Spaces:
Sleeping
Sleeping
Create app.py
Browse filesAdd code for mistral implementation.
app.py
ADDED
@@ -0,0 +1,197 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
"""
|
2 |
+
Question Answering with Retrieval QA and LangChain Language Models featuring Chroma Vector Stores
|
3 |
+
This script uses the LangChain Language Model API to answer questions using Retrieval QA and Chroma Vector Stores.
|
4 |
+
"""
|
5 |
+
|
6 |
+
import os
|
7 |
+
import streamlit as st
|
8 |
+
from dotenv import load_dotenv
|
9 |
+
from PyPDF2 import PdfReader
|
10 |
+
from langchain.text_splitter import CharacterTextSplitter
|
11 |
+
from langchain.embeddings import HuggingFaceBgeEmbeddings
|
12 |
+
from langchain.vectorstores import FAISS
|
13 |
+
from langchain.chat_models import ChatOpenAI
|
14 |
+
from langchain.memory import ConversationBufferMemory
|
15 |
+
from langchain.chains import ConversationalRetrievalChain
|
16 |
+
from htmlTemplates import css, bot_template, user_template
|
17 |
+
from langchain.llms import HuggingFaceHub
|
18 |
+
|
19 |
+
|
20 |
+
def get_pdf_text(pdf_docs):
|
21 |
+
"""
|
22 |
+
Extract text from a list of PDF documents.
|
23 |
+
|
24 |
+
Parameters
|
25 |
+
----------
|
26 |
+
pdf_docs : list
|
27 |
+
List of PDF documents to extract text from.
|
28 |
+
|
29 |
+
Returns
|
30 |
+
-------
|
31 |
+
str
|
32 |
+
Extracted text from all the PDF documents.
|
33 |
+
|
34 |
+
"""
|
35 |
+
text = ""
|
36 |
+
for pdf in pdf_docs:
|
37 |
+
pdf_reader = PdfReader(pdf)
|
38 |
+
for page in pdf_reader.pages:
|
39 |
+
text += page.extract_text()
|
40 |
+
return text
|
41 |
+
|
42 |
+
|
43 |
+
def get_text_chunks(text):
|
44 |
+
"""
|
45 |
+
Split the input text into chunks.
|
46 |
+
|
47 |
+
Parameters
|
48 |
+
----------
|
49 |
+
text : str
|
50 |
+
The input text to be split.
|
51 |
+
|
52 |
+
Returns
|
53 |
+
-------
|
54 |
+
list
|
55 |
+
List of text chunks.
|
56 |
+
|
57 |
+
"""
|
58 |
+
text_splitter = CharacterTextSplitter(
|
59 |
+
separator="\n", chunk_size=1500, chunk_overlap=300, length_function=len
|
60 |
+
)
|
61 |
+
chunks = text_splitter.split_text(text)
|
62 |
+
return chunks
|
63 |
+
|
64 |
+
|
65 |
+
def get_vectorstore(text_chunks):
|
66 |
+
"""
|
67 |
+
Generate a vector store from a list of text chunks using HuggingFace BgeEmbeddings.
|
68 |
+
|
69 |
+
Parameters
|
70 |
+
----------
|
71 |
+
text_chunks : list
|
72 |
+
List of text chunks to be embedded.
|
73 |
+
|
74 |
+
Returns
|
75 |
+
-------
|
76 |
+
FAISS
|
77 |
+
A FAISS vector store containing the embeddings of the text chunks.
|
78 |
+
|
79 |
+
"""
|
80 |
+
# embeddings = OpenAIEmbeddings()
|
81 |
+
model = "BAAI/bge-base-en-v1.5"
|
82 |
+
encode_kwargs = {
|
83 |
+
"normalize_embeddings": True
|
84 |
+
} # set True to compute cosine similarity
|
85 |
+
embeddings = HuggingFaceBgeEmbeddings(
|
86 |
+
model_name=model, encode_kwargs=encode_kwargs, model_kwargs={"device": "cpu"}
|
87 |
+
)
|
88 |
+
vectorstore = FAISS.from_texts(texts=text_chunks, embedding=embeddings)
|
89 |
+
return vectorstore
|
90 |
+
|
91 |
+
|
92 |
+
def get_conversation_chain(vectorstore):
|
93 |
+
"""
|
94 |
+
Create a conversational retrieval chain using a vector store and a language model.
|
95 |
+
|
96 |
+
Parameters
|
97 |
+
----------
|
98 |
+
vectorstore : FAISS
|
99 |
+
A FAISS vector store containing the embeddings of the text chunks.
|
100 |
+
|
101 |
+
Returns
|
102 |
+
-------
|
103 |
+
ConversationalRetrievalChain
|
104 |
+
A conversational retrieval chain for generating responses.
|
105 |
+
|
106 |
+
"""
|
107 |
+
llm = HuggingFaceHub(
|
108 |
+
repo_id="mistralai/Mistral-7B-Instruct-v0.1",
|
109 |
+
model_kwargs={"temperature": 0.5, "max_length": 512},
|
110 |
+
)
|
111 |
+
# llm = ChatOpenAI(temperature=0, model="gpt-3.5-turbo-0613")
|
112 |
+
|
113 |
+
memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)
|
114 |
+
conversation_chain = ConversationalRetrievalChain.from_llm(
|
115 |
+
llm=llm, retriever=vectorstore.as_retriever(), memory=memory
|
116 |
+
)
|
117 |
+
return conversation_chain
|
118 |
+
|
119 |
+
|
120 |
+
def handle_userinput(user_question):
|
121 |
+
"""
|
122 |
+
Handle user input and generate a response using the conversational retrieval chain.
|
123 |
+
|
124 |
+
Parameters
|
125 |
+
----------
|
126 |
+
user_question : str
|
127 |
+
The user's question.
|
128 |
+
|
129 |
+
"""
|
130 |
+
response = st.session_state.conversation({"question": user_question})
|
131 |
+
st.session_state.chat_history = response["chat_history"]
|
132 |
+
|
133 |
+
for i, message in enumerate(st.session_state.chat_history):
|
134 |
+
if i % 2 == 0:
|
135 |
+
st.write(
|
136 |
+
user_template.replace("{{MSG}}", message.content),
|
137 |
+
unsafe_allow_html=True,
|
138 |
+
)
|
139 |
+
else:
|
140 |
+
st.write(
|
141 |
+
bot_template.replace("{{MSG}}", message.content), unsafe_allow_html=True
|
142 |
+
)
|
143 |
+
|
144 |
+
|
145 |
+
def main():
|
146 |
+
|
147 |
+
st.set_page_config(
|
148 |
+
page_title="Chat with a Bot that tries to answer questions about multiple PDFs",
|
149 |
+
page_icon=":books:",
|
150 |
+
)
|
151 |
+
|
152 |
+
st.markdown("# Chat with a Bot")
|
153 |
+
st.markdown("This bot tries to answer questions about multiple PDFs.")
|
154 |
+
|
155 |
+
st.write(css, unsafe_allow_html=True)
|
156 |
+
|
157 |
+
# set huggingface hub token in st.text_input widget
|
158 |
+
# then hide the input
|
159 |
+
huggingface_token = st.text_input("Enter your HuggingFace Hub token", type="password")
|
160 |
+
#openai_api_key = st.text_input("Enter your OpenAI API key", type="password")
|
161 |
+
|
162 |
+
# set this key as an environment variable
|
163 |
+
os.environ["HUGGINGFACEHUB_API_TOKEN"] = huggingface_token
|
164 |
+
#os.environ["OPENAI_API_KEY"] = openai_api_key
|
165 |
+
|
166 |
+
if "conversation" not in st.session_state:
|
167 |
+
st.session_state.conversation = None
|
168 |
+
if "chat_history" not in st.session_state:
|
169 |
+
st.session_state.chat_history = None
|
170 |
+
|
171 |
+
st.header("Chat with a Bot 🤖���� that tries to answer questions about multiple PDFs :books:")
|
172 |
+
user_question = st.text_input("Ask a question about your documents:")
|
173 |
+
if user_question:
|
174 |
+
handle_userinput(user_question)
|
175 |
+
|
176 |
+
with st.sidebar:
|
177 |
+
st.subheader("Your documents")
|
178 |
+
pdf_docs = st.file_uploader(
|
179 |
+
"Upload your PDFs here and click on 'Process'", accept_multiple_files=True
|
180 |
+
)
|
181 |
+
if st.button("Process"):
|
182 |
+
with st.spinner("Processing"):
|
183 |
+
# get pdf text
|
184 |
+
raw_text = get_pdf_text(pdf_docs)
|
185 |
+
|
186 |
+
# get the text chunks
|
187 |
+
text_chunks = get_text_chunks(raw_text)
|
188 |
+
|
189 |
+
# create vector store
|
190 |
+
vectorstore = get_vectorstore(text_chunks)
|
191 |
+
|
192 |
+
# create conversation chain
|
193 |
+
st.session_state.conversation = get_conversation_chain(vectorstore)
|
194 |
+
|
195 |
+
|
196 |
+
if __name__ == "__main__":
|
197 |
+
main()
|