Spaces:
Sleeping
Sleeping
Upload 3 files
Browse files- app.py +131 -0
- pdf_reader.py +121 -0
- requirements.txt +5 -0
app.py
ADDED
@@ -0,0 +1,131 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
from pdf_reader import *
|
3 |
+
|
4 |
+
# Creating Session State Variable
|
5 |
+
if 'API_Key' not in st.session_state:
|
6 |
+
st.session_state['API_Key'] = ''
|
7 |
+
if 'Pinecone_API_Key' not in st.session_state:
|
8 |
+
st.session_state['Pinecone_API_Key'] =''
|
9 |
+
if 'summary' not in st.session_state:
|
10 |
+
st.session_state.summary = ''
|
11 |
+
if 'history' not in st.session_state:
|
12 |
+
st.session_state.history = {}
|
13 |
+
if 'chat' not in st.session_state:
|
14 |
+
st.session_state.chat = ''
|
15 |
+
if 'counter' not in st.session_state:
|
16 |
+
st.session_state.counter = 1
|
17 |
+
|
18 |
+
st.title('PDF Chat Bot')
|
19 |
+
|
20 |
+
#********SIDE BAR Funtionality started*******
|
21 |
+
|
22 |
+
# Sidebar to capture the API keys
|
23 |
+
st.session_state['API_Key'] = st.sidebar.text_input("What's your OPENAI API key?",type="password")
|
24 |
+
# File uploader widget
|
25 |
+
uploaded_file = st.sidebar.file_uploader("Choose a PDF file", type=["pdf"])
|
26 |
+
|
27 |
+
load_button = st.sidebar.button("UPLOAD", key="load_button")
|
28 |
+
|
29 |
+
#If the bove button is clicked, pushing the data to Pinecone...
|
30 |
+
if load_button:
|
31 |
+
#Proceed only if API keys are provided
|
32 |
+
if st.session_state['API_Key'] != '' and uploaded_file is not None:
|
33 |
+
file = save_pdf(uploaded_file)
|
34 |
+
file = "uploaded.pdf"
|
35 |
+
st.session_state.summary = load_db_sum(file, st.session_state['API_Key'])
|
36 |
+
st.session_state.chat = load_db(file, st.session_state['API_Key'])
|
37 |
+
st.session_state.history = {}
|
38 |
+
|
39 |
+
elif st.session_state['API_Key'] == '':
|
40 |
+
st.sidebar.error("Please enter your OpenAI API key.")
|
41 |
+
elif uploaded_file is None:
|
42 |
+
st.sidebar.error("Please attach a PDF file.")
|
43 |
+
|
44 |
+
#********SIDE BAR Funtionality ended*****
|
45 |
+
|
46 |
+
|
47 |
+
if st.session_state['API_Key'] != '' and uploaded_file is not None:
|
48 |
+
file = "uploaded.pdf"
|
49 |
+
st.markdown('<br>', unsafe_allow_html=True)
|
50 |
+
st.markdown("#### **Summary**")
|
51 |
+
st.markdown('<hr style="margin: -10px 0; border-top: 1px solid black;">', unsafe_allow_html=True)
|
52 |
+
st.write(st.session_state.summary)
|
53 |
+
|
54 |
+
# create a variable for the chat
|
55 |
+
conversation = {}
|
56 |
+
|
57 |
+
#Captures User Inputs
|
58 |
+
user_input = st.text_input('Ask about the PDF',key="prompt") # The box for the text prompt
|
59 |
+
# document_count = st.slider('No.Of links to return 🔗 - (0 LOW || 5 HIGH)', 0, 5, 2,step=1)
|
60 |
+
|
61 |
+
submit = st.button("SUBMIT")
|
62 |
+
|
63 |
+
if submit:
|
64 |
+
#Proceed only if API keys are provided
|
65 |
+
if st.session_state.summary == '':
|
66 |
+
st.error("Please upload the PDF file.")
|
67 |
+
|
68 |
+
# user_input = request.form['user_input']
|
69 |
+
else:
|
70 |
+
result = st.session_state.chat({"question": user_input})
|
71 |
+
answer_text = str(result['answer'])
|
72 |
+
question_text = str(result['question'])
|
73 |
+
user = "User"
|
74 |
+
chatbot = "Chat Bot"
|
75 |
+
conversation.update({user: question_text, chatbot: answer_text})
|
76 |
+
|
77 |
+
user_hist = f"[{st.session_state.counter}] {user}"
|
78 |
+
chat_hist = f"[{st.session_state.counter}] {chatbot}"
|
79 |
+
st.session_state.history.update({user_hist : question_text})
|
80 |
+
st.session_state.history.update({chat_hist : answer_text})
|
81 |
+
st.session_state.counter += 1
|
82 |
+
|
83 |
+
|
84 |
+
st.markdown('<br>', unsafe_allow_html=True)
|
85 |
+
st.markdown("#### **Conversation**")
|
86 |
+
st.markdown('<hr style="margin: -10px 0; border-top: 1px solid black;">', unsafe_allow_html=True)
|
87 |
+
|
88 |
+
table_data = list(conversation.items())
|
89 |
+
|
90 |
+
# Display the table with keys bolded using HTML
|
91 |
+
html_table = """
|
92 |
+
<style>
|
93 |
+
table, tr {border:hidden;}
|
94 |
+
table, td {border:hidden;}
|
95 |
+
</style>
|
96 |
+
<table><tr><th><strong></strong></th><th></th></tr>
|
97 |
+
"""
|
98 |
+
for key, value in table_data:
|
99 |
+
html_table += f"<tr><td style='width: 90px;'><strong>{key}:</strong></td><td>{value}</td></tr>"
|
100 |
+
html_table += "</table>"
|
101 |
+
st.markdown(html_table, unsafe_allow_html=True)
|
102 |
+
|
103 |
+
|
104 |
+
st.markdown('<br>', unsafe_allow_html=True)
|
105 |
+
st.markdown("#### **Chat History**")
|
106 |
+
st.markdown('<hr style="margin: -10px 0; border-top: 1px solid black;">', unsafe_allow_html=True)
|
107 |
+
|
108 |
+
table_data2 = list(st.session_state.history.items())
|
109 |
+
|
110 |
+
# Display the table with keys bolded using HTML
|
111 |
+
html_table = """
|
112 |
+
<style>
|
113 |
+
table, tr {border:hidden;}
|
114 |
+
table, td {border:hidden;}
|
115 |
+
</style>
|
116 |
+
<table><tr><th><strong></strong></th><th></th></tr>
|
117 |
+
"""
|
118 |
+
for key, value in table_data2:
|
119 |
+
key = key[4:]
|
120 |
+
html_table += f"<tr><td style='width: 90px;'><strong>{key}:</strong></td><td>{value}</td></tr>"
|
121 |
+
html_table += "</table>"
|
122 |
+
st.markdown(html_table, unsafe_allow_html=True)
|
123 |
+
|
124 |
+
elif st.session_state['API_Key'] == '':
|
125 |
+
st.error("Please enter your OpenAI API key.")
|
126 |
+
elif uploaded_file is None:
|
127 |
+
st.session_state.summary = ''
|
128 |
+
st.error("Please upload the PDF file.")
|
129 |
+
|
130 |
+
|
131 |
+
|
pdf_reader.py
ADDED
@@ -0,0 +1,121 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Imports
|
2 |
+
from langchain.embeddings.openai import OpenAIEmbeddings
|
3 |
+
from langchain.text_splitter import CharacterTextSplitter, RecursiveCharacterTextSplitter
|
4 |
+
from langchain.vectorstores import DocArrayInMemorySearch
|
5 |
+
from langchain.chains.summarize import load_summarize_chain
|
6 |
+
from langchain.chains import RetrievalQA, ConversationalRetrievalChain
|
7 |
+
from langchain.memory import ConversationBufferMemory
|
8 |
+
from langchain.chat_models import ChatOpenAI
|
9 |
+
from langchain.document_loaders import TextLoader
|
10 |
+
from langchain.document_loaders import PyPDFLoader
|
11 |
+
from langchain.prompts import PromptTemplate
|
12 |
+
from langchain.llms import OpenAI
|
13 |
+
import tiktoken
|
14 |
+
import os
|
15 |
+
import sys
|
16 |
+
sys.path.append('../..')
|
17 |
+
|
18 |
+
import datetime
|
19 |
+
current_date = datetime.datetime.now().date()
|
20 |
+
if current_date < datetime.date(2023, 9, 2):
|
21 |
+
llm_name = "gpt-3.5-turbo-0301"
|
22 |
+
else:
|
23 |
+
llm_name = "gpt-3.5-turbo"
|
24 |
+
|
25 |
+
def load_db(file, api_key):
|
26 |
+
os.environ['OPENAI_API_KEY'] = api_key
|
27 |
+
# load documents
|
28 |
+
loader = PyPDFLoader(file)
|
29 |
+
# loader = file
|
30 |
+
documents = loader.load()
|
31 |
+
# documents = loader.read()
|
32 |
+
# split documents
|
33 |
+
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=150)
|
34 |
+
docs = text_splitter.split_documents(documents)
|
35 |
+
# define embedding
|
36 |
+
embeddings = OpenAIEmbeddings()
|
37 |
+
# create vector database from data
|
38 |
+
db = DocArrayInMemorySearch.from_documents(docs, embeddings)
|
39 |
+
|
40 |
+
# add in the prompt
|
41 |
+
prompt_template_doc = """
|
42 |
+
|
43 |
+
Use the following pieces of context to answer the question at the end. {context}
|
44 |
+
You can also look into chat history. {chat_history}
|
45 |
+
If you still can't find the answer, please respond: "Please ask a question related to the document."
|
46 |
+
|
47 |
+
Question: {question}
|
48 |
+
Answer:
|
49 |
+
"""
|
50 |
+
prompt_doc = PromptTemplate(
|
51 |
+
template=prompt_template_doc,
|
52 |
+
input_variables=["context", "question", "chat_history"],
|
53 |
+
)
|
54 |
+
# define retriever
|
55 |
+
retriever = db.as_retriever(search_type="similarity", search_kwargs={"k": 4})
|
56 |
+
# keeps a buffer of history and process it
|
57 |
+
memory = ConversationBufferMemory(
|
58 |
+
memory_key="chat_history",
|
59 |
+
output_key="answer",
|
60 |
+
return_messages=True
|
61 |
+
)
|
62 |
+
# create a chatbot chain
|
63 |
+
qa = ConversationalRetrievalChain.from_llm(
|
64 |
+
llm=ChatOpenAI(model_name=llm_name, temperature=0),
|
65 |
+
chain_type="stuff",
|
66 |
+
retriever=retriever,
|
67 |
+
combine_docs_chain_kwargs={"prompt": prompt_doc},
|
68 |
+
memory=memory
|
69 |
+
)
|
70 |
+
return qa
|
71 |
+
|
72 |
+
def load_db_sum(file, api_key):
|
73 |
+
os.environ['OPENAI_API_KEY'] = api_key
|
74 |
+
# load documents
|
75 |
+
loader = PyPDFLoader(file)
|
76 |
+
# loader = file
|
77 |
+
documents = loader.load()
|
78 |
+
# documents = loader.read()
|
79 |
+
# split documents
|
80 |
+
text_splitter = RecursiveCharacterTextSplitter(chunk_size=4000, chunk_overlap=150)
|
81 |
+
docs = text_splitter.split_documents(documents)
|
82 |
+
# create string of documents
|
83 |
+
str_docs = str(documents)
|
84 |
+
|
85 |
+
# define number of tokens from text
|
86 |
+
def num_tokens_from_string(string: str, encoding_name: str) -> int:
|
87 |
+
encoding = tiktoken.encoding_for_model(encoding_name)
|
88 |
+
num_tokens = len(encoding.encode(string))
|
89 |
+
return num_tokens
|
90 |
+
|
91 |
+
# get tokens
|
92 |
+
num_tokens = num_tokens_from_string(str_docs, llm_name)
|
93 |
+
model_max_tokens = 4097
|
94 |
+
# define embedding
|
95 |
+
embeddings = OpenAIEmbeddings()
|
96 |
+
# create vector database from data
|
97 |
+
db = DocArrayInMemorySearch.from_documents(docs, embeddings)
|
98 |
+
# define retriever
|
99 |
+
retriever = db.as_retriever(search_type="similarity", search_kwargs={"k": 4})
|
100 |
+
#Keeps a buffer of history and process it
|
101 |
+
memory = ConversationBufferMemory(
|
102 |
+
memory_key="chat_history",
|
103 |
+
output_key="answer",
|
104 |
+
return_messages=True
|
105 |
+
)
|
106 |
+
|
107 |
+
# create a chatbot chain based on tokens
|
108 |
+
if num_tokens < model_max_tokens:
|
109 |
+
chain = load_summarize_chain(llm=OpenAI(temperature=0, model="text-davinci-003", openai_api_key=api_key), chain_type="stuff")
|
110 |
+
qa = chain.run(documents)
|
111 |
+
else:
|
112 |
+
chain = load_summarize_chain(llm=OpenAI(temperature=0, model="text-davinci-003", openai_api_key=api_key), chain_type="map_reduce")
|
113 |
+
qa = chain.run(documents)
|
114 |
+
|
115 |
+
return qa
|
116 |
+
|
117 |
+
def save_pdf(pdf_file):
|
118 |
+
with open("uploaded.pdf", "wb") as file:
|
119 |
+
file.write(pdf_file.getvalue())
|
120 |
+
file = "uploaded.pdf"
|
121 |
+
return file
|
requirements.txt
ADDED
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
langchain
|
2 |
+
tiktoken
|
3 |
+
os
|
4 |
+
sys
|
5 |
+
streamlit
|