-
Notifications
You must be signed in to change notification settings - Fork 13
/
document_analyzer.py
359 lines (312 loc) · 17 KB
/
document_analyzer.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
'''
Description: This file contains the code for the streamlit chatbot with documents.
The answer can be read out loud via text to speech.
run the app within the directory:
conda activate ./.conda
streamlit run ./document_analyter.demo.py
'''
import os
import streamlit as st
# loading the OpenAI api key from .env
from dotenv import load_dotenv, find_dotenv
from helperfunctions import *
import os
from langchain.embeddings.openai import OpenAIEmbeddings
#from langchain.vectorstores import Chroma
from langchain.vectorstores import FAISS
import azure.cognitiveservices.speech as speechsdk
from langchain.docstore.document import Document
from langchain.prompts import PromptTemplate
import streamlit as st
import ftfy
import azure.cognitiveservices.speech as speechsdk
if 'priorproject' not in st.session_state:
st.session_state.priorproject=''
if 'language' not in st.session_state:
st.session_state.language='en-US'
load_dotenv(find_dotenv(), override=True)
os.environ["OPENAI_API_BASE"] = os.environ["AZURE_OPENAI_ENDPOINT"]
os.environ["OPENAI_API_KEY"] = os.environ["AZURE_OPENAI_API_KEY"]
os.environ["OPENAI_API_VERSION"] = os.environ["AZURE_OPENAI_API_VERSION"]
os.environ["OPENAI_API_TYPE"] = "azure"
speech_config = speechsdk.SpeechConfig(subscription=os.environ.get('SPEECH_KEY'), region=os.environ.get('SPEECH_REGION'))
speech_config.speech_recognition_language=st.session_state.language
audio_config = speechsdk.audio.AudioConfig(use_default_microphone=True)
speech_recognizer = speechsdk.SpeechRecognizer(speech_config=speech_config, audio_config=audio_config)
if 'project' not in st.session_state:
refresh_project_list()
# Set Vector list
if 'vector_index_list' not in st.session_state and 'project' in st.session_state:
refresh_vector_index_list()
if 'k' not in st.session_state:
st.session_state.k=1
if 'topic_list' not in st.session_state:
refresh_topic_list()
if len(st.session_state.topic_list)>0:
if 'topic' not in st.session_state:
st.session_state.topic = st.session_state.topic_list[0]
if len(st.session_state.topic_list)>0:
if 'query_list' not in st.session_state:
load_topic()
if 'vector_index_list' not in st.session_state and 'project' in st.session_state:
refresh_vector_index_list()
if os.path.isfile("projects/"+st.session_state.project+"/logo.png"):
st.image("projects/"+st.session_state.project+'/logo.png')
else:
st.image('img.png')
st.subheader('Azure OpenAI - Document Analyzer')
if 'document_name' not in st.session_state:
st.session_state.document_name = ''
with st.sidebar:
with st.container():
st.subheader("Project")
with st.expander("Edit Projects"):
with st.form(key="Add new project",clear_on_submit=True):
new_project_name=st.text_input('New Project Name')
col1,col2 = st.columns([1,1])
with col1:
submitted = st.form_submit_button("Add project")
with col2:
del_project = st.form_submit_button('Delete project')
if del_project:
delete_project(st.session_state.project)
if submitted:
add_project(new_project_name)
st.selectbox('Select project', st.session_state.project_list,key='project',label_visibility="hidden")
if 'project' in st.session_state and st.session_state.project != st.session_state.priorproject:
loadproject()
st.session_state.priorproject=st.session_state.project
if(len(st.session_state.project_list)>0):
with st.container():
st.subheader("Document")
with st.expander("Add new Document"):
# file uploader widget
uploaded_file = st.file_uploader('Upload a file:', type=['pdf'])
# token size number widget
token_size = st.number_input('Token size:', min_value=128, max_value=1024, value=512)
# add data button widget
analyze = st.button('Analyze with Azure AI Document Intelligence and create Vector Index')
if analyze and uploaded_file:
jsonfilename = "projects/"+st.session_state.project+'/files/'+uploaded_file.name+'.json'
if os.path.isfile(jsonfilename):
print('JSON File '+jsonfilename+' already exists. Skipping Analysis.')
else:
with st.spinner('Analyzing '+uploaded_file.name+' with Azure AI Document Intelligence...'):
#get full path of uploaded file
bytes_data = uploaded_file.read()
file_name = os.path.join("projects/"+st.session_state.project+'/files/', uploaded_file.name)
with open(file_name, 'wb') as f:
f.write(bytes_data)
from analyzer import analyze_general_documents as di
jsonfilename = di(st.session_state.project,uploaded_file.name)
st.success('JSON File '+jsonfilename+' created successfully.')
with st.spinner('Reading, chunking and embedding Document Intelligence Results from '+uploaded_file.name+'...'):
from indexer import createindex as ci
st.session_state.vector_index_name=ci(st.session_state.project,jsonfilename,uploaded_file.name,token_size)
load_embeddings()
refresh_vector_index_list()
st.success('Document Intelligence Results chunked and embedded. Vector store '+ st.session_state.document_name+' created successfully.')
st.selectbox('Select Document',st.session_state.vector_index_list ,index=0,key="vector_index_name",on_change=load_embeddings)
if len(st.session_state.vector_index_list)>0:
with st.container():
st.subheader("Topic")
with st.expander("Edit Topics"):
with st.form(key="Add new Topic",clear_on_submit=True):
new_topic_name=st.text_input('New Topic Name')
col1,col2 = st.columns([1,1])
with col1:
submitted = st.form_submit_button("Add Topic")
with col2:
del_topic = st.form_submit_button('Delete Topic')
if del_topic:
delete_topic(st.session_state.topic)
if submitted:
add_topic(new_topic_name)
st.selectbox('Select Topics', st.session_state.topic_list,key="topic",on_change=load_topic)
if len(st.session_state.topic_list)>0:
with st.form(key="Ground Truth",clear_on_submit=True):
col61,col62 = st.columns([2,1])
with col61:
ground_truth=st.text_input('Pages to check against:',key="ground_truth",value=str(",".join(str(x) for x in getgroundtruthpages())))
with col62:
st.form_submit_button("Add Ground Truth",on_click=setgroundtruthpages)
if 'vs' not in st.session_state:
load_embeddings()
if(len(st.session_state.topic_list)>0):
tab1, tab2,tab3 = st.tabs(["Document Viewer","Context Queries","Question Answering"])
#Tab 1: Document Viewer
with tab1:
col1, col2,col3,col4,col5 = st.columns([2.8,0.8,1.2,1,1])
with col1:
col21, col22 = st.columns([1,1])
with col21:
startpage=st.number_input('Start Page',min_value=1, max_value=1000,key="startpage")
with col22:
endpage=st.number_input('End Page',min_value=1, max_value=1000,key="endpage")
with col2:
show=st.button('Show Pages')
with col3:
fulldocument=st.button('Full document',on_click=resetpage)
with col4:
tables=st.button('Show tables')
with col5:
keyvalues=st.button('Show Key Values')
if tables:
with st.expander("Tables",expanded=True):
if 'tables' in st.session_state:
st.markdown(st.session_state.tables)
if keyvalues:
with st.expander("Key Values",expanded=True):
if 'keyvalues' in st.session_state:
st.json(st.session_state.keyvalues)
if show or fulldocument:
if fulldocument:
startpage=1
endpage=len(st.session_state.pagecontent)
for i in range(startpage,endpage):
if i==1:
st.markdown('***Begin of Document***')
st.markdown(ftfy.fix_encoding(st.session_state.pagecontent[str(i)]))
st.markdown('***Page '+str(i)+' of '+str(len(st.session_state.pagecontent))+'***')
if i==len(st.session_state.pagecontent)-1:
st.markdown('***End of Document***')
if startpage==endpage:
if startpage==1:
st.markdown('***Begin of Document***')
st.markdown(ftfy.fix_encoding(st.session_state.pagecontent[str(startpage)]))
st.markdown('***Page '+str(startpage)+' of '+str(len(st.session_state.pagecontent))+'***')
if endpage==len(st.session_state.pagecontent):
st.markdown('***Begin of Document***')
#Tab 2: Context Queries
with tab2:
col1,col2,col5 = st.columns([0.2,2,0.4])
with col1:
stt = st.button(':studio_microphone:',key="stt")
with col2:
with st.form(key="Add new Query",clear_on_submit=True):
col3,col4 = st.columns([3.4,1])
with col3:
new_query_name=st.text_input('New Query',label_visibility='collapsed')
with col4:
# k number input widget
add_query_button = st.form_submit_button("Add Query")
if stt:
st.info("Speak into your microphone.")
speech_recognition_result = speech_recognizer.recognize_once_async().get()
if speech_recognition_result.reason == speechsdk.ResultReason.RecognizedSpeech:
print("Recognized: {}".format(speech_recognition_result.text))
add_query(speech_recognition_result.text)
st.session_state.query=speech_recognition_result.text
elif speech_recognition_result.reason == speechsdk.ResultReason.NoMatch:
st.warning("No speech could be recognized: {}".format(speech_recognition_result.no_match_details))
elif speech_recognition_result.reason == speechsdk.ResultReason.Canceled:
cancellation_details = speech_recognition_result.cancellation_details
st.warning("Speech Recognition canceled: {}".format(cancellation_details.reason))
if cancellation_details.reason == speechsdk.CancellationReason.Error:
st.warning("Error details: {}".format(cancellation_details.error_details))
if add_query_button:
add_query(new_query_name)
st.session_state.query = new_query_name
getcontext()
with col5:
st.session_state.k=st.number_input('k',value=1,min_value=1, max_value=20,)
st.text("List of stored context queries:")
col1,col2= st.columns([3.35,1])
with col1:
st.selectbox('Select a Query',st.session_state.query_list ,index=0,label_visibility='collapsed',key="query")
with col2:
col91,col92 = st.columns([1,1])
with col91:
run_query=st.button('Query')
with col92:
del_query = st.button('Delete',on_click=delete_query)
if run_query or stt:
getcontext()
if 'context' in st.session_state:
with st.expander("Query Results",expanded=False):
for i,content in enumerate(st.session_state.querycontent):
st.header('Result from page '+st.session_state.querypages[i]+" with score "+str(st.session_state.queryscores[i])+":")
st.markdown(content)
with st.expander(st.session_state.reducedpages,expanded=True):
st.header('Query: ')
st.markdown(st.session_state.query)
st.header('Pages according to Similarity Search: ')
st.markdown(st.session_state.context)
#Tab 3: Question Answering
with tab3:
col81, col82,col83 = st.columns([1, 1,1])
with col81:
enable_tts=st.checkbox('Text to Speech')
# Select language
st.session_state.language = st.selectbox('Select language',("en-US", "de-DE"),index=0,label_visibility='collapsed')
with col82:
st.session_state.model = "gpt-35-turbo" #st.selectbox('Model',("gpt-35-turbo","text-davinci-003"),index=0)
st.markdown("**Model:**")
st.write(st.session_state.model)
with col83:
# t number input widget
st.session_state.t = st.number_input('Temperature', min_value=0.0, max_value=1.0, value=0.0)
#q = st.text_input('Ask a question about the content of your file: '+st.session_state.document_name,key="questionwidget")
# user's question text input widget
col200,col201 = st.columns([0.4,5])
with col200:
stt2 = st.button(':studio_microphone:',key="stt2")
with col201:
with st.form(key="Add new Question",clear_on_submit=True):
col203,col204 = st.columns([3.9,1])
with col203:
new_question_name=st.text_input('New Question',label_visibility='collapsed')
with col204:
# k number input widget
add_question_button = st.form_submit_button("Add Question")
if stt2:
st.info("Speak into your microphone.")
speech_recognition_result = speech_recognizer.recognize_once_async().get()
if speech_recognition_result.reason == speechsdk.ResultReason.RecognizedSpeech:
print("Recognized: {}".format(speech_recognition_result.text))
add_question(speech_recognition_result.text)
st.session_state.question=speech_recognition_result.text
elif speech_recognition_result.reason == speechsdk.ResultReason.NoMatch:
st.warning("No speech could be recognized: {}".format(speech_recognition_result.no_match_details))
elif speech_recognition_result.reason == speechsdk.ResultReason.Canceled:
cancellation_details = speech_recognition_result.cancellation_details
st.warning("Speech Recognition canceled: {}".format(cancellation_details.reason))
if cancellation_details.reason == speechsdk.CancellationReason.Error:
st.warning("Error details: {}".format(cancellation_details.error_details))
if add_question_button:
add_question(new_question_name)
st.session_state.question = new_question_name
askquestion()
st.text("List of stored questions:")
col111,col112= st.columns([3.8,1])
with col111:
st.selectbox('Select a Question',st.session_state.question_list ,index=0,label_visibility='collapsed',key="question")
with col112:
col191,col192 = st.columns([0.8,1.1])
with col191:
run_question=st.button('Ask')
with col192:
del_question = st.button('Delete',on_click=delete_question,key="delquestion")
if run_question or stt2:
askquestion()
if 'answer' in st.session_state:
st.header('Question: ')
st.markdown(st.session_state.question)
# text area widget for the LLM answer
#st.text_area('Azure OpenAI Answer: ', value=answer,height=200)
st.header('Azure OpenAI Answer: ')
with st.container():
st.markdown(st.session_state.answer)
sourcetext="Source: "+st.session_state.vector_index_name+" - Pages:"+str(",".join(str(x) for x in st.session_state.sourcepages))
if enable_tts:
synthesize_text(st.session_state.answer)
st.divider()
with st.expander(sourcetext):
counter=1
st.header('Sources:')
for page in st.session_state.sourcepages:
st.divider()
st.write("Source: "+st.session_state.document_name+" - Page:"+str(page))
st.divider()
st.markdown(ftfy.fix_encoding(st.session_state.pagecontent[str(page)]))
counter=counter+1