Update app.py

app.py

@@ -3,19 +3,37 @@ import base64
 import io
 from huggingface_hub import InferenceClient
 from gtts import gTTS
+from audiorecorder import audiorecorder
+import speech_recognition as sr
 
-client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
-pre_prompt = "Hola, te llamarás Chaman 3.0, una IA conductual. Tus principios son el transhumanismo ecológico."
-pre_prompt_sent = False
+def recognize_speech(audio_data):
+    st.info("Reconociendo audio...")
+
+    recognizer = sr.Recognizer()
+    audio_recording = sr.AudioFile(audio_data)
+
+    try:
+        with audio_recording as source:
+            audio = recognizer.record(source)
+
+        audio_text = recognizer.recognize_google(audio, language="es-ES")
+        st.subheader("Texto Reconocido:")
+        st.write(audio_text)
+
+    except sr.UnknownValueError:
+        st.warning("No se pudo reconocer el audio. ¿Intentaste grabar algo?")
+        audio_text = ""
+
+    except sr.RequestError as e:
+        st.error(f"Error en el reconocimiento de voz: {e}")
+        audio_text = ""
+
+    st.success("Reconocimiento de voz completado.")
+    return audio_text
 
 def format_prompt(message, history):
-    global pre_prompt_sent
     prompt = "<s>"
 
-    if not pre_prompt_sent and all(f"[INST] {pre_prompt} [/INST]" not in user_prompt for user_prompt, _ in history):
-        prompt += f"[INST] {pre_prompt} [/INST]"
-        pre_prompt_sent = True
-
     for user_prompt, bot_response in history:
         prompt += f"[INST] {user_prompt} [/INST]"
         prompt += f" {bot_response}</s> "
@@ -30,8 +48,9 @@ def text_to_speech(text, speed=1.3):
     audio_fp.seek(0)
     return audio_fp
 
-def generate(user_input, history, temperature=None, max_new_tokens=512, top_p=0.95, repetition_penalty=1.0):
-    global pre_prompt_sent
+def generate(audio_text, history, temperature=None, max_new_tokens=512, top_p=0.95, repetition_penalty=1.0):
+    client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
+
     temperature = float(temperature) if temperature is not None else 0.9
     if temperature < 1e-2:
         temperature = 1e-2
@@ -46,7 +65,7 @@ def generate(user_input, history, temperature=None, max_new_tokens=512, top_p=0.
         seed=42,
     )
 
-    formatted_prompt = format_prompt(user_input, history)
+    formatted_prompt = format_prompt(audio_text, history)
     stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=True)
     response = ""
 
@@ -57,23 +76,34 @@ def generate(user_input, history, temperature=None, max_new_tokens=512, top_p=0.
     audio_file = text_to_speech(response, speed=1.3)
     return response, audio_file
 
-if "history" not in st.session_state:
-    st.session_state.history = []
+def main():
+    st.title("Grabación de Audio y Reconocimiento de Voz")
 
-with st.container():
-    user_input = st.text_input(label="Usuario", value="Saludos")
-    output, audio_file = generate(user_input, history=st.session_state.history)
-    st.text_area("Respuesta", height=400, value=output, key="output_text", disabled=True)
+    if "history" not in st.session_state:
+        st.session_state.history = []
 
-    if user_input:
-        st.session_state.history.append((user_input, output))
+    audio_data = audiorecorder("Habla para grabar", "Deteniendo la grabación...")
 
-    st.write("Presiona el botón y comienza a hablar...")
+    if not audio_data.empty():
+        st.audio(audio_data.export().read(), format="audio/wav")
+        audio_data.export("audio.wav", format="wav")
+        st.write(f"Frame rate: {audio_data.frame_rate}, Frame width: {audio_data.frame_width}, Duration: {audio_data.duration_seconds} seconds")
 
-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
-    )
+        audio_text = recognize_speech("audio.wav")
+        
+        output, audio_file = generate(audio_text, history=st.session_state.history)  
+        st.text_area("Respuesta", height=100, value=output, key="output_text", disabled=True)
+            
+        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()