Spaces:
Build error
Build error
Upload 2 files
Browse files- Step4_voice.py +213 -0
- requirement.txt +4 -0
Step4_voice.py
ADDED
@@ -0,0 +1,213 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Library
|
2 |
+
import openai
|
3 |
+
import streamlit as st
|
4 |
+
import pandas as pd
|
5 |
+
from datetime import datetime
|
6 |
+
from TTS.api import TTS
|
7 |
+
import whisper
|
8 |
+
from audio_recorder import record
|
9 |
+
|
10 |
+
# Custom Streamlit app title and icon
|
11 |
+
st.set_page_config(
|
12 |
+
page_title="VietAI Bot",
|
13 |
+
page_icon=":robot_face:",
|
14 |
+
)
|
15 |
+
|
16 |
+
# Set the title
|
17 |
+
st.title("[VietAI-NTI] ChatGPT")
|
18 |
+
|
19 |
+
# Sidebar Configuration
|
20 |
+
st.sidebar.title(":gear: Model Configuration")
|
21 |
+
|
22 |
+
# Set OPENAI API
|
23 |
+
openai.api_key = st.sidebar.text_input('Your OpenAI API key here:')
|
24 |
+
|
25 |
+
# User Input and AI Response
|
26 |
+
user_input_type = st.sidebar.selectbox("Choose input type:", ["Chat", "Record Audio"])
|
27 |
+
|
28 |
+
# Model Name Selector
|
29 |
+
model_name = st.sidebar.selectbox(
|
30 |
+
"Select a Model",
|
31 |
+
["gpt-3.5-turbo", "gpt-4"], # Add more model names as needed
|
32 |
+
key="model_name",
|
33 |
+
)
|
34 |
+
|
35 |
+
# Temperature Slider
|
36 |
+
temperature = st.sidebar.slider(
|
37 |
+
":thermometer: Temperature",
|
38 |
+
min_value=0.2,
|
39 |
+
max_value=2.0,
|
40 |
+
value=1.0,
|
41 |
+
step=0.1,
|
42 |
+
key="temperature",
|
43 |
+
)
|
44 |
+
|
45 |
+
# Max tokens Slider
|
46 |
+
max_tokens = st.sidebar.slider(
|
47 |
+
":straight_ruler: Max Tokens",
|
48 |
+
min_value=1,
|
49 |
+
max_value=4095,
|
50 |
+
value=256,
|
51 |
+
step=1,
|
52 |
+
key="max_tokens",
|
53 |
+
)
|
54 |
+
|
55 |
+
# Top p Slider
|
56 |
+
# top_p = st.sidebar.slider(
|
57 |
+
# "🎯 Top P",
|
58 |
+
# min_value=0.00,
|
59 |
+
# max_value=1.00,
|
60 |
+
# value=1.00,
|
61 |
+
# step=0.01,
|
62 |
+
# key="top_p",
|
63 |
+
# )
|
64 |
+
|
65 |
+
# Presence penalty Slider
|
66 |
+
# presence_penalty = st.sidebar.slider(
|
67 |
+
# "🚫 Presence penalty",
|
68 |
+
# min_value=0.00,
|
69 |
+
# max_value=2.00,
|
70 |
+
# value=0.00,
|
71 |
+
# step=0.01,
|
72 |
+
# key="presence_penalty",
|
73 |
+
# )
|
74 |
+
|
75 |
+
# Frequency penalty Slider
|
76 |
+
# frequency_penalty = st.sidebar.slider(
|
77 |
+
# "🤐 Frequency penalty",
|
78 |
+
# min_value=0.00,
|
79 |
+
# max_value=2.00,
|
80 |
+
# value=0.00,
|
81 |
+
# step=0.01,
|
82 |
+
# key="frequency_penalty",
|
83 |
+
# )
|
84 |
+
|
85 |
+
# TEXT2SPEECH MODEL
|
86 |
+
# Instantiate the TTS class
|
87 |
+
tts = TTS(TTS().list_models()[13])
|
88 |
+
def convert_2_speech(given_text):
|
89 |
+
tts.tts_to_file(text=given_text, file_path="response.wav")
|
90 |
+
return("response.wav")
|
91 |
+
|
92 |
+
# SPEECH2TEXT MODEL
|
93 |
+
model_whisper = whisper.load_model("tiny.en")
|
94 |
+
def convert_2_text(speech):
|
95 |
+
user_message = model_whisper.transcribe(speech)["text"]
|
96 |
+
return user_message
|
97 |
+
|
98 |
+
# CHAT MODEL
|
99 |
+
# Initialize DataFrame to store chat history
|
100 |
+
chat_history_df = pd.DataFrame(columns=["Timestamp", "Chat"])
|
101 |
+
|
102 |
+
# Reset Button
|
103 |
+
if st.sidebar.button(":arrows_counterclockwise: Reset Chat"):
|
104 |
+
# Save the chat history to the DataFrame before clearing it
|
105 |
+
if st.session_state.messages:
|
106 |
+
timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
|
107 |
+
chat_history = "\n".join([f"{m['role']}: {m['content']}" for m in st.session_state.messages])
|
108 |
+
new_entry = pd.DataFrame({"Timestamp": [timestamp], "Chat": [chat_history]})
|
109 |
+
chat_history_df = pd.concat([chat_history_df, new_entry], ignore_index=True)
|
110 |
+
|
111 |
+
# Save the DataFrame to a CSV file
|
112 |
+
chat_history_df.to_csv("chat_history.csv", index=False)
|
113 |
+
|
114 |
+
# Clear the chat messages and reset the full response
|
115 |
+
st.session_state.messages = []
|
116 |
+
full_response = ""
|
117 |
+
|
118 |
+
# Initialize Chat Messages
|
119 |
+
if "messages" not in st.session_state:
|
120 |
+
st.session_state.messages = []
|
121 |
+
|
122 |
+
# Initialize full_response outside the user input check
|
123 |
+
full_response = ""
|
124 |
+
|
125 |
+
# Display Chat History
|
126 |
+
for message in st.session_state.messages:
|
127 |
+
with st.chat_message(message["role"]):
|
128 |
+
st.markdown(message["content"])
|
129 |
+
|
130 |
+
# User Input and AI Response
|
131 |
+
# if st.sidebar.button(":microphone: Record Audio"):
|
132 |
+
# st.audio(None, format="audio/wav", start_time=0)
|
133 |
+
# recorded_audio = st.audio_recorder(key="audio_recorder")
|
134 |
+
|
135 |
+
# if recorded_audio:
|
136 |
+
# st.success("Recording complete. Click 'Generate Response' to transcribe and generate AI response.")
|
137 |
+
|
138 |
+
|
139 |
+
|
140 |
+
# User Input and AI Response
|
141 |
+
# For "Chat mode"
|
142 |
+
if user_input_type == "Chat":
|
143 |
+
if prompt := st.chat_input("What is up?"):
|
144 |
+
# System
|
145 |
+
st.session_state.messages.append({"role": "system", "content": "You are a helpful assistant named Jarvis"})
|
146 |
+
|
147 |
+
# User
|
148 |
+
st.session_state.messages.append({"role": "user", "content": prompt})
|
149 |
+
with st.chat_message("user"):
|
150 |
+
st.markdown(prompt)
|
151 |
+
|
152 |
+
# Assistant
|
153 |
+
with st.chat_message("assistant"):
|
154 |
+
with st.status("Generating response..."):
|
155 |
+
message_placeholder = st.empty()
|
156 |
+
for response in openai.ChatCompletion.create(
|
157 |
+
model=model_name, # Use the selected model name
|
158 |
+
messages=[
|
159 |
+
{"role": m["role"], "content": m["content"]}
|
160 |
+
for m in st.session_state.messages
|
161 |
+
],
|
162 |
+
temperature=temperature, # Set temperature
|
163 |
+
max_tokens=max_tokens, # Set max tokens
|
164 |
+
top_p=top_p, # Set top p
|
165 |
+
frequency_penalty=frequency_penalty, # Set frequency penalty
|
166 |
+
presence_penalty=presence_penalty, # Set presence penalty
|
167 |
+
stream=True,
|
168 |
+
):
|
169 |
+
full_response += response.choices[0].delta.get("content", "")
|
170 |
+
message_placeholder.markdown(full_response + "▌")
|
171 |
+
message_placeholder.markdown(full_response)
|
172 |
+
|
173 |
+
st.session_state.messages.append({"role": "assistant", "content": full_response})
|
174 |
+
st.audio(convert_2_speech(full_response))
|
175 |
+
|
176 |
+
elif user_input_type == "Record Audio":
|
177 |
+
# Record audio when the "Record Audio" button is clicked
|
178 |
+
if st.button("Record Audio"):
|
179 |
+
st.write("Recording... Please speak for 10 seconds.")
|
180 |
+
output = record(seconds=10, filename='my_recording.wav')
|
181 |
+
st.write("Recording complete!")
|
182 |
+
|
183 |
+
# Convert the recorded audio to text using the Whisper model
|
184 |
+
user_message = convert_2_text(output)
|
185 |
+
|
186 |
+
# Display the transcribed text as user input
|
187 |
+
st.session_state.messages.append({"role": "user", "content": user_message})
|
188 |
+
with st.chat_message("user"):
|
189 |
+
st.markdown(user_message)
|
190 |
+
|
191 |
+
# Assistant
|
192 |
+
with st.chat_message("assistant"):
|
193 |
+
with st.status("Generating response..."):
|
194 |
+
message_placeholder = st.empty()
|
195 |
+
for response in openai.ChatCompletion.create(
|
196 |
+
model=model_name, # Use the selected model name
|
197 |
+
messages=[
|
198 |
+
{"role": m["role"], "content": m["content"]}
|
199 |
+
for m in st.session_state.messages
|
200 |
+
],
|
201 |
+
temperature=temperature, # Set temperature
|
202 |
+
max_tokens=max_tokens, # Set max tokens
|
203 |
+
top_p=top_p, # Set top p
|
204 |
+
frequency_penalty=frequency_penalty, # Set frequency penalty
|
205 |
+
presence_penalty=presence_penalty, # Set presence penalty
|
206 |
+
stream=True,
|
207 |
+
):
|
208 |
+
full_response += response.choices[0].delta.get("content", "")
|
209 |
+
message_placeholder.markdown(full_response + "▌")
|
210 |
+
message_placeholder.markdown(full_response)
|
211 |
+
|
212 |
+
st.session_state.messages.append({"role": "assistant", "content": full_response})
|
213 |
+
st.audio(convert_2_speech(full_response))
|
requirement.txt
ADDED
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
1 |
+
openai
|
2 |
+
openai-whisper
|
3 |
+
streamlit
|
4 |
+
TTS == 0.17.4
|