import streamlit as st
from transformers import pipeline
import torch
from PIL import Image
import io
import librosa
import numpy as np
import logging
import tempfile
import os
from streamlit.runtime.uploaded_file_manager import UploadedFile
from diffusers import DiffusionPipeline

# Configurar página
st.set_page_config(
    page_title="Aplicação de IA Multi-Modal",
    page_icon="🤖",
    layout="wide"
)

# Configurar logging
logging.basicConfig(
    filename='app_errors.log',
    level=logging.ERROR,
    format='%(asctime)s - %(levelname)s - %(message)s'
)

# Cache para evitar recarregar modelos a cada execução
@st.cache_resource(show_spinner=False)
def load_models():
    \"\"\"Carrega todos os modelos com cache para melhor performance\"\"\"
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    logging.info(f"Usando dispositivo: {device}")
    models = {}
    
    try:
        # Modelos de texto
        models['sentiment_analysis'] = pipeline(
            "sentiment-analysis", 
            model="cardiffnlp/twitter-roberta-base-sentiment-latest",
            device=device
        )
        
        models['text_classification'] = pipeline(
            "text-classification", 
            model="distilbert-base-uncased-finetuned-sst-2-english",
            device=device
        )
        
        models['summarization'] = pipeline(
            "summarization", 
            model="facebook/bart-large-cnn",
            device=device,
            max_length=150,
            min_length=30
        )
        
        models['question_answering'] = pipeline(
            "question-answering", 
            model="deepset/roberta-base-squad2",
            device=device
        )
        
        models['translation'] = pipeline(
            "translation", 
            model="Helsinki-NLP/opus-mt-tc-big-en-pt",
            device=device
        )
        
        models['text_generation'] = pipeline(
            "text-generation",
            model="gpt2",
            device=device,
            max_length=100
        )
        
        models['ner'] = pipeline(
            "ner",
            model="dbmdz/bert-large-cased-finetuned-conll03-english",
            device=device,
            aggregation_strategy="simple"
        )
        
        # Modelos de imagem
        models['image_classification'] = pipeline(
            "image-classification",
            model="google/vit-base-patch16-224",
            device=device
        )
        
        models['object_detection'] = pipeline(
            "object-detection",
            model="facebook/detr-resnet-50",
            device=device
        )
        
        models['image_segmentation'] = pipeline(
            "image-segmentation",
            model="facebook/detr-resnet-50-panoptic",
            device=device
        )
        
        models['facial_recognition'] = pipeline(
            "image-classification",
            model="mo-thecreator/vit-Facial-Expression-Recognition",
            device=device
        )
        
        # Modelos de áudio
        models['speech_to_text'] = pipeline(
            "automatic-speech-recognition",
            model="openai/whisper-base",
            device=device
        )
        
        models['audio_classification'] = pipeline(
            "audio-classification",
            model="superb/hubert-base-superb-er",
            device=device
        )
        
        # Modelos generativos
        models['text_to_image'] = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0",
            torch_dtype=torch.float16,
            use_safetensors=True,
            variant="fp16"
        )
        
    except Exception as e:
        st.error(f"Erro crítico ao carregar modelos: {str(e)}")
        logging.exception("Erro ao carregar modelos")
        return {}
    
    return models

# (As outras funções ficam iguais ao teu app original)

def main():
    st.title("🤖 Aplicação de IA Multi-Modal Avançada")
    st.markdown("---")
    
    # Carregar modelos
    with st.spinner("Carregando modelos de IA... (Isso pode levar alguns minutos na primeira execução)"):
        models = load_models()
    
    if not models:
        st.error("Falha crítica ao carregar os modelos. Verifique os logs para mais detalhes.")
        return
    
    st.write("✅ Aplicação carregada com sucesso!")
    # (Resto do teu código)

if __name__ == "__main__":
    main()