Update ui/tabs.py
Browse files- ui/tabs.py +47 -53
ui/tabs.py
CHANGED
|
@@ -53,24 +53,20 @@ def create_all_tabs(audio_service: AudioService, chat_service: ChatService,
|
|
| 53 |
with gr.Tab("Stream Object Detection"):
|
| 54 |
create_streaming_object_detection()
|
| 55 |
def create_sambanova_voice_tab(sambanova_service):
|
| 56 |
-
"""Tạo tab Sambanova Voice AI -
|
| 57 |
|
| 58 |
with gr.Blocks() as sambanova_tab:
|
| 59 |
gr.Markdown("## 🤖 Sambanova Voice AI - Llama 3.2 3B")
|
| 60 |
-
gr.Markdown("Trò chuyện voice-to-voice với model Llama 3.2 3B
|
| 61 |
|
| 62 |
-
# State variables
|
| 63 |
chatbot = gr.Chatbot(
|
| 64 |
type="messages",
|
| 65 |
value=[],
|
| 66 |
label="💬 Hội thoại Voice",
|
| 67 |
-
height=400
|
| 68 |
-
elem_id="sambanova_main_chatbot"
|
| 69 |
-
)
|
| 70 |
-
conversation_state = gr.State(
|
| 71 |
-
value=[],
|
| 72 |
-
elem_id="sambanova_main_state"
|
| 73 |
)
|
|
|
|
| 74 |
|
| 75 |
with gr.Row():
|
| 76 |
with gr.Column(scale=1):
|
|
@@ -78,7 +74,7 @@ def create_sambanova_voice_tab(sambanova_service):
|
|
| 78 |
gr.Markdown("""
|
| 79 |
**Cách sử dụng:**
|
| 80 |
1. Nhấn **Bắt đầu Voice Chat**
|
| 81 |
-
2. Nói "Computer", "Hey", "Hello" hoặc "Xin chào"
|
| 82 |
3. Trò chuyện tự nhiên bằng giọng nói
|
| 83 |
4. Mỗi phiên giới hạn 90 giây
|
| 84 |
|
|
@@ -86,30 +82,35 @@ def create_sambanova_voice_tab(sambanova_service):
|
|
| 86 |
✅ Voice-to-Text conversation
|
| 87 |
✅ Llama 3.2 3B model
|
| 88 |
✅ Stop-word activation
|
| 89 |
-
✅ Multi-language support
|
| 90 |
""")
|
| 91 |
|
| 92 |
-
#
|
| 93 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 94 |
|
| 95 |
-
#
|
| 96 |
-
with gr.Accordion("📝 Text Input
|
| 97 |
text_input = gr.Textbox(
|
| 98 |
label="Nhập tin nhắn",
|
| 99 |
-
placeholder="Nhập tin nhắn text
|
| 100 |
-
lines=
|
| 101 |
-
elem_id="sambanova_text_input"
|
| 102 |
)
|
| 103 |
-
text_send_btn = gr.Button("📤 Gửi tin nhắn",
|
| 104 |
|
| 105 |
# Controls
|
| 106 |
with gr.Row():
|
| 107 |
-
clear_btn = gr.Button("🗑️ Xóa hội thoại"
|
| 108 |
status_display = gr.Textbox(
|
| 109 |
label="Trạng thái",
|
| 110 |
-
value="✅ Sẵn sàng
|
| 111 |
-
interactive=False
|
| 112 |
-
elem_id="sambanova_status"
|
| 113 |
)
|
| 114 |
|
| 115 |
with gr.Column(scale=2):
|
|
@@ -120,28 +121,11 @@ def create_sambanova_voice_tab(sambanova_service):
|
|
| 120 |
- **Provider**: Sambanova AI
|
| 121 |
- **Context**: 128K tokens
|
| 122 |
- **Specialization**: Instruction following
|
| 123 |
-
- **
|
| 124 |
-
|
| 125 |
-
**Voice Features:**
|
| 126 |
-
- Real-time speech recognition
|
| 127 |
-
- Voice activity detection
|
| 128 |
-
- Stop-word activation
|
| 129 |
-
- Text-based responses
|
| 130 |
""")
|
| 131 |
-
|
| 132 |
-
# Conversation display
|
| 133 |
-
gr.Markdown("### 💬 Cuộc trò chuyện")
|
| 134 |
-
|
| 135 |
-
# Additional outputs handling
|
| 136 |
-
def handle_additional_outputs(new_chatbot, new_state):
|
| 137 |
-
"""Xử lý additional outputs từ streaming"""
|
| 138 |
-
return new_chatbot, new_state, "✅ Đang trò chuyện..."
|
| 139 |
-
|
| 140 |
-
# Event handlers
|
| 141 |
-
def clear_conversation():
|
| 142 |
-
"""Xóa hội thoại"""
|
| 143 |
-
return [], [], "🔄 Đã xóa hội thoại"
|
| 144 |
|
|
|
|
| 145 |
def send_text_message(text, current_chatbot, current_state):
|
| 146 |
"""Gửi tin nhắn text"""
|
| 147 |
if not text.strip():
|
|
@@ -172,17 +156,11 @@ def create_sambanova_voice_tab(sambanova_service):
|
|
| 172 |
error_msg = f"❌ Lỗi: {str(e)}"
|
| 173 |
return current_chatbot, current_state, error_msg
|
| 174 |
|
| 175 |
-
|
| 176 |
-
|
| 177 |
-
|
| 178 |
-
outputs=[chatbot, conversation_state, status_display]
|
| 179 |
-
)
|
| 180 |
-
|
| 181 |
-
clear_btn.click(
|
| 182 |
-
clear_conversation,
|
| 183 |
-
outputs=[chatbot, conversation_state, status_display]
|
| 184 |
-
)
|
| 185 |
|
|
|
|
| 186 |
text_send_btn.click(
|
| 187 |
send_text_message,
|
| 188 |
inputs=[text_input, chatbot, conversation_state],
|
|
@@ -191,6 +169,22 @@ def create_sambanova_voice_tab(sambanova_service):
|
|
| 191 |
lambda: "", # Clear text input
|
| 192 |
outputs=[text_input]
|
| 193 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 194 |
|
| 195 |
return sambanova_tab
|
| 196 |
def create_voice_coding_tab(voice_coding_service):
|
|
|
|
| 53 |
with gr.Tab("Stream Object Detection"):
|
| 54 |
create_streaming_object_detection()
|
| 55 |
def create_sambanova_voice_tab(sambanova_service):
|
| 56 |
+
"""Tạo tab Sambanova Voice AI - Simplified version"""
|
| 57 |
|
| 58 |
with gr.Blocks() as sambanova_tab:
|
| 59 |
gr.Markdown("## 🤖 Sambanova Voice AI - Llama 3.2 3B")
|
| 60 |
+
gr.Markdown("Trò chuyện voice-to-voice với model Llama 3.2 3B")
|
| 61 |
|
| 62 |
+
# State variables - KHÔNG dùng elem_id cho State
|
| 63 |
chatbot = gr.Chatbot(
|
| 64 |
type="messages",
|
| 65 |
value=[],
|
| 66 |
label="💬 Hội thoại Voice",
|
| 67 |
+
height=400
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 68 |
)
|
| 69 |
+
conversation_state = gr.State(value=[])
|
| 70 |
|
| 71 |
with gr.Row():
|
| 72 |
with gr.Column(scale=1):
|
|
|
|
| 74 |
gr.Markdown("""
|
| 75 |
**Cách sử dụng:**
|
| 76 |
1. Nhấn **Bắt đầu Voice Chat**
|
| 77 |
+
2. Nói "Computer", "Hey", "Hello" hoặc "Xin chào"
|
| 78 |
3. Trò chuyện tự nhiên bằng giọng nói
|
| 79 |
4. Mỗi phiên giới hạn 90 giây
|
| 80 |
|
|
|
|
| 82 |
✅ Voice-to-Text conversation
|
| 83 |
✅ Llama 3.2 3B model
|
| 84 |
✅ Stop-word activation
|
|
|
|
| 85 |
""")
|
| 86 |
|
| 87 |
+
# Tạo stream
|
| 88 |
+
try:
|
| 89 |
+
stream = sambanova_service.create_stream()
|
| 90 |
+
webrtc = stream.ui()
|
| 91 |
+
webrtc_available = True
|
| 92 |
+
except Exception as e:
|
| 93 |
+
print(f"⚠️ Cannot create WebRTC stream: {e}")
|
| 94 |
+
webrtc_available = False
|
| 95 |
+
gr.Markdown("### ❌ Voice Input Not Available")
|
| 96 |
+
gr.Markdown("Voice features are currently unavailable. Please use text input below.")
|
| 97 |
|
| 98 |
+
# Text input fallback
|
| 99 |
+
with gr.Accordion("📝 Text Input", open=True):
|
| 100 |
text_input = gr.Textbox(
|
| 101 |
label="Nhập tin nhắn",
|
| 102 |
+
placeholder="Nhập tin nhắn text...",
|
| 103 |
+
lines=3
|
|
|
|
| 104 |
)
|
| 105 |
+
text_send_btn = gr.Button("📤 Gửi tin nhắn", variant="primary")
|
| 106 |
|
| 107 |
# Controls
|
| 108 |
with gr.Row():
|
| 109 |
+
clear_btn = gr.Button("🗑️ Xóa hội thoại")
|
| 110 |
status_display = gr.Textbox(
|
| 111 |
label="Trạng thái",
|
| 112 |
+
value="✅ Sẵn sàng" if webrtc_available else "⚠️ Text mode only",
|
| 113 |
+
interactive=False
|
|
|
|
| 114 |
)
|
| 115 |
|
| 116 |
with gr.Column(scale=2):
|
|
|
|
| 121 |
- **Provider**: Sambanova AI
|
| 122 |
- **Context**: 128K tokens
|
| 123 |
- **Specialization**: Instruction following
|
| 124 |
+
- **Temperature**: 0.1
|
| 125 |
+
- **Top-P**: 0.1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 126 |
""")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 127 |
|
| 128 |
+
# Event handlers cho text input
|
| 129 |
def send_text_message(text, current_chatbot, current_state):
|
| 130 |
"""Gửi tin nhắn text"""
|
| 131 |
if not text.strip():
|
|
|
|
| 156 |
error_msg = f"❌ Lỗi: {str(e)}"
|
| 157 |
return current_chatbot, current_state, error_msg
|
| 158 |
|
| 159 |
+
def clear_conversation():
|
| 160 |
+
"""Xóa hội thoại"""
|
| 161 |
+
return [], [], "🔄 Đã xóa hội thoại"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 162 |
|
| 163 |
+
# Kết nối events cho text input
|
| 164 |
text_send_btn.click(
|
| 165 |
send_text_message,
|
| 166 |
inputs=[text_input, chatbot, conversation_state],
|
|
|
|
| 169 |
lambda: "", # Clear text input
|
| 170 |
outputs=[text_input]
|
| 171 |
)
|
| 172 |
+
|
| 173 |
+
clear_btn.click(
|
| 174 |
+
clear_conversation,
|
| 175 |
+
outputs=[chatbot, conversation_state, status_display]
|
| 176 |
+
)
|
| 177 |
+
|
| 178 |
+
# Kết nối events cho WebRTC nếu available
|
| 179 |
+
if webrtc_available:
|
| 180 |
+
def handle_webrtc_outputs(new_chatbot, new_state):
|
| 181 |
+
"""Xử lý outputs từ WebRTC"""
|
| 182 |
+
return new_chatbot, new_state, "✅ Đang trò chuyện..."
|
| 183 |
+
|
| 184 |
+
webrtc.on_additional_outputs(
|
| 185 |
+
handle_webrtc_outputs,
|
| 186 |
+
outputs=[chatbot, conversation_state, status_display]
|
| 187 |
+
)
|
| 188 |
|
| 189 |
return sambanova_tab
|
| 190 |
def create_voice_coding_tab(voice_coding_service):
|