Update app.py

app.py

@@ -1,6 +1,5 @@
 import io
 import base64
-import threading
 import numpy as np
 from gtts import gTTS
 import streamlit as st
@@ -18,7 +17,6 @@ if "history" not in st.session_state:
 if "pre_prompt_sent" not in st.session_state:
     st.session_state.pre_prompt_sent = False
 
-gatherUsageStats = "false"
 pre_prompt_text = "eres una IA conductual, tus respuestas serán breves."
 
 def recognize_speech(audio_data, show_messages=True):
@@ -40,34 +38,6 @@ def recognize_speech(audio_data, show_messages=True):
 
     return audio_text
 
-def start_stream():
-    recognizer = sr.Recognizer()
-    with sr.AudioFile(temp_audio_file_path) as source:
-        audio = recognizer.record(source)
-
-    audio_data = np.frombuffer(audio.frame_data, dtype=np.int16)
-    silent_chunks = detect_silence(audio_data)
-
-    for silent_chunk in silent_chunks:
-        audio_text = recognize_speech(silent_chunk)
-        st.subheader("Texto Reconocido:")
-        st.write(audio_text)
-
-        # Actualiza el historial con el texto reconocido
-        st.session_state.history.append((audio_text, ""))
-
-        # Genera la respuesta de texto a voz usando el texto reconocido como prompt
-        response, audio_file = generate(audio_text, st.session_state.history)
-
-        # Muestra la respuesta y reproduce el audio
-        display_recognition_result(audio_text, response, audio_file)
-
-def detect_silence(audio_data, silence_threshold=5000, silence_duration=5000):
-    is_silence = lambda x: max(x) < silence_threshold
-    chunks = [audio_data[i:i+silence_duration] for i in range(0, len(audio_data), silence_duration)]
-    silent_chunks = [chunk for chunk in chunks if is_silence(chunk)]
-    return silent_chunks
-
 def format_prompt(message, history):
     prompt = "<s>"
 
@@ -127,39 +97,40 @@ def display_recognition_result(audio_text, output, audio_file):
             f"""<audio autoplay="autoplay" controls="controls" src="data:audio/mp3;base64,{base64.b64encode(audio_file.read()).decode()}" type="audio/mp3" id="audio_player"></audio>""",
             unsafe_allow_html=True)
 
-def voice_activity_detection(audio_data):
-    return vad.is_speech(audio_data, sample_rate)
-
-def start_stream():
-    recognizer = sr.Recognizer()
-    with sr.AudioFile(temp_audio_file_path) as source:
-        audio = recognizer.record(source)
-
-    audio_data = np.frombuffer(audio.frame_data, dtype=np.int16)
-    silent_chunks = detect_silence(audio_data)
-
-    for silent_chunk in silent_chunks:
-        audio_text = recognize_speech(silent_chunk)
-        st.subheader("Texto Reconocido:")
-        st.write(audio_text)
-
-audio = mic_recorder(start_prompt="▶️", stop_prompt="🛑", key='recorder')
-
-if audio:
-    st.audio(audio['bytes'])
-
-    audio_bytes = audio["bytes"]
-    sample_width = audio["sample_width"]
-    sample_rate = audio["sample_rate"]
-
-    with wave.open(temp_audio_file_path, 'wb') as wave_file:
-        wave_file.setnchannels(1)
-        wave_file.setsampwidth(sample_width)
-        wave_file.setframerate(sample_rate)
-        wave_file.writeframes(audio_bytes)
+def detect_silence(audio_data, sample_rate, silence_threshold=5000, silence_duration=5000):
+    is_silence = lambda x: max(x) < silence_threshold
+    chunks = [audio_data[i:i+silence_duration] for i in range(0, len(audio_data), silence_duration)]
+    silent_chunks = [chunk for chunk in chunks if is_silence(chunk)]
+    return silent_chunks
 
-vad = webrtcvad.Vad(1)
-channels = [1]
-mapping = [c - 1 for c in channels]
-interval_size = 10 
-downsample = 1
+def main():
+    if not st.session_state.pre_prompt_sent:
+        st.session_state.pre_prompt_sent = True
+
+    audio = mic_recorder(start_prompt="▶️", stop_prompt="🛑", key='recorder')
+
+    if audio:
+        st.audio(audio['bytes'], format="audio/wav")
+        audio_bytes = audio["bytes"]
+        sample_width = audio["sample_width"]
+        sample_rate = audio["sample_rate"]
+
+        with wave.open(temp_audio_file_path, 'wb') as wave_file:
+            wave_file.setnchannels(1)
+            wave_file.setsampwidth(sample_width)
+            wave_file.setframerate(sample_rate)
+            wave_file.writeframes(audio_bytes)
+
+        audio_data = np.frombuffer(audio_bytes, dtype=np.int16)
+        silent_chunks = detect_silence(audio_data, sample_rate)
+
+        for silent_chunk in silent_chunks:
+            audio_text = recognize_speech(silent_chunk)
+            st.subheader("Texto Reconocido:")
+            st.write(audio_text)
+            st.session_state.history.append((audio_text, ""))
+            response, audio_file = generate(audio_text, st.session_state.history)
+            display_recognition_result(audio_text, response, audio_file)
+
+if __name__ == "__main__":
+    main()