Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,133 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
from dotenv import load_dotenv
|
3 |
+
from PyPDF2 import PdfReader
|
4 |
+
from langchain.text_splitter import CharacterTextSplitter
|
5 |
+
from langchain.embeddings import OpenAIEmbeddings, HuggingFaceInstructEmbeddings, HuggingFaceEmbeddings
|
6 |
+
from langchain.vectorstores import FAISS
|
7 |
+
from langchain.chat_models import ChatOpenAI
|
8 |
+
from langchain.memory import ConversationBufferMemory
|
9 |
+
from langchain.chains import ConversationalRetrievalChain
|
10 |
+
from langchain.chat_models.gigachat import GigaChat
|
11 |
+
from htmlTemplates import css, bot_template, user_template
|
12 |
+
from langchain.llms import HuggingFaceHub, LlamaCpp
|
13 |
+
from huggingface_hub import snapshot_download, hf_hub_download
|
14 |
+
|
15 |
+
|
16 |
+
|
17 |
+
repo_name = "IlyaGusev/saiga_mistral_7b_gguf"
|
18 |
+
model_name = "model-q4_K.gguf"
|
19 |
+
|
20 |
+
#snapshot_download(repo_id=repo_name, local_dir=".", allow_patterns=model_name)
|
21 |
+
|
22 |
+
|
23 |
+
def get_pdf_text(pdf_docs):
|
24 |
+
text = ""
|
25 |
+
for pdf in pdf_docs:
|
26 |
+
pdf_reader = PdfReader(pdf)
|
27 |
+
for page in pdf_reader.pages:
|
28 |
+
text += page.extract_text()
|
29 |
+
|
30 |
+
return text
|
31 |
+
|
32 |
+
|
33 |
+
def get_text_chunks(text):
|
34 |
+
text_splitter = CharacterTextSplitter(separator="\n",
|
35 |
+
chunk_size=1000, # 1000
|
36 |
+
chunk_overlap=200, # 200
|
37 |
+
length_function=len
|
38 |
+
)
|
39 |
+
chunks = text_splitter.split_text(text)
|
40 |
+
|
41 |
+
return chunks
|
42 |
+
|
43 |
+
|
44 |
+
#def get_vectorstore(text_chunks):
|
45 |
+
#embeddings = OpenAIEmbeddings()
|
46 |
+
#embeddings = HuggingFaceInstructEmbeddings(model_name="hkunlp/instructor-xl")
|
47 |
+
#embeddings = HuggingFaceEmbeddings(model_name="intfloat/multilingual-e5-large")
|
48 |
+
#embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/paraphrase-multilingual-mpnet-base-v2")
|
49 |
+
#vectorstore = FAISS.from_texts(texts=text_chunks, embedding=embeddings)
|
50 |
+
|
51 |
+
#return vectorstore
|
52 |
+
|
53 |
+
def get_vectorstore(text_chunks, embedding_model_name="intfloat/multilingual-e5-large"):
|
54 |
+
embeddings = HuggingFaceEmbeddings(model_name=embedding_model_name)
|
55 |
+
vectorstore = FAISS.from_texts(texts=text_chunks, embedding=embeddings)
|
56 |
+
return vectorstore
|
57 |
+
|
58 |
+
|
59 |
+
|
60 |
+
def get_conversation_chain(vectorstore, model_name):
|
61 |
+
|
62 |
+
|
63 |
+
llm = GigaChat(profanity=False,
|
64 |
+
verify_ssl_certs=False
|
65 |
+
)
|
66 |
+
|
67 |
+
memory = ConversationBufferMemory(memory_key='chat_history',
|
68 |
+
input_key='question',
|
69 |
+
output_key='answer',
|
70 |
+
return_messages=True)
|
71 |
+
|
72 |
+
conversation_chain = ConversationalRetrievalChain.from_llm(llm=llm,
|
73 |
+
retriever=vectorstore.as_retriever(),
|
74 |
+
memory=memory,
|
75 |
+
return_source_documents=True
|
76 |
+
)
|
77 |
+
|
78 |
+
return conversation_chain
|
79 |
+
|
80 |
+
|
81 |
+
def handle_userinput(user_question):
|
82 |
+
|
83 |
+
response = st.session_state.conversation({'question': user_question})
|
84 |
+
|
85 |
+
st.session_state.chat_history = response['chat_history']
|
86 |
+
|
87 |
+
st.session_state.retrieved_text = response['source_documents']
|
88 |
+
|
89 |
+
for i, (message, text) in enumerate(zip(st.session_state.chat_history, st.session_state.retrieved_text)):
|
90 |
+
if i % 3 == 0:
|
91 |
+
st.write(user_template.replace(
|
92 |
+
"{{MSG}}", message.content), unsafe_allow_html=True)
|
93 |
+
else:
|
94 |
+
st.write(bot_template.replace(
|
95 |
+
"{{MSG}}", message.content), unsafe_allow_html=True)
|
96 |
+
print(text)
|
97 |
+
st.write(bot_template.replace(
|
98 |
+
"{{MSG}}", str(text.page_content)), unsafe_allow_html=True)
|
99 |
+
|
100 |
+
|
101 |
+
st.set_page_config(page_title="Chat with multiple PDFs",
|
102 |
+
page_icon=":books:")
|
103 |
+
st.write(css, unsafe_allow_html=True)
|
104 |
+
|
105 |
+
if "conversation" not in st.session_state:
|
106 |
+
st.session_state.conversation = None
|
107 |
+
if "chat_history" not in st.session_state:
|
108 |
+
st.session_state.chat_history = None
|
109 |
+
|
110 |
+
st.header("Chat with multiple PDFs :books:")
|
111 |
+
user_question = st.text_input("Ask a question about your documents: ")
|
112 |
+
|
113 |
+
if user_question:
|
114 |
+
handle_userinput(user_question)
|
115 |
+
|
116 |
+
with st.sidebar:
|
117 |
+
st.subheader("Your documents")
|
118 |
+
embedding_model_name = st.selectbox("Select embedding model", ["intfloat/multilingual-e5-large", "sentence-transformers/paraphrase-multilingual-mpnet-base-v2"])
|
119 |
+
pdf_docs = st.file_uploader(
|
120 |
+
"Upload your PDFs here and click on 'Process'", accept_multiple_files=True)
|
121 |
+
if st.button("Process"):
|
122 |
+
with st.spinner("Processing"):
|
123 |
+
# get pdf text
|
124 |
+
raw_text = get_pdf_text(pdf_docs)
|
125 |
+
|
126 |
+
# get the text chunks
|
127 |
+
text_chunks = get_text_chunks(raw_text)
|
128 |
+
|
129 |
+
# create vector store
|
130 |
+
vectorstore = get_vectorstore(text_chunks, embedding_model_name)
|
131 |
+
|
132 |
+
# create conversation chain
|
133 |
+
st.session_state.conversation = get_conversation_chain(vectorstore, model_name)
|