Update app.py
Browse files
app.py
CHANGED
|
@@ -98,12 +98,11 @@ async def websocket_endpoint(ws: WebSocket):
|
|
| 98 |
del sessions[stream_sid]
|
| 99 |
print(f"Session cleaned up for stream {stream_sid}")
|
| 100 |
|
| 101 |
-
|
| 102 |
async def process_user_audio(ws: WebSocket, stream_sid: str, audio_chunk: np.ndarray):
|
| 103 |
"""The main logic loop: STT -> LLM -> (Tool/TTS)"""
|
| 104 |
print(f"[{stream_sid}] Processing audio chunk of size {len(audio_chunk)}...")
|
| 105 |
|
| 106 |
-
# 1. Speech-to-Text
|
| 107 |
user_text = await transcribe_audio_chunk(audio_chunk)
|
| 108 |
if not user_text:
|
| 109 |
print(f"[{stream_sid}] No text transcribed.")
|
|
@@ -122,7 +121,6 @@ async def process_user_audio(ws: WebSocket, stream_sid: str, audio_chunk: np.nda
|
|
| 122 |
if chunk is None: break
|
| 123 |
yield chunk
|
| 124 |
|
| 125 |
-
# 2. Start LLM and TTS tasks concurrently
|
| 126 |
llm_task = asyncio.create_task(get_llm_response(sessions[stream_sid]["messages"], llm_chunk_handler))
|
| 127 |
tts_task = asyncio.create_task(stream_and_send_audio(ws, stream_sid, tts_text_iterator()))
|
| 128 |
|
|
@@ -133,13 +131,10 @@ async def process_user_audio(ws: WebSocket, stream_sid: str, audio_chunk: np.nda
|
|
| 133 |
if assistant_message and assistant_message.get("content"):
|
| 134 |
sessions[stream_sid]["messages"].append(assistant_message)
|
| 135 |
|
| 136 |
-
# 3. Handle Tool Calls if any
|
| 137 |
if tool_calls:
|
| 138 |
-
# Before executing, add the assistant's tool request to history
|
| 139 |
sessions[stream_sid]["messages"].append(assistant_message)
|
| 140 |
|
| 141 |
for tool_call_data in tool_calls:
|
| 142 |
-
# Recreate a simple object that mimics the structure tool_handler expects
|
| 143 |
tool_call = type('ToolCall', (), {
|
| 144 |
'id': tool_call_data.get('id'),
|
| 145 |
'function': type('Function', (), tool_call_data.get('function'))
|
|
@@ -149,7 +144,6 @@ async def process_user_audio(ws: WebSocket, stream_sid: str, audio_chunk: np.nda
|
|
| 149 |
tool_result_message = execute_tool_call(tool_call)
|
| 150 |
sessions[stream_sid]["messages"].append(tool_result_message)
|
| 151 |
|
| 152 |
-
# 4. Get a final response from the LLM after executing the tool
|
| 153 |
final_tts_queue = asyncio.Queue()
|
| 154 |
async def final_llm_chunk_handler(chunk): await final_tts_queue.put(chunk)
|
| 155 |
async def final_tts_iterator():
|
|
@@ -186,5 +180,7 @@ async def stream_and_send_audio(ws: WebSocket, stream_sid: str, text_iterator):
|
|
| 186 |
|
| 187 |
if __name__ == "__main__":
|
| 188 |
import uvicorn
|
| 189 |
-
|
| 190 |
-
|
|
|
|
|
|
|
|
|
| 98 |
del sessions[stream_sid]
|
| 99 |
print(f"Session cleaned up for stream {stream_sid}")
|
| 100 |
|
| 101 |
+
|
| 102 |
async def process_user_audio(ws: WebSocket, stream_sid: str, audio_chunk: np.ndarray):
|
| 103 |
"""The main logic loop: STT -> LLM -> (Tool/TTS)"""
|
| 104 |
print(f"[{stream_sid}] Processing audio chunk of size {len(audio_chunk)}...")
|
| 105 |
|
|
|
|
| 106 |
user_text = await transcribe_audio_chunk(audio_chunk)
|
| 107 |
if not user_text:
|
| 108 |
print(f"[{stream_sid}] No text transcribed.")
|
|
|
|
| 121 |
if chunk is None: break
|
| 122 |
yield chunk
|
| 123 |
|
|
|
|
| 124 |
llm_task = asyncio.create_task(get_llm_response(sessions[stream_sid]["messages"], llm_chunk_handler))
|
| 125 |
tts_task = asyncio.create_task(stream_and_send_audio(ws, stream_sid, tts_text_iterator()))
|
| 126 |
|
|
|
|
| 131 |
if assistant_message and assistant_message.get("content"):
|
| 132 |
sessions[stream_sid]["messages"].append(assistant_message)
|
| 133 |
|
|
|
|
| 134 |
if tool_calls:
|
|
|
|
| 135 |
sessions[stream_sid]["messages"].append(assistant_message)
|
| 136 |
|
| 137 |
for tool_call_data in tool_calls:
|
|
|
|
| 138 |
tool_call = type('ToolCall', (), {
|
| 139 |
'id': tool_call_data.get('id'),
|
| 140 |
'function': type('Function', (), tool_call_data.get('function'))
|
|
|
|
| 144 |
tool_result_message = execute_tool_call(tool_call)
|
| 145 |
sessions[stream_sid]["messages"].append(tool_result_message)
|
| 146 |
|
|
|
|
| 147 |
final_tts_queue = asyncio.Queue()
|
| 148 |
async def final_llm_chunk_handler(chunk): await final_tts_queue.put(chunk)
|
| 149 |
async def final_tts_iterator():
|
|
|
|
| 180 |
|
| 181 |
if __name__ == "__main__":
|
| 182 |
import uvicorn
|
| 183 |
+
# Hugging Face Spaces expects the app to run on port 7860
|
| 184 |
+
port = int(os.environ.get("PORT", 7860))
|
| 185 |
+
print(f"Starting RentBot server on port {port}...")
|
| 186 |
+
uvicorn.run(app, host="0.0.0.0", port=port)
|