Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import os
|
3 |
+
from langchain.chains import RetrievalQA, ConversationalRetrievalChain
|
4 |
+
from langchain.chat_models import ChatOpenAI
|
5 |
+
from langchain.document_loaders import TextLoader
|
6 |
+
from langchain.text_splitter import CharacterTextSplitter
|
7 |
+
from langchain.embeddings import OpenAIEmbeddings
|
8 |
+
from langchain.vectorstores import FAISS
|
9 |
+
from langchain.memory import ConversationBufferMemory
|
10 |
+
import re
|
11 |
+
|
12 |
+
# Initialize the Streamlit app
|
13 |
+
st.title('Document-Based Q&A System')
|
14 |
+
|
15 |
+
# API Key input securely
|
16 |
+
api_key = st.text_input("Enter your OpenAI API key:", type="password")
|
17 |
+
if api_key:
|
18 |
+
os.environ["OPENAI_API_KEY"] = api_key
|
19 |
+
st.success("API Key has been set!")
|
20 |
+
|
21 |
+
# File uploader
|
22 |
+
uploaded_file = st.file_uploader("Upload your document", type=['txt'])
|
23 |
+
if uploaded_file is not None:
|
24 |
+
# Read and process the document
|
25 |
+
text_data = uploaded_file.getvalue().decode("utf-8")
|
26 |
+
text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
|
27 |
+
data = text_splitter.split_documents(text_data)
|
28 |
+
|
29 |
+
# Create vector store
|
30 |
+
embeddings = OpenAIEmbeddings()
|
31 |
+
vectorstore = FAISS.from_documents(data, embedding=embeddings)
|
32 |
+
|
33 |
+
# Create conversation chain
|
34 |
+
llm = ChatOpenAI(temperature=0.3, model_name="gpt-4-turbo")
|
35 |
+
memory = ConversationBufferMemory(
|
36 |
+
memory_key='chat_history', return_messages=True, output_key='answer')
|
37 |
+
conversation_chain = ConversationalRetrievalChain.from_llm(
|
38 |
+
llm=llm,
|
39 |
+
chain_type="stuff",
|
40 |
+
retriever=vectorstore.as_retriever(),
|
41 |
+
memory=memory,
|
42 |
+
return_source_documents=True
|
43 |
+
)
|
44 |
+
|
45 |
+
# Question input
|
46 |
+
query = st.text_input("Ask a question about the document:")
|
47 |
+
if query:
|
48 |
+
result = conversation_chain({"question": query})
|
49 |
+
answer = result["answer"]
|
50 |
+
st.write("Answer:", answer)
|
51 |
+
|
52 |
+
# Optionally display source text snippets
|
53 |
+
if st.checkbox("Show source text snippets"):
|
54 |
+
st.write("Source documents:")
|
55 |
+
for i in result["source_documents"]:
|
56 |
+
res = re.search(r'^[^\n]*', i.page_content)
|
57 |
+
st.write(i.page_content[res.span()[0]:res.span()[1]])
|