Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -16,8 +16,142 @@ from bs4 import BeautifulSoup
|
|
16 |
from collections import deque
|
17 |
from audio_recorder_streamlit import audio_recorder
|
18 |
|
19 |
-
|
20 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
def chat_with_file_contents(prompt, file_content):
|
23 |
conversation = [{'role': 'system', 'content': 'You are a helpful assistant.'}]
|
@@ -74,4 +208,39 @@ def main():
|
|
74 |
else:
|
75 |
if st.button(f"Chat about Section {i+1}"):
|
76 |
st.write('Reasoning with your inputs...')
|
77 |
-
response = chat_with_model(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
from collections import deque
|
17 |
from audio_recorder_streamlit import audio_recorder
|
18 |
|
19 |
+
|
20 |
+
def generate_filename(prompt, file_type):
|
21 |
+
central = pytz.timezone('US/Central')
|
22 |
+
safe_date_time = datetime.now(central).strftime("%m%d_%I%M")
|
23 |
+
safe_prompt = "".join(x for x in prompt if x.isalnum())[:45]
|
24 |
+
return f"{safe_date_time}_{safe_prompt}.{file_type}"
|
25 |
+
|
26 |
+
def chat_with_model(prompt, document_section):
|
27 |
+
model = model_choice
|
28 |
+
conversation = [{'role': 'system', 'content': 'You are a helpful assistant.'}]
|
29 |
+
conversation.append({'role': 'user', 'content': prompt})
|
30 |
+
if len(document_section)>0:
|
31 |
+
conversation.append({'role': 'assistant', 'content': document_section})
|
32 |
+
response = openai.ChatCompletion.create(model=model, messages=conversation)
|
33 |
+
#return response
|
34 |
+
return response['choices'][0]['message']['content']
|
35 |
+
|
36 |
+
def transcribe_audio(openai_key, file_path, model):
|
37 |
+
OPENAI_API_URL = "https://api.openai.com/v1/audio/transcriptions"
|
38 |
+
headers = {
|
39 |
+
"Authorization": f"Bearer {openai_key}",
|
40 |
+
}
|
41 |
+
with open(file_path, 'rb') as f:
|
42 |
+
data = {'file': f}
|
43 |
+
response = requests.post(OPENAI_API_URL, headers=headers, files=data, data={'model': model})
|
44 |
+
if response.status_code == 200:
|
45 |
+
st.write(response.json())
|
46 |
+
|
47 |
+
response2 = chat_with_model(response.json().get('text'), '') # *************************************
|
48 |
+
st.write('Responses:')
|
49 |
+
#st.write(response)
|
50 |
+
st.write(response2)
|
51 |
+
return response.json().get('text')
|
52 |
+
else:
|
53 |
+
st.write(response.json())
|
54 |
+
st.error("Error in API call.")
|
55 |
+
return None
|
56 |
+
|
57 |
+
def save_and_play_audio(audio_recorder):
|
58 |
+
audio_bytes = audio_recorder()
|
59 |
+
if audio_bytes:
|
60 |
+
filename = generate_filename("Recording", "wav")
|
61 |
+
with open(filename, 'wb') as f:
|
62 |
+
f.write(audio_bytes)
|
63 |
+
st.audio(audio_bytes, format="audio/wav")
|
64 |
+
return filename
|
65 |
+
return None
|
66 |
+
|
67 |
+
def create_file(filename, prompt, response):
|
68 |
+
if filename.endswith(".txt"):
|
69 |
+
with open(filename, 'w') as file:
|
70 |
+
file.write(f"Prompt:\n{prompt}\nResponse:\n{response}")
|
71 |
+
elif filename.endswith(".htm"):
|
72 |
+
with open(filename, 'w') as file:
|
73 |
+
file.write(f"<h1>Prompt:</h1> <p>{prompt}</p> <h1>Response:</h1> <p>{response}</p>")
|
74 |
+
elif filename.endswith(".md"):
|
75 |
+
with open(filename, 'w') as file:
|
76 |
+
file.write(f"# Prompt:\n{prompt}\n# Response:\n{response}")
|
77 |
+
def truncate_document(document, length):
|
78 |
+
return document[:length]
|
79 |
+
def divide_document(document, max_length):
|
80 |
+
return [document[i:i+max_length] for i in range(0, len(document), max_length)]
|
81 |
+
def get_table_download_link(file_path):
|
82 |
+
with open(file_path, 'r') as file:
|
83 |
+
data = file.read()
|
84 |
+
b64 = base64.b64encode(data.encode()).decode()
|
85 |
+
file_name = os.path.basename(file_path)
|
86 |
+
ext = os.path.splitext(file_name)[1] # get the file extension
|
87 |
+
if ext == '.txt':
|
88 |
+
mime_type = 'text/plain'
|
89 |
+
elif ext == '.py':
|
90 |
+
mime_type = 'text/plain'
|
91 |
+
elif ext == '.xlsx':
|
92 |
+
mime_type = 'text/plain'
|
93 |
+
elif ext == '.csv':
|
94 |
+
mime_type = 'text/plain'
|
95 |
+
elif ext == '.htm':
|
96 |
+
mime_type = 'text/html'
|
97 |
+
elif ext == '.md':
|
98 |
+
mime_type = 'text/markdown'
|
99 |
+
else:
|
100 |
+
mime_type = 'application/octet-stream' # general binary data type
|
101 |
+
href = f'<a href="data:{mime_type};base64,{b64}" target="_blank" download="{file_name}">{file_name}</a>'
|
102 |
+
return href
|
103 |
+
|
104 |
+
|
105 |
+
|
106 |
+
|
107 |
+
|
108 |
+
|
109 |
+
def CompressXML(xml_text):
|
110 |
+
root = ET.fromstring(xml_text)
|
111 |
+
for elem in list(root.iter()):
|
112 |
+
if isinstance(elem.tag, str) and 'Comment' in elem.tag:
|
113 |
+
elem.parent.remove(elem)
|
114 |
+
return ET.tostring(root, encoding='unicode', method="xml")
|
115 |
+
|
116 |
+
def read_file_content(file,max_length):
|
117 |
+
if file.type == "application/json":
|
118 |
+
content = json.load(file)
|
119 |
+
return str(content)
|
120 |
+
elif file.type == "text/html" or file.type == "text/htm":
|
121 |
+
content = BeautifulSoup(file, "html.parser")
|
122 |
+
return content.text
|
123 |
+
elif file.type == "application/xml" or file.type == "text/xml":
|
124 |
+
tree = ET.parse(file)
|
125 |
+
root = tree.getroot()
|
126 |
+
xml = CompressXML(ET.tostring(root, encoding='unicode'))
|
127 |
+
return xml
|
128 |
+
elif file.type == "text/markdown" or file.type == "text/md":
|
129 |
+
md = mistune.create_markdown()
|
130 |
+
content = md(file.read().decode())
|
131 |
+
return content
|
132 |
+
elif file.type == "text/plain":
|
133 |
+
return file.getvalue().decode()
|
134 |
+
else:
|
135 |
+
return ""
|
136 |
+
|
137 |
+
|
138 |
+
# Sidebar and global
|
139 |
+
openai.api_key = os.getenv('OPENAI_KEY')
|
140 |
+
st.set_page_config(page_title="GPT Streamlit Document Reasoner",layout="wide")
|
141 |
+
menu = ["htm", "txt", "xlsx", "csv", "md", "py"] #619
|
142 |
+
choice = st.sidebar.selectbox("Output File Type:", menu)
|
143 |
+
model_choice = st.sidebar.radio("Select Model:", ('gpt-3.5-turbo', 'gpt-3.5-turbo-0301'))
|
144 |
+
|
145 |
+
# Audio, transcribe, GPT:
|
146 |
+
filename = save_and_play_audio(audio_recorder)
|
147 |
+
if filename is not None:
|
148 |
+
transcription = transcribe_audio(openai.api_key, filename, "whisper-1")
|
149 |
+
st.write(transcription)
|
150 |
+
gptOutput = chat_with_model(transcription, '') # *************************************
|
151 |
+
filename = generate_filename(transcription, choice)
|
152 |
+
create_file(filename, transcription, gptOutput)
|
153 |
+
st.sidebar.markdown(get_table_download_link(filename), unsafe_allow_html=True)
|
154 |
+
|
155 |
|
156 |
def chat_with_file_contents(prompt, file_content):
|
157 |
conversation = [{'role': 'system', 'content': 'You are a helpful assistant.'}]
|
|
|
208 |
else:
|
209 |
if st.button(f"Chat about Section {i+1}"):
|
210 |
st.write('Reasoning with your inputs...')
|
211 |
+
response = chat_with_model(user_prompt, section)
|
212 |
+
st.write('Response:')
|
213 |
+
st.write(response)
|
214 |
+
document_responses[i] = response
|
215 |
+
filename = generate_filename(f"{user_prompt}_Section_{i+1}", choice)
|
216 |
+
create_file(filename, user_prompt, response)
|
217 |
+
st.sidebar.markdown(get_table_download_link(filename), unsafe_allow_html=True)
|
218 |
+
|
219 |
+
# ... document_responses logic remains the same
|
220 |
+
|
221 |
+
all_files = [file for file in glob.glob("*.{}".format(choice))]
|
222 |
+
|
223 |
+
for file in all_files:
|
224 |
+
col1, col2, col3 = st.sidebar.columns([5,1,1]) # adjust the ratio as needed
|
225 |
+
with col1:
|
226 |
+
st.markdown(get_table_download_link(file), unsafe_allow_html=True)
|
227 |
+
with col2:
|
228 |
+
if st.button("π", key="read_"+file): # search emoji button
|
229 |
+
with open(file, 'r') as f:
|
230 |
+
file_contents = f.read()
|
231 |
+
file_content_area = st.text_area("File Contents:", file_contents, height=100)
|
232 |
+
if st.button('π¬ Chat with file content'):
|
233 |
+
st.write('Reasoning with your inputs...')
|
234 |
+
response = chat_with_file_contents(user_prompt, file_contents)
|
235 |
+
st.write('Response:')
|
236 |
+
st.write(response)
|
237 |
+
filename = generate_filename(user_prompt, choice)
|
238 |
+
create_file(filename, user_prompt, response)
|
239 |
+
st.sidebar.markdown(get_table_download_link(filename), unsafe_allow_html=True)
|
240 |
+
with col3:
|
241 |
+
if st.button("π", key="delete_"+file):
|
242 |
+
os.remove(file)
|
243 |
+
st.experimental_rerun()
|
244 |
+
|
245 |
+
if __name__ == "__main__":
|
246 |
+
main()
|