(
+        memory_key="chat_history",
+        output_key="answer",
+        return_messages=True
+    )
+
+    # create a chatbot chain based on tokens
+    if num_tokens < model_max_tokens:
+        chain = load_summarize_chain(llm=OpenAI(temperature=0, model="text-davinci-003", openai_api_key=api_key), chain_type="stuff")
+        qa = chain.run(documents)
+    else:
+        chain = load_summarize_chain(llm=OpenAI(temperature=0, model="text-davinci-003", openai_api_key=api_key), chain_type="map_reduce")
+        qa = chain.run(documents)
+
+    return qa 
+
+def save_pdf(pdf_file):
+    with open("uploaded.pdf", "wb") as file:
+        file.write(pdf_file.getvalue())
+    file = "uploaded.pdf"
+    return file

requirements.txt

@@ -0,0 +1,5 @@
+langchain
+tiktoken
+os
+sys
+streamlit