Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -7,35 +7,68 @@ import json
|
|
7 |
import mistune
|
8 |
import pytz
|
9 |
import math
|
|
|
|
|
10 |
from datetime import datetime
|
11 |
from openai import ChatCompletion
|
12 |
from xml.etree import ElementTree as ET
|
13 |
from bs4 import BeautifulSoup
|
14 |
from collections import deque
|
|
|
15 |
|
16 |
openai.api_key = os.getenv('OPENAI_KEY')
|
17 |
-
st.set_page_config(
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
choice = st.sidebar.selectbox("Output file type:", menu)
|
23 |
-
choicePrefix = "Output file type is "
|
24 |
-
if choice == "txt":
|
25 |
-
st.sidebar.write(choicePrefix + "Text File.")
|
26 |
-
elif choice == "htm":
|
27 |
-
st.sidebar.write(choicePrefix + "HTML5.")
|
28 |
-
elif choice == "md":
|
29 |
-
st.sidebar.write(choicePrefix + "Markdown.")
|
30 |
-
elif choice == "py":
|
31 |
-
st.sidebar.write(choicePrefix + "Python Code.")
|
32 |
|
33 |
def generate_filename(prompt, file_type):
|
34 |
central = pytz.timezone('US/Central')
|
35 |
safe_date_time = datetime.now(central).strftime("%m%d_%I%M")
|
36 |
-
safe_prompt = "".join(x for x in prompt if x.isalnum())[:
|
37 |
return f"{safe_date_time}_{safe_prompt}.{file_type}"
|
38 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
39 |
def create_file(filename, prompt, response):
|
40 |
if filename.endswith(".txt"):
|
41 |
with open(filename, 'w') as file:
|
@@ -46,21 +79,10 @@ def create_file(filename, prompt, response):
|
|
46 |
elif filename.endswith(".md"):
|
47 |
with open(filename, 'w') as file:
|
48 |
file.write(f"# Prompt:\n{prompt}\n# Response:\n{response}")
|
49 |
-
|
50 |
def truncate_document(document, length):
|
51 |
return document[:length]
|
52 |
-
|
53 |
def divide_document(document, max_length):
|
54 |
return [document[i:i+max_length] for i in range(0, len(document), max_length)]
|
55 |
-
|
56 |
-
def chat_with_model(prompt, document_section):
|
57 |
-
model = "gpt-3.5-turbo"
|
58 |
-
conversation = [{'role': 'system', 'content': 'You are a helpful assistant.'}]
|
59 |
-
conversation.append({'role': 'user', 'content': prompt})
|
60 |
-
conversation.append({'role': 'assistant', 'content': document_section})
|
61 |
-
response = openai.ChatCompletion.create(model=model, messages=conversation)
|
62 |
-
return response['choices'][0]['message']['content']
|
63 |
-
|
64 |
def get_table_download_link(file_path):
|
65 |
with open(file_path, 'r') as file:
|
66 |
data = file.read()
|
@@ -69,6 +91,12 @@ def get_table_download_link(file_path):
|
|
69 |
ext = os.path.splitext(file_name)[1] # get the file extension
|
70 |
if ext == '.txt':
|
71 |
mime_type = 'text/plain'
|
|
|
|
|
|
|
|
|
|
|
|
|
72 |
elif ext == '.htm':
|
73 |
mime_type = 'text/html'
|
74 |
elif ext == '.md':
|
@@ -78,6 +106,22 @@ def get_table_download_link(file_path):
|
|
78 |
href = f'<a href="data:{mime_type};base64,{b64}" target="_blank" download="{file_name}">{file_name}</a>'
|
79 |
return href
|
80 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
81 |
def CompressXML(xml_text):
|
82 |
root = ET.fromstring(xml_text)
|
83 |
for elem in list(root.iter()):
|
@@ -107,14 +151,14 @@ def read_file_content(file,max_length):
|
|
107 |
return ""
|
108 |
|
109 |
def main():
|
110 |
-
user_prompt = st.text_area("
|
111 |
|
112 |
collength, colupload = st.columns([2,3]) # adjust the ratio as needed
|
113 |
with collength:
|
114 |
#max_length = 12000 - optimal for gpt35 turbo. 2x=24000 for gpt4. 8x=96000 for gpt4-32k.
|
115 |
-
max_length = st.slider("
|
116 |
with colupload:
|
117 |
-
uploaded_file = st.file_uploader("
|
118 |
|
119 |
document_sections = deque()
|
120 |
document_responses = {}
|
@@ -136,7 +180,7 @@ def main():
|
|
136 |
st.markdown(f"**Section {i+1}**\n{document_responses[i]}")
|
137 |
else:
|
138 |
if st.button(f"Chat about Section {i+1}"):
|
139 |
-
st.write('
|
140 |
response = chat_with_model(user_prompt, section)
|
141 |
st.write('Response:')
|
142 |
st.write(response)
|
@@ -146,7 +190,7 @@ def main():
|
|
146 |
st.sidebar.markdown(get_table_download_link(filename), unsafe_allow_html=True)
|
147 |
|
148 |
if st.button('π¬ Chat'):
|
149 |
-
st.write('
|
150 |
response = chat_with_model(user_prompt, ''.join(list(document_sections)))
|
151 |
st.write('Response:')
|
152 |
st.write(response)
|
@@ -155,15 +199,18 @@ def main():
|
|
155 |
create_file(filename, user_prompt, response)
|
156 |
st.sidebar.markdown(get_table_download_link(filename), unsafe_allow_html=True)
|
157 |
|
158 |
-
all_files = glob.glob("
|
|
|
|
|
|
|
159 |
for file in all_files:
|
160 |
-
col1,
|
161 |
with col1:
|
162 |
st.markdown(get_table_download_link(file), unsafe_allow_html=True)
|
163 |
-
with
|
164 |
-
if st.button("π", key=file):
|
165 |
os.remove(file)
|
166 |
st.experimental_rerun()
|
167 |
-
|
168 |
if __name__ == "__main__":
|
169 |
-
main()
|
|
|
7 |
import mistune
|
8 |
import pytz
|
9 |
import math
|
10 |
+
import requests
|
11 |
+
|
12 |
from datetime import datetime
|
13 |
from openai import ChatCompletion
|
14 |
from xml.etree import ElementTree as ET
|
15 |
from bs4 import BeautifulSoup
|
16 |
from collections import deque
|
17 |
+
from audio_recorder_streamlit import audio_recorder
|
18 |
|
19 |
openai.api_key = os.getenv('OPENAI_KEY')
|
20 |
+
st.set_page_config(page_title="GPT Streamlit Document Reasoner",layout="wide")
|
21 |
+
|
22 |
+
menu = ["htm", "txt", "xlsx", "csv", "md", "py"] #619
|
23 |
+
choice = st.sidebar.selectbox("Output File Type:", menu)
|
24 |
+
model_choice = st.sidebar.radio("Select Model:", ('gpt-3.5-turbo', 'gpt-3.5-turbo-0301'))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
|
26 |
def generate_filename(prompt, file_type):
|
27 |
central = pytz.timezone('US/Central')
|
28 |
safe_date_time = datetime.now(central).strftime("%m%d_%I%M")
|
29 |
+
safe_prompt = "".join(x for x in prompt if x.isalnum())[:45]
|
30 |
return f"{safe_date_time}_{safe_prompt}.{file_type}"
|
31 |
|
32 |
+
def chat_with_model(prompt, document_section):
|
33 |
+
model = model_choice
|
34 |
+
conversation = [{'role': 'system', 'content': 'You are a helpful assistant.'}]
|
35 |
+
conversation.append({'role': 'user', 'content': prompt})
|
36 |
+
if len(document_section)>0:
|
37 |
+
conversation.append({'role': 'assistant', 'content': document_section})
|
38 |
+
response = openai.ChatCompletion.create(model=model, messages=conversation)
|
39 |
+
return response
|
40 |
+
#return response['choices'][0]['message']['content']
|
41 |
+
|
42 |
+
def transcribe_audio(openai_key, file_path, model):
|
43 |
+
OPENAI_API_URL = "https://api.openai.com/v1/audio/transcriptions"
|
44 |
+
headers = {
|
45 |
+
"Authorization": f"Bearer {openai_key}",
|
46 |
+
}
|
47 |
+
with open(file_path, 'rb') as f:
|
48 |
+
data = {'file': f}
|
49 |
+
response = requests.post(OPENAI_API_URL, headers=headers, files=data, data={'model': model})
|
50 |
+
if response.status_code == 200:
|
51 |
+
st.write(response.json())
|
52 |
+
response2 = chat_with_model(response.json().get('text'), '')
|
53 |
+
st.write('Responses:')
|
54 |
+
#st.write(response)
|
55 |
+
st.write(response2)
|
56 |
+
return response.json().get('text')
|
57 |
+
else:
|
58 |
+
st.write(response.json())
|
59 |
+
st.error("Error in API call.")
|
60 |
+
return None
|
61 |
+
|
62 |
+
def save_and_play_audio(audio_recorder):
|
63 |
+
audio_bytes = audio_recorder()
|
64 |
+
if audio_bytes:
|
65 |
+
filename = generate_filename("Recording", "wav")
|
66 |
+
with open(filename, 'wb') as f:
|
67 |
+
f.write(audio_bytes)
|
68 |
+
st.audio(audio_bytes, format="audio/wav")
|
69 |
+
return filename
|
70 |
+
return None
|
71 |
+
|
72 |
def create_file(filename, prompt, response):
|
73 |
if filename.endswith(".txt"):
|
74 |
with open(filename, 'w') as file:
|
|
|
79 |
elif filename.endswith(".md"):
|
80 |
with open(filename, 'w') as file:
|
81 |
file.write(f"# Prompt:\n{prompt}\n# Response:\n{response}")
|
|
|
82 |
def truncate_document(document, length):
|
83 |
return document[:length]
|
|
|
84 |
def divide_document(document, max_length):
|
85 |
return [document[i:i+max_length] for i in range(0, len(document), max_length)]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
86 |
def get_table_download_link(file_path):
|
87 |
with open(file_path, 'r') as file:
|
88 |
data = file.read()
|
|
|
91 |
ext = os.path.splitext(file_name)[1] # get the file extension
|
92 |
if ext == '.txt':
|
93 |
mime_type = 'text/plain'
|
94 |
+
elif ext == '.py':
|
95 |
+
mime_type = 'text/plain'
|
96 |
+
elif ext == '.xlsx':
|
97 |
+
mime_type = 'text/plain'
|
98 |
+
elif ext == '.csv':
|
99 |
+
mime_type = 'text/plain'
|
100 |
elif ext == '.htm':
|
101 |
mime_type = 'text/html'
|
102 |
elif ext == '.md':
|
|
|
106 |
href = f'<a href="data:{mime_type};base64,{b64}" target="_blank" download="{file_name}">{file_name}</a>'
|
107 |
return href
|
108 |
|
109 |
+
|
110 |
+
|
111 |
+
# Audio, transcribe, GPT:
|
112 |
+
filename = save_and_play_audio(audio_recorder)
|
113 |
+
if filename is not None:
|
114 |
+
transcription = transcribe_audio(openai.api_key, filename, "whisper-1")
|
115 |
+
st.write(transcription)
|
116 |
+
gptOutput = chat_with_model(transcription, '') # push transcript through as prompt
|
117 |
+
filename = generate_filename(transcription, choice)
|
118 |
+
create_file(filename, transcription, gptOutput)
|
119 |
+
st.sidebar.markdown(get_table_download_link(filename), unsafe_allow_html=True)
|
120 |
+
|
121 |
+
|
122 |
+
|
123 |
+
|
124 |
+
|
125 |
def CompressXML(xml_text):
|
126 |
root = ET.fromstring(xml_text)
|
127 |
for elem in list(root.iter()):
|
|
|
151 |
return ""
|
152 |
|
153 |
def main():
|
154 |
+
user_prompt = st.text_area("Enter prompts, instructions & questions:", '', height=100)
|
155 |
|
156 |
collength, colupload = st.columns([2,3]) # adjust the ratio as needed
|
157 |
with collength:
|
158 |
#max_length = 12000 - optimal for gpt35 turbo. 2x=24000 for gpt4. 8x=96000 for gpt4-32k.
|
159 |
+
max_length = st.slider("File section length for large files", min_value=1000, max_value=128000, value=12000, step=1000)
|
160 |
with colupload:
|
161 |
+
uploaded_file = st.file_uploader("Add a file for context:", type=["xml", "json", "xlsx","csv","html", "htm", "md", "txt"])
|
162 |
|
163 |
document_sections = deque()
|
164 |
document_responses = {}
|
|
|
180 |
st.markdown(f"**Section {i+1}**\n{document_responses[i]}")
|
181 |
else:
|
182 |
if st.button(f"Chat about Section {i+1}"):
|
183 |
+
st.write('Reasoning with your inputs...')
|
184 |
response = chat_with_model(user_prompt, section)
|
185 |
st.write('Response:')
|
186 |
st.write(response)
|
|
|
190 |
st.sidebar.markdown(get_table_download_link(filename), unsafe_allow_html=True)
|
191 |
|
192 |
if st.button('π¬ Chat'):
|
193 |
+
st.write('Reasoning with your inputs...')
|
194 |
response = chat_with_model(user_prompt, ''.join(list(document_sections)))
|
195 |
st.write('Response:')
|
196 |
st.write(response)
|
|
|
199 |
create_file(filename, user_prompt, response)
|
200 |
st.sidebar.markdown(get_table_download_link(filename), unsafe_allow_html=True)
|
201 |
|
202 |
+
all_files = glob.glob("*.*")
|
203 |
+
all_files = [file for file in all_files if len(os.path.splitext(file)[0]) >= 20] # exclude files with short names
|
204 |
+
all_files.sort(key=lambda x: (os.path.splitext(x)[1], x), reverse=True) # sort by file type and file name in descending order
|
205 |
+
|
206 |
for file in all_files:
|
207 |
+
col1, col3 = st.sidebar.columns([5,1]) # adjust the ratio as needed
|
208 |
with col1:
|
209 |
st.markdown(get_table_download_link(file), unsafe_allow_html=True)
|
210 |
+
with col3:
|
211 |
+
if st.button("π", key="delete_"+file):
|
212 |
os.remove(file)
|
213 |
st.experimental_rerun()
|
214 |
+
|
215 |
if __name__ == "__main__":
|
216 |
+
main()
|