Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,183 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import os
|
3 |
+
from langchain_community.vectorstores import FAISS
|
4 |
+
from langchain_community.document_loaders import PyPDFLoader
|
5 |
+
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
6 |
+
from langchain_community.embeddings import HuggingFaceEmbeddings
|
7 |
+
from langchain_huggingface import HuggingFaceEndpoint # Updated import
|
8 |
+
from langchain.chains import ConversationalRetrievalChain
|
9 |
+
from langchain.memory import ConversationBufferMemory
|
10 |
+
import tempfile
|
11 |
+
|
12 |
+
api_token = os.getenv("HF_TOKEN")
|
13 |
+
list_llm = ["meta-llama/Meta-Llama-3-8B-Instruct", "mistralai/Mistral-7B-Instruct-v0.2"]
|
14 |
+
list_llm_simple = [os.path.basename(llm) for llm in list_llm]
|
15 |
+
|
16 |
+
def load_doc(uploaded_files):
|
17 |
+
try:
|
18 |
+
temp_files = []
|
19 |
+
for uploaded_file in uploaded_files:
|
20 |
+
temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".pdf")
|
21 |
+
temp_file.write(uploaded_file.read())
|
22 |
+
temp_file.close()
|
23 |
+
temp_files.append(temp_file.name)
|
24 |
+
|
25 |
+
loaders = [PyPDFLoader(x) for x in temp_files]
|
26 |
+
pages = []
|
27 |
+
for loader in loaders:
|
28 |
+
pages.extend(loader.load())
|
29 |
+
|
30 |
+
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1024, chunk_overlap=64)
|
31 |
+
doc_splits = text_splitter.split_documents(pages)
|
32 |
+
|
33 |
+
for temp_file in temp_files:
|
34 |
+
os.remove(temp_file) # Clean up temporary files
|
35 |
+
|
36 |
+
return doc_splits
|
37 |
+
except Exception as e:
|
38 |
+
st.error(f"Error loading document: {e}")
|
39 |
+
return []
|
40 |
+
|
41 |
+
def create_db(splits):
|
42 |
+
try:
|
43 |
+
embeddings = HuggingFaceEmbeddings()
|
44 |
+
vectordb = FAISS.from_documents(splits, embeddings)
|
45 |
+
return vectordb
|
46 |
+
except Exception as e:
|
47 |
+
st.error(f"Error creating vector database: {e}")
|
48 |
+
return None
|
49 |
+
|
50 |
+
def initialize_llmchain(llm_model, vector_db):
|
51 |
+
try:
|
52 |
+
llm = HuggingFaceEndpoint(
|
53 |
+
repo_id=llm_model,
|
54 |
+
huggingfacehub_api_token=api_token,
|
55 |
+
temperature=0.5,
|
56 |
+
max_new_tokens=4096,
|
57 |
+
top_k=3,
|
58 |
+
)
|
59 |
+
memory = ConversationBufferMemory(
|
60 |
+
memory_key="chat_history",
|
61 |
+
output_key='answer',
|
62 |
+
return_messages=True
|
63 |
+
)
|
64 |
+
|
65 |
+
retriever = vector_db.as_retriever()
|
66 |
+
qa_chain = ConversationalRetrievalChain.from_llm(
|
67 |
+
llm,
|
68 |
+
retriever=retriever,
|
69 |
+
chain_type="stuff",
|
70 |
+
memory=memory,
|
71 |
+
return_source_documents=True,
|
72 |
+
verbose=False,
|
73 |
+
)
|
74 |
+
return qa_chain
|
75 |
+
except Exception as e:
|
76 |
+
st.error(f"Error initializing LLM chain: {e}")
|
77 |
+
return None
|
78 |
+
|
79 |
+
def initialize_database(uploaded_files):
|
80 |
+
try:
|
81 |
+
doc_splits = load_doc(uploaded_files)
|
82 |
+
if not doc_splits:
|
83 |
+
return None, "Failed to load documents."
|
84 |
+
vector_db = create_db(doc_splits)
|
85 |
+
if vector_db is None:
|
86 |
+
return None, "Failed to create vector database."
|
87 |
+
return vector_db, "Database created!"
|
88 |
+
except Exception as e:
|
89 |
+
st.error(f"Error initializing database: {e}")
|
90 |
+
return None, "Failed to initialize database."
|
91 |
+
|
92 |
+
def initialize_LLM(llm_option, vector_db):
|
93 |
+
try:
|
94 |
+
llm_name = list_llm[llm_option]
|
95 |
+
qa_chain = initialize_llmchain(llm_name, vector_db)
|
96 |
+
if qa_chain is None:
|
97 |
+
return None, "Failed to initialize QA chain."
|
98 |
+
return qa_chain, "QA chain initialized. Chatbot is ready!"
|
99 |
+
except Exception as e:
|
100 |
+
st.error(f"Error initializing LLM: {e}")
|
101 |
+
return None, "Failed to initialize LLM."
|
102 |
+
|
103 |
+
def format_chat_history(chat_history):
|
104 |
+
formatted_chat_history = []
|
105 |
+
for user_message, bot_message in chat_history:
|
106 |
+
formatted_chat_history.append(f"User: {user_message}\nAssistant: {bot_message}\n")
|
107 |
+
return formatted_chat_history
|
108 |
+
|
109 |
+
def conversation(qa_chain, message, history):
|
110 |
+
try:
|
111 |
+
formatted_chat_history = format_chat_history(history)
|
112 |
+
response = qa_chain.invoke({"question": message, "chat_history": formatted_chat_history})
|
113 |
+
response_answer = response["answer"]
|
114 |
+
response_sources = response["source_documents"]
|
115 |
+
|
116 |
+
sources = []
|
117 |
+
for doc in response_sources:
|
118 |
+
sources.append({
|
119 |
+
"content": doc.page_content.strip(),
|
120 |
+
"page": doc.metadata["page"] + 1
|
121 |
+
})
|
122 |
+
|
123 |
+
new_history = history + [(message, response_answer)]
|
124 |
+
return qa_chain, new_history, response_answer, sources
|
125 |
+
except Exception as e:
|
126 |
+
st.error(f"Error in conversation: {e}")
|
127 |
+
return qa_chain, history, "", []
|
128 |
+
|
129 |
+
def main():
|
130 |
+
st.sidebar.title("PDF Chatbot")
|
131 |
+
|
132 |
+
st.sidebar.markdown("### Step 1 - Upload PDF documents and Initialize RAG pipeline")
|
133 |
+
uploaded_files = st.sidebar.file_uploader("Upload PDF documents", type="pdf", accept_multiple_files=True)
|
134 |
+
|
135 |
+
if uploaded_files:
|
136 |
+
if st.sidebar.button("Create vector database"):
|
137 |
+
with st.spinner("Creating vector database..."):
|
138 |
+
vector_db, db_message = initialize_database(uploaded_files)
|
139 |
+
st.sidebar.success(db_message)
|
140 |
+
st.session_state['vector_db'] = vector_db
|
141 |
+
|
142 |
+
if 'vector_db' not in st.session_state:
|
143 |
+
st.session_state['vector_db'] = None
|
144 |
+
|
145 |
+
if 'qa_chain' not in st.session_state:
|
146 |
+
st.session_state['qa_chain'] = None
|
147 |
+
|
148 |
+
if 'chat_history' not in st.session_state:
|
149 |
+
st.session_state['chat_history'] = []
|
150 |
+
|
151 |
+
st.sidebar.markdown("### Select Large Language Model (LLM)")
|
152 |
+
llm_option = st.sidebar.radio("Available LLMs", list_llm_simple)
|
153 |
+
|
154 |
+
if st.sidebar.button("Initialize Question Answering Chatbot"):
|
155 |
+
with st.spinner("Initializing QA chatbot..."):
|
156 |
+
qa_chain, llm_message = initialize_LLM(list_llm_simple.index(llm_option), st.session_state['vector_db'])
|
157 |
+
st.session_state['qa_chain'] = qa_chain
|
158 |
+
st.sidebar.success(llm_message)
|
159 |
+
|
160 |
+
st.title("Chat with your Document")
|
161 |
+
|
162 |
+
if st.session_state['qa_chain']:
|
163 |
+
message = st.text_input("Ask a question")
|
164 |
+
|
165 |
+
if st.button("Submit"):
|
166 |
+
with st.spinner("Generating response..."):
|
167 |
+
qa_chain, chat_history, response_answer, sources = conversation(st.session_state['qa_chain'], message, st.session_state['chat_history'])
|
168 |
+
st.session_state['qa_chain'] = qa_chain
|
169 |
+
st.session_state['chat_history'] = chat_history
|
170 |
+
|
171 |
+
st.markdown("### Chatbot Response")
|
172 |
+
|
173 |
+
# Display the chat history in a chat-like interface
|
174 |
+
for i, (user_msg, bot_msg) in enumerate(st.session_state['chat_history']):
|
175 |
+
st.markdown(f"**User:** {user_msg}")
|
176 |
+
st.markdown(f"**Assistant:** {bot_msg}")
|
177 |
+
|
178 |
+
with st.expander("Relevant context from the source document"):
|
179 |
+
for source in sources:
|
180 |
+
st.text_area(f"Source - Page {source['page']}", value=source["content"], height=100)
|
181 |
+
|
182 |
+
if __name__ == "__main__":
|
183 |
+
main()
|