-
Notifications
You must be signed in to change notification settings - Fork 1
/
main.py
171 lines (151 loc) · 9.05 KB
/
main.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
import uuid
import json
import openai
import streamlit as st
from langdetect import detect
import modules
# Some global parameters
top_n = 3 # Top n search results from Bing will be used to generate answers.
n_questions = 5 # The number of questions that ChatGPT needs to come up with.
enable_logging = True # Turn on or off whether to log user feedback to Google Forms.
st.set_page_config(layout='wide')
chatgpt_provider = 'azure' # openai or azure
if chatgpt_provider == 'openai':
openai.api_key = st.secrets.openai_keys.key
elif chatgpt_provider == 'azure':
openai.api_type = 'azure'
openai.api_version = '2023-03-15-preview'
openai.api_base = st.secrets.azure_openai.api_base
openai.api_key = st.secrets.azure_openai.api_key
# Some variables to control the execution of this program
if 'generated' not in st.session_state:
# Whether the generation is completed
st.session_state.generated = 0
if 'generation_id' not in st.session_state:
# Random ID for anonymous feedback
st.session_state.generation_id = uuid.uuid4()
for i in range(n_questions):
# Whether the "I like this one" button is clicked for each question
if f'question_{i}_feedback' not in st.session_state:
st.session_state[f'question_{i}_feedback'] = 0
# Layout of the sidebar
with st.sidebar:
st.markdown('# :information_source: User Info')
st.markdown('### What is lateral reading?')
st.markdown('Lateral reading is a critical thinking approach used to evaluate the credibility and accuracy of '
'information found online by stepping away from the initial source and exploring other sources to '
'verify its authenticity.')
st.markdown('### How can lateral reading help fight online misinformation?')
st.markdown('Lateral reading can reduce the risk of being misled by misinformation, propaganda, and other forms '
'of disinformation. By cross-referencing information with multiple sources, users can become more '
'informed and responsible consumers of information, promoting a better online community.')
st.markdown('### How does this tool support lateral reading?')
st.markdown('All you need is to copy and paste a text into the input box and then click on the "Probe" '
'button. This tool (powered by [OpenAI](https://openai.com/)) will generate **five** '
'questions you may want to ask and provide answers by summarizing relevant documents found by [Bing]('
'https://bing.com/). You can give feedback on the generated content. The generation time depends on '
'the responsiveness of API calls. Please wait till the generation completes '
'before interacting with the page.')
# Building the main page
st.markdown('# :bulb: ReadProbe [](https://github.com/DakeZhang1998/ReadProbe)')
st.markdown('An AI-powered tool to support **lateral reading**. (NEW) Now GPT-4 is behind the scene!')
st.markdown('### :pushpin: Instructions')
st.markdown('Please read the information on the left before using this tool. The length of your input should be more '
'than 10 words (enough context for meaningful generation) and less than 2000 words (due to the limit of '
'ChatGPT). A desired input would consist of statements, rather than questions (e.g., who is the current '
'prime minister of Canada) or requests (e.g., please find information about the current prime minister of '
'Canada). Only English is supported at the current stage.')
st.markdown('### :robot_face: Start Here')
demo = 'Demo:\n' \
'Elon Musk\'s account was briefly suspended by an outgoing Twitter employee.'
input_text = st.text_area(label='Help me probe into:', value=demo)
col1, col2 = st.columns(2)
with col1:
probe_button = st.button(':mag: Probe')
with col2:
refresh_button = st.button(':arrows_counterclockwise: Reset')
if refresh_button:
modules.refresh(top_n=top_n)
# Interactions
if probe_button or st.session_state.generated == 1:
if probe_button and st.session_state.generated == 1:
# If the content has been generated but the user still clicks on the "probe" button, rerun the program.
modules.refresh(top_n=top_n)
# Input Check
with st.spinner('Checking input ...'):
if len(input_text.strip()) == 0:
st.warning('Please input your text.')
st.stop()
if detect(input_text) != 'en':
st.warning('Only English is supported for now. Please try another input.')
st.stop()
if chatgpt_provider == 'openai' and not modules.input_check(input_text):
st.warning('Your input may contain harmful content. Please try another input.')
st.stop()
if len(input_text.split()) < 10:
st.warning('Your input is too short. Please try longer input.')
st.stop()
if len(input_text.split()) > 2000:
st.warning('Your input is too long. Please try shorter input.')
st.stop()
# Generate questions
with st.spinner('Generating questions ...'):
questions = modules.generate_questions(input_text, provider=chatgpt_provider)
# Search online and generate answers
records = []
for i, question in enumerate(questions):
record = [question]
with st.expander(f'**{i + 1}\. {question}**', expanded=True):
with st.spinner('Searching online ...'):
search_result = modules.bing_search(question, top_n=top_n)
with st.spinner('Generating answers ...'):
record.append([url for url, doc in search_result])
answer = modules.summarize(question, [doc for url, doc in search_result], provider=chatgpt_provider)
if len(answer) == 0:
st.warning('Azure OpenAI servers are at capacity. Please try again later.')
st.stop()
modules.refresh(top_n=top_n)
answer = str(answer).replace('$', '\\$')
record.append(answer)
seen = set()
# Insert attribution citations
for ind, (url, doc) in enumerate(search_result):
if f'[{ind + 1}]' in answer:
seen.add(ind)
answer = answer.replace(f'[{ind + 1}]', f'[[{ind + 1}]]({url})')
st.markdown(f'> {answer}')
# Show URLs to the relevant web documents
for j in range(len(search_result)):
output = f'{j + 1}. {search_result[j][0]}'
if j not in seen:
output = f'{output} (This source is not directly leveraged in attribution but could be ' \
f'potentially relevant.)'
st.markdown(output)
# User feedback: I like this one
button = st.button(':thumbsup: I like this one', key=f'thumbsup_for_q{i + 1}')
if button or st.session_state[f'question_{i}_feedback'] == 1:
st.info('Thanks for your feedback!')
if button or st.session_state[f'question_{i}_feedback'] == 0 and enable_logging:
modules.log_data(log_id=str(st.session_state.generation_id), user_input=input_text,
output=json.dumps(record), action='thumbsup')
st.session_state[f'question_{i}_feedback'] = 1
records.append(record)
if probe_button and st.session_state.generated == 0 and enable_logging:
modules.log_data(log_id=str(st.session_state.generation_id), user_input=input_text, output=json.dumps(records),
action='generation')
st.session_state.generated = 1
st.markdown('### :warning: Disclaimer')
st.markdown('Content generation in this web application is performed mainly by calling [OpenAI](https://openai.com/) '
'APIs and [Bing](https://bing.com/) APIs. Even though we perform some processing to improve the quality of '
'the generated content, due to our limited knowledge of the models behind those services, we can not make '
'sure the generated content won\'t contain biased, unethical, incorrect, or controversial information. '
'Please be aware of the potential risks of using this tool. **The content generated here doesn\'t '
'represent the opinions of the developers in this project.** If you notice potentially malicious content '
'generated by this app, please open an [issue](https://github.com/DakeZhang1998/ReadProbe/issues) to '
'report. Your valuable feedback is deeply appreciated.')
st.markdown('### Privacy')
st.markdown('Your usage data will be collected anonymously to improve this tool, including your inputs and '
'thumbs-ups. Be aware not to include sensitive data in your inputs. Your personal information will '
'not be collected, such as your IP address, location, etc. Using this tool indicates you agree with our'
'privacy policy.')