added idle timeout of 10 secs

main.py

@@ -88,9 +88,13 @@ def main():
                 # record audio
                 #display("recording...",color=ERROR)
                 led(LED_RED,"on")
-                record_audio(filename_input)
+                timeout = record_audio(filename_input)
                 #display("recording stopped.",color=ERROR)
                 led(LED_RED,"off")
+
+                if timeout:
+                    end_conversation(persona)
+                    main()
 
                 # transcribe audio to text with whisper-1 model
                 user_text = transcribe_audio(filename_input)

utils/gpio.py

@@ -1,12 +1,12 @@
 # simple LED on/off functionality
-import os
-
-if os.name == "posix":
-  import RPi.GPIO as GPIO
+import platform
 
+if platform.system() == 'Linux':
+    import RPi.GPIO as GPIO
+
 
 def led(LED, state="on"):
-  if os.name == "posix":
+  if platform.system() == 'Linux':
     GPIO.setmode(GPIO.BCM)
     GPIO.setup(LED, GPIO.OUT)
     if state == "on":

utils/recording.py

@@ -2,6 +2,7 @@
 import soundfile as sf
 import numpy as np
 import sounddevice # needed to suppres linux error messsages in terminal
+import time
 
 def record_audio(filename):
     CHUNK = 1024
@@ -10,7 +11,9 @@ def record_audio(filename):
     RATE = 22050  # samples per second
     THRESHOLD = 1000  # Adjust this value to set the sensitivity of the voice detection
     SILENCE_LIMIT = 1.1  # Number of seconds of silence before stopping the recording
-
+    TIMEOUT_AFTER_SECONDS = 10  # Maximum recording time in seconds
+    IS_TIMEOUT = False
+
     p = pyaudio.PyAudio()
     stream = p.open(format=FORMAT,
                     channels=CHANNELS,
@@ -19,14 +22,20 @@ def record_audio(filename):
                     frames_per_buffer=CHUNK)
 
     # Wait for the user to start speaking
+    start_time = time.time()
     while True:
         data = np.frombuffer(stream.read(CHUNK), dtype=np.int16)
-        if np.abs(data).mean() > THRESHOLD: #check mean amplitude of the chunk
+        if np.abs(data).mean() > THRESHOLD:  # check mean amplitude of the chunk
+            break
+        if (time.time() - start_time) >= TIMEOUT_AFTER_SECONDS:
+            IS_TIMEOUT = True
             break
 
-    # Record until the user stops speaking
+    # Record for the specified duration or until the user stops speaking
     frames = []
     silent_frames = 0
+
+    start_time = time.time()
     while True:
         data = np.frombuffer(stream.read(CHUNK), dtype=np.int16)
         frames.append(data)
@@ -43,4 +52,9 @@ def record_audio(filename):
 
     # Convert the frames to a single audio file
     frames = np.concatenate(frames, axis=0)
-    sf.write(filename, frames, RATE)
+    sf.write(filename, frames, RATE)
+
+    if IS_TIMEOUT:
+        return True
+    else:
+        return False