Upload 2 files
Browse files- chatpdf.py +98 -0
- requirements.txt +10 -0
chatpdf.py
ADDED
@@ -0,0 +1,98 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import pickle
|
3 |
+
import streamlit as st
|
4 |
+
from streamlit_extras.add_vertical_space import add_vertical_space
|
5 |
+
from PyPDF2 import PdfReader
|
6 |
+
from openai.embeddings_utils import get_embedding
|
7 |
+
import openai
|
8 |
+
from dotenv import load_dotenv
|
9 |
+
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
10 |
+
from langchain.embeddings.openai import OpenAIEmbeddings
|
11 |
+
from langchain.vectorstores import FAISS
|
12 |
+
from langchain.llms import OpenAI
|
13 |
+
from langchain.chains.question_answering import load_qa_chain
|
14 |
+
from langchain.callbacks import get_openai_callback
|
15 |
+
# Sidebar contents
|
16 |
+
with st.sidebar:
|
17 |
+
st.title('🤗LLM Chat App💬')
|
18 |
+
st.markdown('''
|
19 |
+
## About
|
20 |
+
OpenAI based LLM-powered chatbot built using:
|
21 |
+
- [OpenAI](https://platform.openai.com/docs/models) LLM model
|
22 |
+
- [Streamlit](https://streamlit.io/)
|
23 |
+
- [LangChain](https://python.langchain.com/)
|
24 |
+
''')
|
25 |
+
add_vertical_space(5)
|
26 |
+
st.write('Made with ❤️ by Harry')
|
27 |
+
|
28 |
+
|
29 |
+
# Load environment variables
|
30 |
+
# load_dotenv()
|
31 |
+
|
32 |
+
# # Retrieve OpenAI API key
|
33 |
+
# openai_api_key = os.getenv("OPENAI_API_KEY")
|
34 |
+
# if openai_api_key is None:
|
35 |
+
# raise ValueError("The OPENAI_API_KEY environment variable is not set")
|
36 |
+
|
37 |
+
# # Set the OpenAI API key for the OpenAI library
|
38 |
+
# openai.api_key = openai_api_key
|
39 |
+
|
40 |
+
def extract_text_from_pdf(pdf):
|
41 |
+
pdf_reader = PdfReader(pdf)
|
42 |
+
text = ""
|
43 |
+
for page in pdf_reader.pages:
|
44 |
+
text += page.extract_text()
|
45 |
+
return text
|
46 |
+
def get_embeddings(text_list):
|
47 |
+
return [get_embedding(text) for text in text_list]
|
48 |
+
def main():
|
49 |
+
st.header("Chat with PDF 💬")
|
50 |
+
# Upload a PDF file
|
51 |
+
pdf = st.file_uploader("Upload your PDF file", type='pdf')
|
52 |
+
|
53 |
+
if pdf is not None:
|
54 |
+
# Extract text from the PDF
|
55 |
+
|
56 |
+
text = extract_text_from_pdf(pdf)
|
57 |
+
# Split text into chunks
|
58 |
+
text_splitter = RecursiveCharacterTextSplitter(
|
59 |
+
chunk_size=1000,
|
60 |
+
chunk_overlap=200,
|
61 |
+
length_function=len
|
62 |
+
)
|
63 |
+
chunks = text_splitter.split_text(text=text)
|
64 |
+
# chunks data with langchain
|
65 |
+
#chunks = [text[i:i+chunk_size] for i in range(0, len(text), chunk_size - chunk_overlap)]
|
66 |
+
st.write("PDF content successfully extracted.")
|
67 |
+
#st.write("Below is chunks data")
|
68 |
+
#st.write(chunks)
|
69 |
+
|
70 |
+
# Create or load embeddings
|
71 |
+
store_name = pdf.name[:-4]
|
72 |
+
st.write(f'Processing: {store_name}')
|
73 |
+
|
74 |
+
if os.path.exists(f"{store_name}.pkl"):
|
75 |
+
with open(f"{store_name}.pkl", "rb") as f:
|
76 |
+
VectorStore = pickle.load(f)
|
77 |
+
st.write('Embeddings loaded from the disk')
|
78 |
+
else:
|
79 |
+
embeddings = OpenAIEmbeddings()
|
80 |
+
VectorStore = FAISS.from_texts(chunks, embedding=embeddings)
|
81 |
+
with open(f"{store_name}.pkl", "wb") as f:
|
82 |
+
pickle.dump(VectorStore, f)
|
83 |
+
st.write('Embeddings created and saved to disk')
|
84 |
+
|
85 |
+
# Accept user questions/query
|
86 |
+
query = st.text_input("Ask questions about your PDF file:")
|
87 |
+
|
88 |
+
if query:
|
89 |
+
docs = VectorStore.similarity_search(query=query, k=3)
|
90 |
+
|
91 |
+
llm = OpenAI(model_name="gpt-3.5-turbo")
|
92 |
+
chain = load_qa_chain(llm=llm, chain_type="stuff")
|
93 |
+
with get_openai_callback() as cb:
|
94 |
+
response = chain.run(input_documents=docs, question=query)
|
95 |
+
print(cb)
|
96 |
+
st.write(response)
|
97 |
+
if __name__ == '__main__':
|
98 |
+
main()
|
requirements.txt
ADDED
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
langchain
|
2 |
+
openai==0.28.1
|
3 |
+
PyPDF2
|
4 |
+
python-dotenv
|
5 |
+
streamlit
|
6 |
+
faiss-cpu
|
7 |
+
streamlit-extras
|
8 |
+
tiktoken
|
9 |
+
plotly
|
10 |
+
|