timmy0079 commited on
Commit
dc4f29c
β€’
1 Parent(s): ac6ac21

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +149 -5
app.py CHANGED
@@ -1,7 +1,151 @@
1
- import gradio as gr
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2
 
3
- def greet(name):
4
- return "Hello " + name + "!!"
5
 
6
- iface = gr.Interface(fn=greet, inputs="text", outputs="text")
7
- iface.launch()
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import streamlit as st
2
+ from dotenv import load_dotenv
3
+ from PyPDF2 import PdfReader
4
+ from langchain.text_splitter import CharacterTextSplitter, RecursiveCharacterTextSplitter
5
+ from langchain.embeddings import OpenAIEmbeddings, HuggingFaceInstructEmbeddings
6
+ from langchain.vectorstores import FAISS, Chroma
7
+ from langchain.embeddings import HuggingFaceEmbeddings # General embeddings from HuggingFace models.
8
+ from langchain.chat_models import ChatOpenAI
9
+ from langchain.memory import ConversationBufferMemory
10
+ from langchain.chains import ConversationalRetrievalChain
11
+ from htmlTemplates import css, bot_template, user_template
12
+ from langchain.llms import HuggingFaceHub, LlamaCpp, CTransformers # For loading transformer models.
13
+ from langchain.document_loaders import PyPDFLoader, TextLoader, JSONLoader, CSVLoader
14
+ import tempfile # μž„μ‹œ νŒŒμΌμ„ μƒμ„±ν•˜κΈ° μœ„ν•œ λΌμ΄λΈŒλŸ¬λ¦¬μž…λ‹ˆλ‹€.
15
+ import os
16
 
 
 
17
 
18
+ # PDF λ¬Έμ„œλ‘œλΆ€ν„° ν…μŠ€νŠΈλ₯Ό μΆ”μΆœν•˜λŠ” ν•¨μˆ˜μž…λ‹ˆλ‹€.
19
+ def get_pdf_text(pdf_docs):
20
+ temp_dir = tempfile.TemporaryDirectory() # μž„μ‹œ 디렉토리λ₯Ό μƒμ„±ν•©λ‹ˆλ‹€.
21
+ temp_filepath = os.path.join(temp_dir.name, pdf_docs.name) # μž„μ‹œ 파일 경둜λ₯Ό μƒμ„±ν•©λ‹ˆλ‹€.
22
+ with open(temp_filepath, "wb") as f: # μž„μ‹œ νŒŒμΌμ„ λ°”μ΄λ„ˆλ¦¬ μ“°κΈ° λͺ¨λ“œλ‘œ μ—½λ‹ˆλ‹€.
23
+ f.write(pdf_docs.getvalue()) # PDF λ¬Έμ„œμ˜ λ‚΄μš©μ„ μž„μ‹œ νŒŒμΌμ— μ”λ‹ˆλ‹€.
24
+ pdf_loader = PyPDFLoader(temp_filepath) # PyPDFLoaderλ₯Ό μ‚¬μš©ν•΄ PDFλ₯Ό λ‘œλ“œν•©λ‹ˆλ‹€.
25
+ pdf_doc = pdf_loader.load() # ν…μŠ€νŠΈλ₯Ό μΆ”μΆœν•©λ‹ˆλ‹€.
26
+ return pdf_doc # μΆ”μΆœν•œ ν…μŠ€νŠΈλ₯Ό λ°˜ν™˜ν•©λ‹ˆλ‹€.
27
+
28
+ # 과제
29
+ # μ•„λž˜ ν…μŠ€νŠΈ μΆ”μΆœ ν•¨μˆ˜λ₯Ό μž‘μ„±
30
+
31
+ def get_text_file(docs):
32
+ pass
33
+
34
+
35
+ def get_csv_file(docs):
36
+ pass
37
+
38
+ def get_json_file(docs):
39
+ pass
40
+
41
+
42
+ # λ¬Έμ„œλ“€μ„ μ²˜λ¦¬ν•˜μ—¬ ν…μŠ€νŠΈ 청크둜 λ‚˜λˆ„λŠ” ν•¨μˆ˜μž…λ‹ˆλ‹€.
43
+ def get_text_chunks(documents):
44
+ text_splitter = RecursiveCharacterTextSplitter(
45
+ chunk_size=1000, # 청크의 크기λ₯Ό μ§€μ •ν•©λ‹ˆλ‹€.
46
+ chunk_overlap=200, # 청크 μ‚¬μ΄μ˜ 쀑볡을 μ§€μ •ν•©λ‹ˆλ‹€.
47
+ length_function=len # ν…μŠ€νŠΈμ˜ 길이λ₯Ό μΈ‘μ •ν•˜λŠ” ν•¨μˆ˜λ₯Ό μ§€μ •ν•©λ‹ˆλ‹€.
48
+ )
49
+
50
+ documents = text_splitter.split_documents(documents) # λ¬Έμ„œλ“€μ„ 청크둜 λ‚˜λˆ•λ‹ˆλ‹€
51
+ return documents # λ‚˜λˆˆ 청크λ₯Ό λ°˜ν™˜ν•©λ‹ˆλ‹€.
52
+
53
+
54
+ # ν…μŠ€νŠΈ μ²­ν¬λ“€λ‘œλΆ€ν„° 벑터 μŠ€ν† μ–΄λ₯Ό μƒμ„±ν•˜λŠ” ν•¨μˆ˜μž…λ‹ˆλ‹€.
55
+ def get_vectorstore(text_chunks):
56
+ # OpenAI μž„λ² λ”© λͺ¨λΈμ„ λ‘œλ“œν•©λ‹ˆλ‹€. (Embedding models - Ada v2)
57
+
58
+ embeddings = OpenAIEmbeddings()
59
+ vectorstore = FAISS.from_documents(text_chunks, embeddings) # FAISS 벑터 μŠ€ν† μ–΄λ₯Ό μƒμ„±ν•©λ‹ˆλ‹€.
60
+
61
+ return vectorstore # μƒμ„±λœ 벑터 μŠ€ν† μ–΄λ₯Ό λ°˜ν™˜ν•©λ‹ˆλ‹€.
62
+
63
+
64
+ def get_conversation_chain(vectorstore):
65
+ gpt_model_name = 'gpt-3.5-turbo'
66
+ llm = ChatOpenAI(model_name = gpt_model_name) #gpt-3.5 λͺ¨λΈ λ‘œλ“œ
67
+
68
+ # λŒ€ν™” 기둝을 μ €μž₯ν•˜κΈ° μœ„ν•œ λ©”λͺ¨λ¦¬λ₯Ό μƒμ„±ν•©λ‹ˆλ‹€.
69
+ memory = ConversationBufferMemory(
70
+ memory_key='chat_history', return_messages=True)
71
+ # λŒ€ν™” 검색 체인을 μƒμ„±ν•©λ‹ˆλ‹€.
72
+ conversation_chain = ConversationalRetrievalChain.from_llm(
73
+ llm=llm,
74
+ retriever=vectorstore.as_retriever(),
75
+ memory=memory
76
+ )
77
+ return conversation_chain
78
+
79
+ # μ‚¬μš©μž μž…λ ₯을 μ²˜λ¦¬ν•˜λŠ” ν•¨μˆ˜μž…λ‹ˆλ‹€.
80
+ def handle_userinput(user_question):
81
+ # λŒ€ν™” 체인을 μ‚¬μš©ν•˜μ—¬ μ‚¬μš©μž μ§ˆλ¬Έμ— λŒ€ν•œ 응닡을 μƒμ„±ν•©λ‹ˆλ‹€.
82
+ response = st.session_state.conversation({'question': user_question})
83
+ # λŒ€ν™” 기둝을 μ €μž₯ν•©λ‹ˆλ‹€.
84
+ st.session_state.chat_history = response['chat_history']
85
+
86
+ for i, message in enumerate(st.session_state.chat_history):
87
+ if i % 2 == 0:
88
+ st.write(user_template.replace(
89
+ "{{MSG}}", message.content), unsafe_allow_html=True)
90
+ else:
91
+ st.write(bot_template.replace(
92
+ "{{MSG}}", message.content), unsafe_allow_html=True)
93
+
94
+
95
+ def main():
96
+ load_dotenv()
97
+ st.set_page_config(page_title="Chat with multiple Files",
98
+ page_icon=":books:")
99
+ st.write(css, unsafe_allow_html=True)
100
+
101
+ if "conversation" not in st.session_state:
102
+ st.session_state.conversation = None
103
+ if "chat_history" not in st.session_state:
104
+ st.session_state.chat_history = None
105
+
106
+ st.header("Chat with multiple Files :")
107
+ user_question = st.text_input("Ask a question about your documents:")
108
+ if user_question:
109
+ handle_userinput(user_question)
110
+
111
+ with st.sidebar:
112
+ openai_key = st.text_input("Paste your OpenAI API key (sk-...)")
113
+ if openai_key:
114
+ os.environ["OPENAI_API_KEY"] = openai_key
115
+
116
+ st.subheader("Your documents")
117
+ docs = st.file_uploader(
118
+ "Upload your PDFs here and click on 'Process'", accept_multiple_files=True)
119
+ if st.button("Process"):
120
+ with st.spinner("Processing"):
121
+ # get pdf text
122
+ doc_list = []
123
+
124
+ for file in docs:
125
+ print('file - type : ', file.type)
126
+ if file.type == 'text/plain':
127
+ # file is .txt
128
+ doc_list.extend(get_text_file(file))
129
+ elif file.type in ['application/octet-stream', 'application/pdf']:
130
+ # file is .pdf
131
+ doc_list.extend(get_pdf_text(file))
132
+ elif file.type == 'text/csv':
133
+ # file is .csv
134
+ doc_list.extend(get_csv_file(file))
135
+ elif file.type == 'application/json':
136
+ # file is .json
137
+ doc_list.extend(get_json_file(file))
138
+
139
+ # get the text chunks
140
+ text_chunks = get_text_chunks(doc_list)
141
+
142
+ # create vector store
143
+ vectorstore = get_vectorstore(text_chunks)
144
+
145
+ # create conversation chain
146
+ st.session_state.conversation = get_conversation_chain(
147
+ vectorstore)
148
+
149
+
150
+ if __name__ == '__main__':
151
+ main()