commit

app.py

@@ -10,7 +10,6 @@
 from backend.speech_proccessing import process_audio_stream
 from backend.openai_models import transcribe_audio, generate_response, generate_image_response, ChatHistory
 
-
 import time, wave
 import asyncio
 import json
@@ -25,7 +24,6 @@
 
 load_dotenv()
 
-
 app = FastAPI()
 
 users_directory = {}    # maintain users database or their chat history in their where each key represents the user_id
@@ -40,19 +38,20 @@ async def get(request: Request):
 
 @app.websocket('/ws/{user_id}')     # will be responsible for handling real time stream of audio chunks and all AI generated responses will be sent to the streamer client 
 async def chat(websocket: WebSocket, user_id: str):
-    i = 0
+
     if user_id not in users_directory: 
         users_directory[user_id] = ChatHistory()   # creates an instance of the ChatHistory class and each user will have their own instance of ChatHistory
 
     chat_history = users_directory[user_id]
-     
+
     await websocket.accept()
 
     audio_queue = asyncio.Queue()
     response_queue = asyncio.Queue()
 
     process_task = asyncio.create_task(process_audio_stream(audio_queue, response_queue))   # It will create asynchrounous task to handle audio_queue in the background, detect speeches in the audio_queue using pre trained Model and add it to response_queue
-
+
+    i = 0
     while True:
 
         try:
@@ -69,6 +68,7 @@ async def chat(websocket: WebSocket, user_id: str):
             # print(audio_queue.qsize())
             if not response_queue.empty():
               logger.info('Speech detected')
+              await asyncio.sleep(0.1)
               await generate_ai_response(response_queue, websocket, user_id, chat_history)   #  for generating ai responses and send it back to the client
 
 
@@ -100,7 +100,7 @@ async def handle_audio_new(websocket: WebSocket, audio_queue):
 
                 if not audio_data:
                     break
-                await audio_queue.put(audio_data)      
+                await audio_queue.put(audio_data)    
 
         return True
     except Exception as e:

backend/speech_proccessing.py

@@ -39,13 +39,10 @@ async def process_audio_stream(audio_queue, response_queue):
     audio_buffer = np.zeros(TARGET_LENGTH, dtype=np.float32)
     audio_chunks = []
     audio_data = b''
-        # Open the audio stream
-
-        # print("Listening... Press Ctrl+C to stop.")
 
     speak = 0
     silence = 0
-        # Continuously read from the stream and append to audio_data
+
 
     while True:
           try: