🚀 Auto-deploy: BotVPS atualizado em 24/03/2026 11:50:13

bot_logic.py

@@ -8,10 +8,9 @@ from orchestrator import (
     handle_message_async, orchestrate_async, format_confirmation_message, 
     format_completion_message, execute_step_async
 )
-from ai_agent import query_agent
+from ai_agent import query_agent_async
-import speech_recognition as sr
+from audio_handler import transcribe_audio, text_to_speech
-from pydub import AudioSegment
+from config import get_config
-from gtts import gTTS
 
 # Configuração de logging
 logging.basicConfig(
@@ -24,14 +23,12 @@ TOKEN = os.getenv("TELEGRAM_BOT_TOKEN")
 ALLOWED_CHAT_ID = os.getenv("TELEGRAM_CHAT_ID")
 
 def synthesize_audio(text: str) -> str:
+    """Wrapper para a síntese de voz centralizada."""
     try:
-        texto_limpo = re.sub(r'[*`#]', '', text)
+        filename = text_to_speech(text)
-        filepath = "/tmp/reply_audio.mp3"
+        return os.path.join("/tmp", filename)
-        tts = gTTS(text=texto_limpo[:500], lang='pt-br', slow=False)
-        tts.save(filepath)
-        return filepath
     except Exception as e:
-        print(f"TTS Error: {e}")
+        logger.error(f"TTS Error: {e}")
     return ""
 
 async def auth_check(update: Update) -> bool:
@@ -42,10 +39,41 @@ async def auth_check(update: Update) -> bool:
 
 async def handle_text(update: Update, context: ContextTypes.DEFAULT_TYPE):
     if not await auth_check(update): return
-    
-    user_msg = update.message.text
-    chat_id = update.message.chat_id
     await update.message.reply_chat_action(action="typing")
+    await process_logic(update, context, update.message.text)
+
+async def handle_voice(update: Update, context: ContextTypes.DEFAULT_TYPE):
+    if not await auth_check(update): return
+    
+    await update.message.reply_chat_action(action="record_voice")
+    
+    # Baixa o arquivo de voz do Telegram
+    voice_file = await context.bot.get_file(update.message.voice.file_id)
+    ogg_path = f"/tmp/{update.message.voice.file_id}.ogg"
+    await voice_file.download_to_drive(ogg_path)
+    
+    try:
+        # Transcreve (STT)
+        text = transcribe_audio(ogg_path)
+        if not text:
+            await update.message.reply_text("Não consegui entender o áudio.")
+            return
+            
+        await update.message.reply_text(f"🎤 Entendi: \"{text}\"")
+        
+        # Processa como se fosse texto (reutiliza handle_text logic)
+        # Gambiarra rápida: cria um objeto Update fake ou chama a lógica diretamente
+        # Melhor: extrair a lógica de handle_text para uma função pura
+        await process_logic(update, context, text, is_voice=True)
+            
+    except Exception as e:
+        logger.error(f"Erro Voz: {e}")
+        await update.message.reply_text(f"Erro ao processar voz: {e}")
+    finally:
+        if os.path.exists(ogg_path): os.remove(ogg_path)
+
+async def process_logic(update: Update, context: ContextTypes.DEFAULT_TYPE, user_msg: str, is_voice: bool = False):
+    chat_id = update.message.chat_id
     
     # 1. COMANDOS DIRETOS
     if user_msg.startswith('/') and user_msg.split()[0] in ['/status', '/tools', '/sync']:
@@ -78,19 +106,24 @@ async def handle_text(update: Update, context: ContextTypes.DEFAULT_TYPE):
             reply = format_confirmation_message(result)
         else:
             reply = format_completion_message(result)
-        await update.message.reply_text(reply)
     else:
         # Fallback AI Agent
-        from config import get_config
-        from ai_agent import query_agent_async
         cfg = get_config()
+        # No Telegram, ainda não estamos mantendo histórico complexo no bot_data (pode ser futuro)
         reply = await query_agent_async(user_msg, override_provider=cfg.get("active_provider"))
-        await update.message.reply_text(reply)
     
-async def handle_voice(update: Update, context: ContextTypes.DEFAULT_TYPE):
+    # Envia resposta (Texto)
-    if not await auth_check(update): return
+    # Remove o bloco <REFINED> se houver, pois no Telegram polui
-    await update.message.reply_text("Processando aúdio...")
+    reply_clean = re.sub(r'<REFINED>.*?</REFINED>', '', reply, flags=re.DOTALL).strip()
-    await update.message.reply_text("Comando de voz recebido (STT não configurado neste passo).")
+    await update.message.reply_text(reply_clean)
+    
+    # Se foi por voz, responde por voz também
+    if is_voice:
+        audio_path = synthesize_audio(reply_clean)
+        if audio_path and os.path.exists(audio_path):
+            with open(audio_path, 'rb') as voice:
+                await update.message.reply_voice(voice)
+            os.remove(audio_path)
 
 def get_telegram_app():
     if not TOKEN: