Spaces:
Runtime error
Runtime error
OpenRAG128
commited on
Commit
β’
3abdd86
1
Parent(s):
f0d87c1
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,130 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import PyPDF2
|
3 |
+
from langchain_community.embeddings import OllamaEmbeddings
|
4 |
+
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
5 |
+
from langchain_community.vectorstores.chroma import Chroma
|
6 |
+
from langchain.chains import ConversationalRetrievalChain
|
7 |
+
from langchain.memory import ChatMessageHistory, ConversationBufferMemory
|
8 |
+
from langchain_groq import ChatGroq
|
9 |
+
|
10 |
+
|
11 |
+
|
12 |
+
# Function to initialize conversation chain with GROQ language model
|
13 |
+
groq_api_key = "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"
|
14 |
+
|
15 |
+
# Initializing GROQ chat with provided API key, model name, and settings
|
16 |
+
llm_groq = ChatGroq(
|
17 |
+
groq_api_key=groq_api_key, model_name="mixtral-8x7b-32768", temperature=0.2
|
18 |
+
)
|
19 |
+
|
20 |
+
# Streamlit app
|
21 |
+
st.set_page_config(page_title="DocDynamo", layout="wide")
|
22 |
+
|
23 |
+
st.title("DocDynamoπ")
|
24 |
+
uploaded_file = st.file_uploader("Please upload a PDF file to begin!", type="pdf")
|
25 |
+
|
26 |
+
st.sidebar.title("DocDynamo By OpenRAG")
|
27 |
+
st.sidebar.markdown(
|
28 |
+
"""
|
29 |
+
π **Introducing DocDynamo by OpenRAG: Your PDF Companion!** π
|
30 |
+
|
31 |
+
Welcome, esteemed users, to the groundbreaking release of DocDynamo on May 21, 2024. At OpenRAG, we are committed to pioneering solutions for modern challenges, and DocDynamo is our latest triumph.
|
32 |
+
|
33 |
+
"""
|
34 |
+
|
35 |
+
)
|
36 |
+
|
37 |
+
st.sidebar.markdown(
|
38 |
+
"""
|
39 |
+
π‘ **How DocDynamo Works**
|
40 |
+
|
41 |
+
Simply upload your PDF, and let DocDynamo work its magic. Once processed, you can ask DocDynamo any question pertaining to the content of your PDF. It's like having a personal assistant at your fingertips, ready to provide instant answers.
|
42 |
+
"""
|
43 |
+
|
44 |
+
)
|
45 |
+
|
46 |
+
st.sidebar.markdown(
|
47 |
+
"""
|
48 |
+
π§ **Get in Touch**
|
49 |
+
|
50 |
+
For inquiries or collaboration proposals, please don't hesitate to reach out to us:
|
51 |
+
π© Email: openrag189@gmail.com
|
52 |
+
π LinkedIn: [OpenRAG](https://www.linkedin.com/company/102036854/admin/dashboard/)
|
53 |
+
πΈ Instagram: [OpenRAG](https://www.instagram.com/open.rag?igsh=MnFwMHd5cjU1OGFj)
|
54 |
+
|
55 |
+
Experience the future of PDF interaction with DocDynamo. Welcome to a new era of efficiency and productivity. OpenRAG: Empowering You Through Innovation. π
|
56 |
+
"""
|
57 |
+
|
58 |
+
)
|
59 |
+
if uploaded_file:
|
60 |
+
# Inform the user that processing has started
|
61 |
+
with st.spinner(f"Processing `{uploaded_file.name}`..."):
|
62 |
+
# Read the PDF file
|
63 |
+
pdf = PyPDF2.PdfReader(uploaded_file)
|
64 |
+
pdf_text = ""
|
65 |
+
for page in pdf.pages:
|
66 |
+
pdf_text += page.extract_text()
|
67 |
+
|
68 |
+
# Split the text into chunks
|
69 |
+
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1200, chunk_overlap=50)
|
70 |
+
texts = text_splitter.split_text(pdf_text)
|
71 |
+
|
72 |
+
# Create metadata for each chunk
|
73 |
+
metadatas = [{"source": f"{i}-pl"} for i in range(len(texts))]
|
74 |
+
|
75 |
+
# Create a Chroma vector store
|
76 |
+
embeddings = OllamaEmbeddings(model="nomic-embed-text")
|
77 |
+
docsearch = Chroma.from_texts(texts, embeddings, metadatas=metadatas)
|
78 |
+
|
79 |
+
# Initialize message history for conversation
|
80 |
+
message_history = ChatMessageHistory()
|
81 |
+
|
82 |
+
# Memory for conversational context
|
83 |
+
memory = ConversationBufferMemory(
|
84 |
+
memory_key="chat_history",
|
85 |
+
output_key="answer",
|
86 |
+
chat_memory=message_history,
|
87 |
+
return_messages=True,
|
88 |
+
)
|
89 |
+
|
90 |
+
# Create a chain that uses the Chroma vector store
|
91 |
+
chain = ConversationalRetrievalChain.from_llm(
|
92 |
+
llm=llm_groq,
|
93 |
+
chain_type="stuff",
|
94 |
+
retriever=docsearch.as_retriever(),
|
95 |
+
memory=memory,
|
96 |
+
return_source_documents=True,
|
97 |
+
)
|
98 |
+
|
99 |
+
st.success(f"Processing `{uploaded_file.name}` done. You can now ask questions!")
|
100 |
+
|
101 |
+
user_input = st.text_input("Ask a question about the PDF:")
|
102 |
+
|
103 |
+
if user_input:
|
104 |
+
# Call the chain with user's message content
|
105 |
+
res = chain.invoke(user_input)
|
106 |
+
answer = res["answer"]
|
107 |
+
source_documents = res["source_documents"]
|
108 |
+
|
109 |
+
text_elements = [] # Initialize list to store text elements
|
110 |
+
|
111 |
+
# Process source documents if available
|
112 |
+
if source_documents:
|
113 |
+
for source_idx, source_doc in enumerate(source_documents):
|
114 |
+
source_name = f"source_{source_idx}"
|
115 |
+
# Create the text element referenced in the message
|
116 |
+
text_elements.append(source_doc.page_content)
|
117 |
+
source_names = [f"source_{idx}" for idx in range(len(source_documents))]
|
118 |
+
|
119 |
+
# Add source references to the answer
|
120 |
+
if source_names:
|
121 |
+
answer += f"\nSources: {', '.join(source_names)}"
|
122 |
+
else:
|
123 |
+
answer += "\nNo sources found"
|
124 |
+
|
125 |
+
# Display the results
|
126 |
+
st.markdown(f"**Answer:** {answer}")
|
127 |
+
|
128 |
+
for idx, element in enumerate(text_elements):
|
129 |
+
with st.expander(f"Source {idx}"):
|
130 |
+
st.write(element)
|