Update app.py

app.py

@@ -1,29 +1,21 @@
-import pyaudio
 import streamlit as st
 import base64
 import io
 from huggingface_hub import InferenceClient
 from gtts import gTTS
+from audiorecorder import audiorecorder
 import speech_recognition as sr
 from pydub import AudioSegment
-import webrtcvad
 
-def perform_vad(audio_data):
-    vad = webrtcvad.Vad()
-    vad.set_mode(1)
-    
-    samples = audio_data.frame_data
-    is_speech = vad.is_speech(samples, sample_rate=audio_data.sample_rate)
-    
-    return is_speech
+if "history" not in st.session_state:
+    st.session_state.history = []
 
-def recognize_speech(device_index, show_messages=True):
+def recognize_speech(audio_data, show_messages=True):
     recognizer = sr.Recognizer()
-    audio_recording = sr.Microphone(device_index=device_index, sample_rate=16000, chunk_size=1024)
+    audio_recording = sr.AudioFile(audio_data)
 
     with audio_recording as source:
-        recognizer.adjust_for_ambient_noise(source, gain=None)
-        audio = recognizer.listen(source, timeout=5, gain=None)
+        audio = recognizer.record(source)
 
     try:
         audio_text = recognizer.recognize_google(audio, language="es-ES")
@@ -44,9 +36,10 @@ def format_prompt(message, history):
     prompt = "<s>"
 
     for user_prompt, bot_response in history:
-        prompt += f"[/s] {user_prompt} [/s] {bot_response}</s> "
+        prompt += f"[INST] {user_prompt} [/INST]"
+        prompt += f" {bot_response}</s> "
 
-    prompt += f"[/s] {message} [/s]"
+    prompt += f"[INST] {message} [/INST]"
     return prompt
 
 def generate(audio_text, history, temperature=None, max_new_tokens=512, top_p=0.95, repetition_penalty=1.0):
@@ -91,39 +84,26 @@ def text_to_speech(text, speed=1.3):
 
 def main():
     st.title("Chatbot de Voz a Voz")
-    available_microphones = pyaudio.PyAudio().get_device_count()
-
-    if available_microphones > 0:
-        st.info(f"Número de micrófonos disponibles: {available_microphones}")
-        selected_microphone = st.selectbox("Selecciona un micrófono", list(range(available_microphones)))
-        st.info(f"Número de micrófono seleccionado: {selected_microphone}")
-        recognizer = sr.Recognizer()
-        audio_recording = sr.Microphone(device_index=selected_microphone, sample_rate=16000, chunk_size=1024)
-
-        st.info("Habla para grabar")
-        with audio_recording as source:
-            recognizer.adjust_for_ambient_noise(source, duration=1)
-            audio_data = recognizer.listen(source, timeout=5)
-
-        st.success("Deteniendo la grabación...")
-
-        if audio_data:
-            st.audio(audio_data.frame_data, format="audio/wav")
-            
-            if perform_vad(audio_data):
-                audio_text = recognize_speech(device_index=selected_microphone)
-                if audio_text:
-                    output, audio_file = generate(audio_text, history=st.session_state.history)
-
-                    if audio_file is not None:
-                        st.markdown(
-                            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
-                        )
-    else:
-        st.warning("No se detectaron micrófonos disponibles. Asegúrate de que tengas un micrófono conectado.")
+    audio_data = sr.Microphone(device_index=-1)
+
+    if not audio_data.empty():
+        st.audio(audio_data.export().read(), format="audio/wav")
+        audio_data.export("audio.wav", format="wav")
+        audio_text = recognize_speech("audio.wav")
+
+        if audio_text:
+            output, audio_file = generate(audio_text, history=st.session_state.history)  
+
+            if audio_text:  
+                st.session_state.history.append((audio_text, output))  
+
+            if audio_file is not None:
+                st.markdown(
+                    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
+                )
 
 if __name__ == "__main__":
-    main()
+    main()