app/audio_processor.py

import base64
import tempfile
import os
from app.config import WHISPER_MODEL, WHISPER_MODEL_PATH
os.environ["XDG_CACHE_HOME"] = str(WHISPER_MODEL_PATH) # "./models"
from app.logger import log
from app.sentiment_model import analyze_sentiment_portgues
from app.translator import translate_long_text
from app.emotion_model import analyze_emotion
import whisper
from moviepy.editor import VideoFileClip
whisper_model = whisper.load_model(WHISPER_MODEL)


# -------------------------
# 🔥 UTIL: salvar base64
# -------------------------
def save_base64_to_file(base64_data, suffix):
    try:
        file_bytes = base64.b64decode(base64_data)
    except Exception:
        raise ValueError("Base64 inválido")

    temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=suffix)

    with open(temp_file.name, "wb") as f:
        f.write(file_bytes)

    return temp_file.name

# -------------------------
# 🔥 UTIL: extrair áudio
# -------------------------
def extract_audio_from_video(video_path):
    temp_audio = tempfile.NamedTemporaryFile(delete=False, suffix=".wav")

    video = VideoFileClip(video_path)
    video.audio.write_audiofile(temp_audio.name)
    video.close()

    return temp_audio.name


# -------------------------
# 🔥 PROCESSAMENTO PRINCIPAL
# -------------------------
def process_audio(video_base64=None, audio_base64=None, logger=None):

    video_path = None
    audio_path = None

    try:
        # -------------------------
        # 🎯 1. ORIGEM DO ÁUDIO
        # -------------------------

        if audio_base64:
            log("Recebido áudio base64", logger=logger)
            audio_path = save_base64_to_file(audio_base64, ".wav")

        elif video_base64:
            log("Recebido vídeo base64", logger=logger)
            video_path = save_base64_to_file(video_base64, ".mp4")

            log("Extraindo áudio do vídeo...", logger=logger)
            audio_path = extract_audio_from_video(video_path)

        else:
            raise ValueError("Nenhum áudio ou vídeo fornecido")

        # -------------------------
        # 🎯 2. WHISPER
        # -------------------------

        log("Transcrevendo com Whisper...", logger=logger)

        result = whisper_model.transcribe(
            audio_path,
            language="pt",
            verbose=False,
            fp16=False  # importante para CPU
        )

        text = " ".join([seg["text"] for seg in result["segments"]])

        # -------------------------
        # 🎯 3. ROBERTA (cardiffnlp/twitter-xlm-roberta)
        # -------------------------
        #sentiment, score = analyze_sentiment_portgues(text, logger)
        #print(f"Sentimento: {sentiment} ({score})")
        #return {
        #    "transcription": text,
        #    "sentiment": sentiment,
        #    "score": score
        #}

        # -------------------------
        # 🎯 3. TRADUÇÃO + EMOÇÃO
        # -------------------------

        print("Traduzindo texto para inglês...")
        text_en = translate_long_text(text)

        print("Analisando emoção...")
        emotion_result = analyze_emotion(text_en)

        return {
            "transcription": text,
            "translation": text_en,

            "analysis": {
                "emotion": emotion_result["emotion"],
                "confidence": emotion_result["confidence"],
                "emotion_raw": emotion_result["emotion_raw"],
                "all_emotions": emotion_result["all_emotions"]
            }
        }
    
    finally:
        # -------------------------
        # 🧹 LIMPEZA (MUITO IMPORTANTE)
        # -------------------------

        if video_path and os.path.exists(video_path):
            os.remove(video_path)

        if audio_path and os.path.exists(audio_path):
            os.remove(audio_path)