Adding voice handler for trash report

pxy_bots/views.py

@@ -1,9 +1,13 @@
+import os
 import json
 import logging
+
+import openai
 from telegram import Update, Bot
 from django.http import JsonResponse
 from django.views.decorators.csrf import csrf_exempt
 from asgiref.sync import sync_to_async
+
 from .models import TelegramBot
 from pxy_langchain.services import LangchainAIService
 from .handlers import (
@@ -15,6 +19,9 @@ from .handlers import (
 
 logger = logging.getLogger(__name__)
 
+# Configura tu API Key de OpenAI
+openai.api_key = os.getenv("OPENAI_API_KEY")
+
 
 # -------------------------------
 # 🛠 Modular local handlers inside views.py
@@ -26,15 +33,6 @@ async def handle_location_message(update):
         return True
     return False
 
-async def handle_voice_or_general_message(update):
-    if update.message.voice:
-        file_id = update.message.voice.file_id
-        await update.message.reply_text(
-            f"🎙 Recibí tu mensaje de voz con ID {file_id}. Pronto calcularé tu CO₂."
-        )
-        return True
-    return False
-
 
 async def dispatch_citizen_commands(update, text):
     if text == "/start":
@@ -92,6 +90,27 @@ async def dispatch_private_commands(update, text):
     return True
 
 
+# -------------------------------
+# 🛠 Voice transcription helper
+# -------------------------------
+
+async def transcribe_with_whisper(update, bot):
+    # 1. Descarga el archivo de voz desde Telegram
+    tg_file = await sync_to_async(bot.get_file)(update.message.voice.file_id)
+    download_path = f"/tmp/{update.message.voice.file_id}.ogg"
+    await sync_to_async(tg_file.download)(download_path)
+
+    # 2. Envía el audio a la API Whisper de OpenAI
+    with open(download_path, "rb") as audio:
+        transcript = openai.Audio.transcribe(
+            model="whisper-1",
+            file=audio,
+            response_format="text",
+            language="es"
+        )
+    return transcript.strip()
+
+
 # -------------------------------
 # 🌐 Main webhook
 # -------------------------------
@@ -101,57 +120,64 @@ async def telegram_webhook(request, bot_name):
     try:
         logger.info(f"Webhook called for bot: {bot_name}")
 
+        # Carga la configuración del bot
         try:
-            bot_instance = await sync_to_async(TelegramBot.objects.get)(name=bot_name, is_active=True)
-            logger.info(f"Loaded bot configuration: {bot_instance}")
+            bot_instance = await sync_to_async(TelegramBot.objects.get)(
+                name=bot_name, is_active=True
+            )
         except TelegramBot.DoesNotExist:
             logger.error(f"Bot '{bot_name}' not found or inactive.")
             return JsonResponse({"error": f"Bot '{bot_name}' not found."}, status=400)
 
         if not bot_instance.assistant:
-            logger.error(f"No assistant configured for bot '{bot_name}'.")
             return JsonResponse({"error": "Assistant not configured."}, status=400)
 
-        if request.method == "POST":
-            try:
-                request_body = json.loads(request.body.decode("utf-8"))
-                update = Update.de_json(request_body, Bot(token=bot_instance.token))
-                logger.info(f"Update received: {update}")
-            except json.JSONDecodeError as e:
-                logger.error(f"Failed to decode JSON: {e}")
-                return JsonResponse({"error": "Invalid JSON payload"}, status=400)
+        if request.method != "POST":
+            return JsonResponse({"error": "Invalid request method"}, status=400)
 
-            if update.message:
-                text = update.message.text or ""
+        # Decodifica el payload de Telegram
+        try:
+            payload = json.loads(request.body.decode("utf-8"))
+            update = Update.de_json(payload, Bot(token=bot_instance.token))
+        except json.JSONDecodeError as e:
+            logger.error(f"Failed to decode JSON: {e}")
+            return JsonResponse({"error": "Invalid JSON payload"}, status=400)
 
-                # Handle location always first
-                if await handle_location_message(update):
-                    return JsonResponse({"status": "ok"})
-
-                # Handle voice or general simple responses next
-                if await handle_voice_or_general_message(update):
-                    return JsonResponse({"status": "ok"})
-
-                # Then dispatch commands per bot
-                if bot_name == "PepeBasuritaCoinsBot":
-                    if await dispatch_citizen_commands(update, text):
-                        return JsonResponse({"status": "ok"})
-                elif bot_name == "PepeCamioncitoBot":
-                    if await dispatch_city_commands(update, text):
-                        return JsonResponse({"status": "ok"})
-                elif bot_name == "PepeMotitoBot":
-                    if await dispatch_private_commands(update, text):
-                        return JsonResponse({"status": "ok"})
-
-                # Otherwise fallback to LLM
-                assistant_instance = await sync_to_async(LangchainAIService)(bot_instance.assistant)
-                bot_response = await sync_to_async(assistant_instance.generate_response)(text)
-                await update.message.reply_text(bot_response)
+        if not update.message:
+            return JsonResponse({"status": "no message"})
 
+        # 1) Geolocalización
+        if await handle_location_message(update):
             return JsonResponse({"status": "ok"})
 
-        logger.warning("Received non-POST request")
-        return JsonResponse({"error": "Invalid request method"}, status=400)
+        # 2) Voz: transcribir y llamar a report_trash
+        if update.message.voice:
+            bot = Bot(token=bot_instance.token)
+            transcript = await transcribe_with_whisper(update, bot)
+            if not transcript:
+                await update.message.reply_text(
+                    "No pude entender tu mensaje de voz. Intenta de nuevo."
+                )
+                return JsonResponse({"status": "ok"})
+            update.message.text = transcript
+            await report_trash(update)
+            return JsonResponse({"status": "ok"})
+
+        # 3) Comandos de texto por bot
+        text = update.message.text or ""
+        if bot_name == "PepeBasuritaCoinsBot" and await dispatch_citizen_commands(update, text):
+            return JsonResponse({"status": "ok"})
+        if bot_name == "PepeCamioncitoBot" and await dispatch_city_commands(update, text):
+            return JsonResponse({"status": "ok"})
+        if bot_name == "PepeMotitoBot" and await dispatch_private_commands(update, text):
+            return JsonResponse({"status": "ok"})
+
+        # 4) Fallback a LLM
+        assistant_instance = await sync_to_async(LangchainAIService)(bot_instance.assistant)
+        bot_response = await sync_to_async(assistant_instance.generate_response)(text)
+        await update.message.reply_text(bot_response)
+
+        return JsonResponse({"status": "ok"})
 
     except Exception as e:
         logger.error(f"Error in webhook: {e}")