app.py

import os
import json
import requests
from fastapi import FastAPI, HTTPException
from pydantic import BaseModel
from google.cloud import storage
from google.auth import exceptions
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
from io import BytesIO
from dotenv import load_dotenv
import uvicorn

load_dotenv()

API_KEY = os.getenv("API_KEY")
GCS_BUCKET_NAME = os.getenv("GCS_BUCKET_NAME")
GOOGLE_APPLICATION_CREDENTIALS_JSON = os.getenv("GOOGLE_APPLICATION_CREDENTIALS_JSON")
HF_API_TOKEN = os.getenv("HF_API_TOKEN")

try:
    credentials_info = json.loads(GOOGLE_APPLICATION_CREDENTIALS_JSON)
    storage_client = storage.Client.from_service_account_info(credentials_info)
    bucket = storage_client.bucket(GCS_BUCKET_NAME)
except (exceptions.DefaultCredentialsError, json.JSONDecodeError, KeyError, ValueError) as e:
    raise RuntimeError(f"Error al cargar credenciales o bucket: {e}")

app = FastAPI()

class DownloadModelRequest(BaseModel):
    model_name: str
    pipeline_task: str
    input_text: str

class GCSHandler:
    def __init__(self, bucket_name):
        self.bucket = storage_client.bucket(bucket_name)

    def file_exists(self, blob_name):
        return self.bucket.blob(blob_name).exists()

    def upload_file(self, blob_name, file_stream):
        blob = self.bucket.blob(blob_name)
        blob.upload_from_file(file_stream)

    def download_file(self, blob_name):
        blob = self.bucket.blob(blob_name)
        if not blob.exists():
            raise HTTPException(status_code=404, detail=f"File '{blob_name}' not found.")
        return BytesIO(blob.download_as_bytes())

def download_model_from_huggingface(model_name):
    url = f"https://huggingface.co/{model_name}/tree/main"
    headers = {"Authorization": f"Bearer {HF_API_TOKEN}"}
    
    # Intentar obtener el árbol de archivos
    try:
        response = requests.get(url, headers=headers)
        if response.status_code == 200:
            # Extraer la lista de archivos del árbol (parseo HTML o JSON depende de la respuesta)
            # Aquí asumimos que el archivo de modelos está disponible
            file_urls = []  # Aquí agregarías la lógica para extraer los enlaces correctos del HTML de la página
            for file_url in file_urls:
                filename = file_url.split("/")[-1]
                blob_name = f"{model_name}/{filename}"
                bucket.blob(blob_name).upload_from_file(BytesIO(requests.get(file_url).content))
        else:
            raise HTTPException(status_code=404, detail="Error al acceder al árbol de archivos de Hugging Face.")
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Error descargando archivos de Hugging Face: {e}")

@app.post("/predict/")
async def predict(request: DownloadModelRequest):
    try:
        gcs_handler = GCSHandler(GCS_BUCKET_NAME)
        model_prefix = request.model_name
        model_files = [
            "pytorch_model.bin",
            "config.json",
            "tokenizer.json",
            "model.safetensors",
        ]
        
        # Verificar si los archivos del modelo están en GCS
        model_files_exist = all(gcs_handler.file_exists(f"{model_prefix}/{file}") for file in model_files)
        
        if not model_files_exist:
            # Descargar el modelo si no existe
            download_model_from_huggingface(model_prefix)
        
        # Descargar los archivos necesarios
        model_files_streams = {file: gcs_handler.download_file(f"{model_prefix}/{file}") for file in model_files if gcs_handler.file_exists(f"{model_prefix}/{file}")}
        
        # Asegurar que los archivos esenciales estén presentes
        config_stream = model_files_streams.get("config.json")
        tokenizer_stream = model_files_streams.get("tokenizer.json")
        
        if not config_stream or not tokenizer_stream:
            raise HTTPException(status_code=500, detail="Required model files missing.")
        
        # Cargar el modelo y el tokenizador
        model = AutoModelForCausalLM.from_pretrained(config_stream)
        tokenizer = AutoTokenizer.from_pretrained(tokenizer_stream)
        
        # Crear un pipeline para la tarea deseada
        pipeline_ = pipeline(request.pipeline_task, model=model, tokenizer=tokenizer)
        
        # Realizar la predicción
        result = pipeline_(request.input_text)
        
        return {"response": result}
    
    except HTTPException as e:
        raise e
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Error: {e}")

if __name__ == "__main__":
    uvicorn.run(app, host="0.0.0.0", port=7860)