app.py

# 修复 huggingface_hub 缺失 cached_download 问题（必须在导入 diffusers 之前）
import huggingface_hub
from huggingface_hub import hf_hub_download
huggingface_hub.cached_download = hf_hub_download  # monkey-patch BEFORE importing diffusers

import gradio as gr
import pandas as pd
import torch
from diffusers import StableDiffusionXLPipeline
from PIL import Image
import os
import zipfile
import tempfile
import uuid

# ==== 模型加载 ====
device = "cuda" if torch.cuda.is_available() else "cpu"
pipe = StableDiffusionXLPipeline.from_pretrained(
    "stabilityai/stable-diffusion-xl-base-1.0",
    torch_dtype=torch.float16 if device == "cuda" else torch.float32,
    variant="fp16" if device == "cuda" else None,
).to(device)

# ==== 生成说明文本 ====
def generate_text(prompt):
    return f"这是一幅描绘：{prompt} 的画面，小朋友可以根据图像发挥想象力哦！"

# ==== 批量 CSV 处理函数 ====
def process_csv(file):
    if not file.name.endswith(".csv"):
        raise gr.Error("请上传一个有效的 CSV 文件（包含 prompt 列）")

    df = pd.read_csv(file.name)
    if "prompt" not in df.columns:
        raise gr.Error("CSV 文件中必须包含名为 'prompt' 的列")

    with tempfile.TemporaryDirectory() as tmpdir:
        output_dir = os.path.join(tmpdir, "output")
        os.makedirs(output_dir, exist_ok=True)

        for idx, row in df.iterrows():
            prompt = row["prompt"]
            image = pipe(prompt=prompt).images[0]
            image_path = os.path.join(output_dir, f"image_{idx+1}.png")
            image.save(image_path)

            text = generate_text(prompt)
            text_path = os.path.join(output_dir, f"text_{idx+1}.txt")
            with open(text_path, "w", encoding="utf-8") as f:
                f.write(text)

        zip_path = os.path.join(tmpdir, f"output_{uuid.uuid4().hex}.zip")
        with zipfile.ZipFile(zip_path, "w") as zipf:
            for file_name in os.listdir(output_dir):
                file_path = os.path.join(output_dir, file_name)
                zipf.write(file_path, arcname=file_name)

        return zip_path

# ==== 单张生成函数 ====
def text_to_image(prompt):
    image = pipe(prompt=prompt).images[0]
    return image

# ==== Gradio 界面 ====
with gr.Blocks() as demo:
    gr.Markdown("## 🎨 AI 图文生成器（批量和单图）")

    with gr.Tab("📂 批量生成（上传CSV）"):
        csv_input = gr.File(label="上传CSV文件（需包含 'prompt' 列）", file_types=[".csv"])
        csv_output = gr.File(label="下载生成的图文ZIP包")
        csv_btn = gr.Button("开始生成")
        csv_btn.click(fn=process_csv, inputs=csv_input, outputs=csv_output)

    with gr.Tab("🖼️ 单图生成（文本转图片）"):
        prompt_input = gr.Textbox(label="输入提示词", placeholder="如：一只飞翔在太空的小猫咪")
        image_output = gr.Image(label="生成的图像", type="pil")
        single_btn = gr.Button("立即生成")
        single_btn.click(fn=text_to_image, inputs=prompt_input, outputs=image_output)

if __name__ == "__main__":
    demo.launch()