Update app.py

app.py

@@ -7,13 +7,19 @@ import speech_recognition as sr
 from pydub import AudioSegment
 from audiorecorder import audiorecorder
 
+pre_prompt_text = "eres una IA conductual, tus respuestas serán breves."
+
 if "history" not in st.session_state:
     st.session_state.history = []
 
+if "pre_prompt_sent" not in st.session_state:
+    st.session_state.pre_prompt_sent = False
+
 def recognize_speech(audio_data, show_messages=True):
     recognizer = sr.Recognizer()
 
     try:
+        audio_data.seek(0)
         audio = gTTS(audio_data.read(), lang='es')
         audio_recording = sr.AudioFile(io.BytesIO(audio_data.read()))
 
@@ -35,6 +41,7 @@ def recognize_speech(audio_data, show_messages=True):
     return audio_text
 
 def format_prompt(message, history):
+    preprompt = "Este es el inicio de la conversación. "
     prompt = "<s>"
 
     for user_prompt, bot_response in history:
@@ -42,7 +49,7 @@ def format_prompt(message, history):
         prompt += f" {bot_response}</s> "
 
     prompt += f"[INST] {message} [/INST]"
-    return prompt
+    return preprompt + prompt
 
 def generate(audio_text, history, temperature=None, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0):
     client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
@@ -76,10 +83,17 @@ 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
 
 def play_audio(audio_file):
@@ -89,7 +103,14 @@ def play_audio(audio_file):
 def main():
     st.title("Chatbot de Voz a Voz")
     audio_text = ""
-    audio_data = audiorecorder("Hablar ▶️", "Detener 🛑")
+
+    if not st.session_state.pre_prompt_sent:
+        st.session_state.pre_prompt_sent = True
+        st.session_state.history.append((pre_prompt_text, ""))
+        pre_prompt_audio_file = text_to_speech(pre_prompt_text)
+        play_audio(pre_prompt_audio_file)
+
+    audio_data = audiorecorder.record("Hablar ▶️", "Detener 🛑")
 
     if not audio_data.empty():
         st.audio(audio_data.export().read(), format="audio/mp3")