Bot to read our local PDF and answer using Openai, Embeddings, Langchain & Streamlit

Chat with PDF  | Chatbot Bio-query use case

In visual code Studio, creat directory usecase3 and creat a file search.py

in termial install the below libraries first

pip install chatpdf

pip install faiss-cpu  

pip install langchain

Steps Load + Split+Store+Get+Generate Bot + use Bot

Down load any PDF say test.pdf from which you want to ask. Copy to your current directory

Try this code first for load and split pdf

Create a file names search.py

#load and split pdf

from langchain.document_loaders import PyPDFLoader

loader = PyPDFLoader("test.pdf")

pages_content = loader.load_and_split()

print(len(pages_content), pages_content)

You will get dictionary output with filename, meta data, page no etc.

Now type as below and Run

#load pdf and split

from langchain.document_loaders import PyPDFLoader

loader = PyPDFLoader("test.pdf")

pages_content = loader.load_and_split()

#print(len(pages_content), pages_content)

# #refer openai api embeddings

from langchain.embeddings.openai import OpenAIEmbeddings

from langchain_openai import OpenAIEmbeddings

from langchain.vectorstores import FAISS

embeddings = OpenAIEmbeddings()

db = FAISS.from_documents(pages_content, embeddings)

db.save_local("faiss_index")  # to save local copy

nwdb = FAISS.load_local("faiss_index", embeddings,

                        allow_dangerous_deserialization=True)

query = "are there any educational qualification"

docs  =  nwdb.similarity_search(query) # to create similarity index

# print(docs)

from langchain.chains import RetrievalQA

from langchain.chat_models import ChatOpenAI

llm = ChatOpenAI()  # creat chatbot

qa_chain = RetrievalQA.from_chain_type(llm, retriever=nwdb.as_retriever())

res =qa_chain({"query" : "are there any educational qualification"})

print(res)

To get the results for your query. Here educational qualification related since i have take my bio-dat as pdf test file.

Now add ask query module in the above code as shown below:

#load pdf and split

from langchain.document_loaders import PyPDFLoader

# loader = PyPDFLoader("test.pdf")

# pages_content = loader.load_and_split()

# print(len(pages_content), pages_content)

# #refer openai api embeddings

from langchain.embeddings.openai import OpenAIEmbeddings

from langchain_openai import OpenAIEmbeddings

from langchain.vectorstores import FAISS

embeddings = OpenAIEmbeddings()

# two below lines are commented to avoid to create vectorbase again

# db = FAISS.from_documents(pages_content, embeddings)

# db.save_local("faiss_index")  # to save local vectorbase copy

# you can see files faiss_index and faiss

nwdb = FAISS.load_local("faiss_index", embeddings,

                        allow_dangerous_deserialization=True)

query = "are there any educational qualification"

docs  =  nwdb.similarity_search(query) # to create similarity index

# print(docs) # Note see faiss_index directory with files

from langchain.chains import RetrievalQA

from langchain.chat_models import ChatOpenAI

llm = ChatOpenAI()  # creat chatbot

qa_chain = RetrievalQA.from_chain_type(llm, retriever=nwdb.as_retriever())

res =qa_chain({"query" : "are there any educational qualification"})

# print(res)

# you can see faiss_index directory under which 2 files index.faiss and index.pkl

#addition for chatbot after app.py

def ask(user_query):

     res = qa_chain({"query": user_query})

     return res["result"]

Now create a front end file named as app.py with code as shown below

import streamlit as st

from search import ask

with st.chat_message("assitant"):

    st.write("Hello! May I helpYou ?")

prompt = st.chat_input("Type question")

if prompt:

    st.chat_message("user").markdown(prompt)

    response = ask(prompt)

    st.chat_message("assistant").markdown(response)

pip install streamlit in the terminal window

Save and Run as streamlit run app.py  to get streamlit.app tab in the browser 

Pl. double click the image to see clearly. 

Now type in the text window Educational qualification in bullet points  

and then press play button. You may get the screen as shown below:

Wow ! we have done a chatbot answering questions from my test.pdf

This is the simplest way with openai, embeddings, langchain and streamlit we could develope our custom made chatbot. This is the beginning. You can change code in app.py by refereinng streamlit documentation  to change your frontent and enjoy!!!😊😊😊😊😊😊😊😊

Please use your PDF and change queries as per your PDF.🎈