-
-
Notifications
You must be signed in to change notification settings - Fork 14
/
ui.py
235 lines (226 loc) · 9.3 KB
/
ui.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
import streamlit as st
import openai
import requests
import base64
import os
import re
from datetime import datetime
from dotenv import load_dotenv
import time
load_dotenv()
# Show logo but resize to the screen https://devxt.com/wp-content/uploads/2023/01/Logo-1024x316.png
st.image(
"https://devxt.com/wp-content/uploads/2023/01/Logo-1024x316.png",
width=300,
)
EZLOCALAI_SERVER = os.getenv("EZLOCALAI_URL", "http://localhost:8091")
EZLOCALAI_API_KEY = os.getenv("EZLOCALAI_API_KEY", "none")
DEFAULT_LLM = os.getenv("DEFAULT_MODEL", "TheBloke/phi-2-dpo-GGUF")
VISION_MODEL = os.getenv("VISION_MODEL", None)
SD_MODEL = os.getenv("SD_MODEL", "stabilityai/sdxl-turbo")
IMG_ENABLED = os.getenv("IMG_ENABLED", "false")
WHISPER_MODEL = os.getenv("WHISPER_MODEL", "base.en")
if "/" in DEFAULT_LLM:
link_to_model = f"https://huggingface.co/{DEFAULT_LLM}"
else:
link_to_model = f"https://huggingface.co/models?search={DEFAULT_LLM}"
st.markdown(
f"""
[](https://github.com/DevXT-LLC/ezlocalai) [](https://hub.docker.com/r/joshxt/ezlocalai)
## ezLocal.ai Demo
**Language model:** [{DEFAULT_LLM}]({link_to_model})
"""
)
if VISION_MODEL:
st.markdown(
f"""
**Vision model:** [{VISION_MODEL}](https://huggingface.co/{VISION_MODEL})
"""
)
if IMG_ENABLED.lower() == "true":
st.markdown(
f"""
**Image Generation model:** [{SD_MODEL}](https://huggingface.co/{SD_MODEL})
"""
)
openai.base_url = f"{EZLOCALAI_SERVER}/v1/"
openai.api_key = EZLOCALAI_API_KEY if EZLOCALAI_API_KEY else EZLOCALAI_SERVER
HEADERS = {
"Content-Type": "application/json",
"Authorization": f"{EZLOCALAI_API_KEY}",
"ngrok-skip-browser-warning": "true",
}
waiting_for_server = False
def get_voices():
return ["Morgan_Freeman", "DukeNukem", "HAL9000"]
# Commented for speed, but this works if we want to get the voices from the server
"""
global EZLOCALAI_SERVER
global HEADERS
global waiting_for_server
while True:
try:
voices = requests.get(
f"{EZLOCALAI_SERVER}/v1/audio/voices", headers=HEADERS
)
voices = voices.json()["voices"]
waiting_for_server = False
return voices
except:
if waiting_for_server == False:
st.spinner("Waiting for server to start...")
waiting_for_server = True
time.sleep(1)
"""
def display_content(content):
global EZLOCALAI_SERVER
global HEADERS
outputs_url = f"{EZLOCALAI_SERVER}/outputs/"
os.makedirs("outputs", exist_ok=True)
if "http://localhost:8091/outputs/" in content:
if outputs_url != "http://localhost:8091/outputs/":
content = content.replace("http://localhost:8091/outputs/", outputs_url)
if "<audio controls>" in content or " " not in content:
try:
audio_response = content.split("data:audio/wav;base64,")[1].split('" type')[
0
]
except:
audio_response = content
file_name = f"outputs/{datetime.now().strftime('%Y-%m-%d-%H-%M-%S')}.wav"
with open(file_name, "wb") as fh:
fh.write(base64.b64decode(audio_response))
st.audio(file_name, format="audio/wav", start_time=0)
if outputs_url in content:
urls = re.findall(f"{re.escape(outputs_url)}[^\"' ]+", content)
urls = urls[0].split("\n\n")
for url in urls:
file_name = url.split("/")[-1]
url = f"{outputs_url}{file_name}"
data = requests.get(url, headers=HEADERS).content
if url.endswith(".jpg") or url.endswith(".png"):
content = content.replace(url, "")
st.image(data, use_column_width=True)
elif url.endswith(".mp4"):
content = content.replace(url, "")
st.audio(data, format="audio/mp4", start_time=0)
elif url.endswith(".wav"):
content = content.replace(url, "")
st.audio(data, format="audio/wav", start_time=0)
st.markdown(f"[Download Audio]({url})")
st.markdown(content, unsafe_allow_html=True)
mode = st.radio(
"Choose which ezLocalai example to try",
["Multimodal Interaction Example", "PDF to Audio Example"],
)
if mode == "PDF to Audio Example":
st.markdown("# PDF to Audio\n\nUpload a PDF file to convert it to audio.")
voice_drop_down = st.selectbox("Text-to-Speech Voice", get_voices(), index=0)
uploaded_file = st.file_uploader("Choose a PDF file", type="pdf")
if uploaded_file:
start_time = time.time()
st.markdown("---")
st.spinner("Thinking...")
tts_response = openai.audio.speech.create(
model="tts",
voice=voice_drop_down,
input=f"data:application/pdf;base64,{base64.b64encode(uploaded_file.read()).decode('utf-8')}",
)
new_audio_url = tts_response.content.decode("utf-8")
end_time = time.time() # Record the end time
elapsed_time = end_time - start_time
# If response time is longer than 60 seconds, split the response time into minutes and seconds
if elapsed_time > 60:
minutes = int(elapsed_time // 60)
seconds = elapsed_time % 60
st.success(f"Response time: {minutes} minutes and {seconds:.2f} seconds")
else:
st.success(f"Response time: {elapsed_time:.2f} seconds")
display_content(new_audio_url)
st.balloons()
else:
st.markdown(
"# Multimodal Interaction Example\n\nThis is a multimodal interaction example. You can upload an image and describe it in the text box. The model will generate a response based on the image and text input.\n\nNote: This is a single interaction example, it is not a conversation, the model will not remember the last thing you said."
)
show_advanced_options = st.checkbox(
"Show Advanced Options", key="show_advanced_options"
)
if show_advanced_options:
SYSTEM_MESSAGE = st.text_area(
"System Prompt",
"",
)
DEFAULT_MAX_TOKENS = st.number_input(
"Max Output Tokens", min_value=10, max_value=300000, value=1024
)
DEFAULT_TEMPERATURE = st.number_input(
"Temperature", min_value=0.0, max_value=1.0, value=0.5
)
DEFAULT_TOP_P = st.number_input(
"Top P", min_value=0.0, max_value=1.0, value=0.9
)
else:
SYSTEM_MESSAGE = ""
DEFAULT_MAX_TOKENS = 1024
DEFAULT_TEMPERATURE = 0.5
DEFAULT_TOP_P = 0.9
with st.form("chat"):
voice_drop_down = st.selectbox(
"Text-to-Speech Response Voice", ["None"] + get_voices(), index=0
)
uploaded_file = st.file_uploader("Upload an image")
prompt = st.text_area("Your Message:", "Describe each stage of this image.")
send = st.form_submit_button("Send")
if prompt != "" and send:
start_time = time.time()
st.markdown("---")
st.spinner("Thinking...")
messages = []
if SYSTEM_MESSAGE != "":
messages.append({"role": "system", "content": SYSTEM_MESSAGE})
if uploaded_file:
messages.append(
{
"role": "user",
"content": [
{"type": "text", "text": prompt},
{
"type": f"{uploaded_file.type.split('/')[0]}_url",
f"{uploaded_file.type.split('/')[0]}_url": {
"url": f"data:{uploaded_file.type};base64,{base64.b64encode(uploaded_file.read()).decode('utf-8')}",
},
},
],
},
)
if uploaded_file.type.startswith("image"):
st.image(uploaded_file, use_column_width=True)
if messages == []:
messages = [
{"role": "user", "content": prompt},
]
extra_body = (
None if voice_drop_down == "None" else {"voice": voice_drop_down}
)
response = openai.chat.completions.create(
model=DEFAULT_LLM,
messages=messages,
temperature=DEFAULT_TEMPERATURE,
max_tokens=DEFAULT_MAX_TOKENS,
top_p=DEFAULT_TOP_P,
stream=False,
extra_body=extra_body,
)
display_content(response.choices[0].message.content)
end_time = time.time() # Record the end time
elapsed_time = end_time - start_time
# If response time is longer than 60 seconds, split the response time into minutes and seconds
if elapsed_time > 60:
minutes = int(elapsed_time // 60)
seconds = elapsed_time % 60
st.success(
f"Response time: {minutes} minutes and {seconds:.2f} seconds"
)
else:
st.success(f"Response time: {elapsed_time:.2f} seconds")
st.balloons()