Update app.py

app.py

@@ -1,13 +1,13 @@
 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
-import pyaudio
-import wave
-import os
+from pydub import AudioSegment
 
-pre_prompt_text = "Eres una IA conductual, tus respuestas deberán ser breves, estóicas y humanistas."
+pre_prompt_text = "You are a behavioral AI, your answers should be brief, stoic and humanistic."
 
 if "history" not in st.session_state:
     st.session_state.history = []
@@ -15,7 +15,6 @@ if "history" not in st.session_state:
 if "pre_prompt_sent" not in st.session_state:
     st.session_state.pre_prompt_sent = False
 
-# Simplified error messages
 def recognize_speech(audio_data, show_messages=True):
     recognizer = sr.Recognizer()
     audio_recording = sr.AudioFile(audio_data)
@@ -26,21 +25,28 @@ def recognize_speech(audio_data, show_messages=True):
     try:
         audio_text = recognizer.recognize_google(audio, language="es-ES")
         if show_messages:
-            st.subheader("Texto Reconocido:")
+            st.subheader("Recognized text:")
             st.write(audio_text)
-            st.success("Reconocimiento de voz completado.")
-    except (sr.UnknownValueError, sr.RequestError) as e:
-        st.error("Error en el reconocimiento de voz. ¿Intentaste grabar algo?")
+            st.success("Completed.")
+    except sr.UnknownValueError:
+        st.warning("The audio could not be recognized. Did you try to record something?")
+        audio_text = ""
+    except sr.RequestError:
+        st.error("Talk to me to get started!")
         audio_text = ""
 
     return audio_text
 
-# Simplified prompt formatting
 def format_prompt(message, history):
-    prompt = "<s>[INST] " + pre_prompt_text + " [/INST]"
+    prompt = "<s>"
+
+    if not st.session_state.pre_prompt_sent:
+        prompt += f"[INST] {pre_prompt_text} [/INST]"
+        st.session_state.pre_prompt_sent = True
 
     for user_prompt, bot_response in history:
-        prompt += f"[INST] {user_prompt} [/INST] {bot_response}</s> "
+        prompt += f"[INST] {user_prompt} [/INST]"
+        prompt += f" {bot_response}</s> "
 
     prompt += f"[INST] {message} [/INST]"
     return prompt
@@ -68,15 +74,23 @@ def generate(audio_text, history, temperature=None, max_new_tokens=512, top_p=0.
         response += response_token.token.text
     
     response = ' '.join(response.split()).replace('</s>', '')
-    return response
+    audio_file = text_to_speech(response, speed=1.3)
+    return response, audio_file
+
+def text_to_speech(text, speed=1.3):
+    tts = gTTS(text=text, lang='es')
+    audio_fp = io.BytesIO()
+    tts.write_to_fp(audio_fp)
+    audio_fp.seek(0)
+    audio = AudioSegment.from_file(audio_fp, format="mp3")
+    modified_speed_audio = audio.speedup(playback_speed=speed)
+    modified_audio_fp = io.BytesIO()
+    modified_speed_audio.export(modified_audio_fp, format="mp3")
+    modified_audio_fp.seek(0)
+    return modified_audio_fp
 
-# Simplified audio playback using os.system and say
-def play_audio(text):
-    os.system(f"say '{text}'")
-
-# Simplified audio playback
 def main():
-    audio_data = audiorecorder("Presiona para hablar", "Deteniendo la grabación...")
+    audio_data = audiorecorder("Push to Play", "Stop Recording...")
 
     if not audio_data.empty():
         st.audio(audio_data.export().read(), format="audio/wav")
@@ -84,8 +98,12 @@ def main():
         audio_text = recognize_speech("audio.wav")
 
         if audio_text:
-            response = generate(audio_text, history=st.session_state.history)  
-            play_audio(response)
+            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)
 
 if __name__ == "__main__":
-    main()
+    main()