Update app.py

app.py

@@ -2,8 +2,10 @@ import streamlit as st
 from huggingface_hub import InferenceClient
 import base64
 from pydub import AudioSegment
-from pydub.playback import play
-import pyttsx3
+from io import BytesIO
+from gtts import gTTS
+import speech_recognition as sr
+from datetime import datetime
 
 client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
 pre_prompt = ""
@@ -25,11 +27,36 @@ def format_prompt(message, history):
     return prompt
 
 def text_to_speech(text):
-    engine = pyttsx3.init()
-    engine.save_to_file(text, "output_pyttsx3.mp3")
-    engine.runAndWait()
-    return "output_pyttsx3.mp3"
+    tts = gTTS(text=text, lang='es')
+    audio_stream = BytesIO()
+    tts.save(audio_stream)
+    audio_stream.seek(0)
+    return audio_stream.read()
+
+def take_user_input():
     
+    r = sr.Recognizer()
+    with sr.Microphone() as source:
+        print('Escuchando....')
+        r.pause_threshold = 1
+        audio = r.listen(source)
+
+    try:
+        print('Reconociendo...')
+        query = r.recognize_google(audio, language='es-ES')
+        if 'salir' in query or 'detener' in query:
+            speak("Hasta luego.")
+            exit()
+    except Exception:
+        query = 'None'
+        else:
+            speak('de nuevo...')
+            query = 'None'
+    except Exception:
+        speak('de nuevo...')
+        query = 'None'
+    return query
+
 def generate(user_input, history, temperature=None, max_new_tokens=512, top_p=0.95, repetition_penalty=1.0):
     global pre_prompt_sent
     temperature = float(temperature) if temperature is not None else 0.9
@@ -47,30 +74,38 @@ def generate(user_input, history, temperature=None, max_new_tokens=512, top_p=0.
     )
 
     formatted_prompt = format_prompt(user_input, history)
-    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=True)
-    response = ""
-
-    for response_token in stream:
-        response += response_token.token.text
-    
-    response = ' '.join(response.split()).replace('</s>', '')
+    try:
+        stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=True)
+        response = ""
+        for response_token in stream:
+            response += response_token.token.text
+        response = ' '.join(response.split()).replace('</s>', '')
+        audio_bytes = text_to_speech(response)
+        return response, audio_bytes
+    except Exception as e:
+        return str(e), None
 
-    audio_file_path = text_to_speech(response) 
+def speak(text):
+    audio_file_path = text_to_speech(text)
     audio_file = open(audio_file_path, 'rb')
     audio_bytes = audio_file.read()
-
-    return response, audio_bytes
+    st.audio(audio_bytes, format="audio/mp3", start_time=0, key="audio_player")
 
 if "history" not in st.session_state:
     st.session_state.history = []
 
-user_input = st.text_input(label="", value="")
+user_input = take_user_input()
 output, audio_bytes = generate(user_input, history=st.session_state.history)
+
+if user_input != 'None':
+    st.session_state.history.append((user_input, output))
+
 st.text_area("Respuesta", value=output, height=400, key="output_text", disabled=True)
 
-st.markdown(
-    f"""
-    <audio autoplay="autoplay" controls="controls" src="data:audio/mp3;base64,{base64.b64encode(audio_bytes).decode()}" type="audio/mp3" id="audio_player"></audio>
-    """,
-    unsafe_allow_html=True
-)
+if audio_bytes is not None:
+    st.markdown(
+        f"""
+        <audio autoplay="autoplay" controls="controls" src="data:audio/mp3;base64,{base64.b64encode(audio_bytes).decode()}" type="audio/mp3" id="audio_player"></audio>
+        """,
+        unsafe_allow_html=True
+    )