Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,144 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
from PyPDF2 import PdfReader
|
3 |
+
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
4 |
+
from langchain_google_genai import GoogleGenerativeAIEmbeddings
|
5 |
+
import streamlit as st
|
6 |
+
import google.generativeai as genai
|
7 |
+
from langchain.vectorstores import FAISS
|
8 |
+
from langchain_google_genai import ChatGoogleGenerativeAI
|
9 |
+
from langchain.chains.question_answering import load_qa_chain
|
10 |
+
from langchain.prompts import PromptTemplate
|
11 |
+
from dotenv import load_dotenv
|
12 |
+
|
13 |
+
load_dotenv()
|
14 |
+
os.getenv("GOOGLE_API_KEY")
|
15 |
+
genai.configure(api_key=os.getenv("GOOGLE_API_KEY"))
|
16 |
+
|
17 |
+
# read all pdf files and return text
|
18 |
+
|
19 |
+
|
20 |
+
def get_pdf_text(pdf_docs):
|
21 |
+
text = ""
|
22 |
+
for pdf in pdf_docs:
|
23 |
+
pdf_reader = PdfReader(pdf)
|
24 |
+
for page in pdf_reader.pages:
|
25 |
+
text += page.extract_text()
|
26 |
+
return text
|
27 |
+
|
28 |
+
# split text into chunks
|
29 |
+
|
30 |
+
|
31 |
+
def get_text_chunks(text):
|
32 |
+
splitter = RecursiveCharacterTextSplitter(
|
33 |
+
chunk_size=10000, chunk_overlap=1000)
|
34 |
+
chunks = splitter.split_text(text)
|
35 |
+
return chunks # list of strings
|
36 |
+
|
37 |
+
# get embeddings for each chunk
|
38 |
+
|
39 |
+
|
40 |
+
def get_vector_store(chunks):
|
41 |
+
embeddings = GoogleGenerativeAIEmbeddings(
|
42 |
+
model="models/embedding-001") # type: ignore
|
43 |
+
vector_store = FAISS.from_texts(chunks, embedding=embeddings)
|
44 |
+
vector_store.save_local("faiss_index")
|
45 |
+
|
46 |
+
|
47 |
+
def get_conversational_chain():
|
48 |
+
prompt_template = """
|
49 |
+
Answer the question as detailed as possible from the provided context, make sure to provide all the details, if the answer is not in
|
50 |
+
provided context just say, "answer is not available in the context", don't provide the wrong answer\n\n
|
51 |
+
Context:\n {context}?\n
|
52 |
+
Question: \n{question}\n
|
53 |
+
|
54 |
+
Answer:
|
55 |
+
"""
|
56 |
+
|
57 |
+
model = ChatGoogleGenerativeAI(model="gemini-pro",
|
58 |
+
client=genai,
|
59 |
+
temperature=0.3,
|
60 |
+
)
|
61 |
+
prompt = PromptTemplate(template=prompt_template,
|
62 |
+
input_variables=["context", "question"])
|
63 |
+
chain = load_qa_chain(llm=model, chain_type="stuff", prompt=prompt)
|
64 |
+
return chain
|
65 |
+
|
66 |
+
|
67 |
+
def clear_chat_history():
|
68 |
+
st.session_state.messages = [
|
69 |
+
{"role": "assistant", "content": "upload some pdfs and ask me a question"}]
|
70 |
+
|
71 |
+
|
72 |
+
def user_input(user_question):
|
73 |
+
embeddings = GoogleGenerativeAIEmbeddings(
|
74 |
+
model="models/embedding-001") # type: ignore
|
75 |
+
|
76 |
+
new_db = FAISS.load_local("faiss_index", embeddings, allow_dangerous_deserialization=True)
|
77 |
+
docs = new_db.similarity_search(user_question)
|
78 |
+
|
79 |
+
chain = get_conversational_chain()
|
80 |
+
|
81 |
+
response = chain(
|
82 |
+
{"input_documents": docs, "question": user_question}, return_only_outputs=True, )
|
83 |
+
|
84 |
+
print(response)
|
85 |
+
return response
|
86 |
+
|
87 |
+
|
88 |
+
def main():
|
89 |
+
st.set_page_config(
|
90 |
+
page_title="Gemini PDF Chatbot",
|
91 |
+
page_icon="🤖"
|
92 |
+
)
|
93 |
+
|
94 |
+
# Sidebar for uploading PDF files
|
95 |
+
with st.sidebar:
|
96 |
+
st.title("Menu:")
|
97 |
+
pdf_docs = st.file_uploader(
|
98 |
+
"Upload your PDF Files and Click on the Submit & Process Button", accept_multiple_files=True)
|
99 |
+
if st.button("Submit & Process"):
|
100 |
+
with st.spinner("Processing..."):
|
101 |
+
raw_text = get_pdf_text(pdf_docs)
|
102 |
+
text_chunks = get_text_chunks(raw_text)
|
103 |
+
get_vector_store(text_chunks)
|
104 |
+
st.success("Done")
|
105 |
+
|
106 |
+
# Main content area for displaying chat messages
|
107 |
+
st.title("Chat with PDF files using Gemini🤖")
|
108 |
+
st.write("Welcome to the chat!")
|
109 |
+
st.sidebar.button('Clear Chat History', on_click=clear_chat_history)
|
110 |
+
|
111 |
+
# Chat input
|
112 |
+
# Placeholder for chat messages
|
113 |
+
|
114 |
+
if "messages" not in st.session_state.keys():
|
115 |
+
st.session_state.messages = [
|
116 |
+
{"role": "assistant", "content": "upload some pdfs and ask me a question"}]
|
117 |
+
|
118 |
+
for message in st.session_state.messages:
|
119 |
+
with st.chat_message(message["role"]):
|
120 |
+
st.write(message["content"])
|
121 |
+
|
122 |
+
if prompt := st.chat_input():
|
123 |
+
st.session_state.messages.append({"role": "user", "content": prompt})
|
124 |
+
with st.chat_message("user"):
|
125 |
+
st.write(prompt)
|
126 |
+
|
127 |
+
# Display chat messages and bot response
|
128 |
+
if st.session_state.messages[-1]["role"] != "assistant":
|
129 |
+
with st.chat_message("assistant"):
|
130 |
+
with st.spinner("Thinking..."):
|
131 |
+
response = user_input(prompt)
|
132 |
+
placeholder = st.empty()
|
133 |
+
full_response = ''
|
134 |
+
for item in response['output_text']:
|
135 |
+
full_response += item
|
136 |
+
placeholder.markdown(full_response)
|
137 |
+
placeholder.markdown(full_response)
|
138 |
+
if response is not None:
|
139 |
+
message = {"role": "assistant", "content": full_response}
|
140 |
+
st.session_state.messages.append(message)
|
141 |
+
|
142 |
+
|
143 |
+
if __name__ == "__main__":
|
144 |
+
main()
|