app.py

import gradio as gr
import random
import os
import requests
import base64
from PIL import Image
from io import BytesIO

HF_TOKEN = os.environ.get("girlToken")

API_BASE = "https://prithivmlmods-qwen-image-edit-2511-loras-fast.hf.space"
INFER_URL = f"{API_BASE}/gradio/infer"
NAMED_API_URL = f"{API_BASE}/gradio_api/call/v2/infer"

LORA_STYLES = [
    'Multiple-Angles', 'Photo-to-Anime', 'Anime-V2', 'Light-Migration',
    'Upscaler', 'Style-Transfer', 'Manga-Tone', 'Anything2Real',
    'Fal-Multiple-Angles', 'Polaroid-Photo', 'Unblur-Anything',
    'Midnight-Noir-Eyes-Spotlight', 'Hyper-Realistic-Portrait',
    'Ultra-Realistic-Portrait', 'Pixar-Inspired-3D', 'Noir-Comic-Book',
    'Any-light', 'Studio-DeLight', 'Cinematic-FlatLog',
]
MAX_SEED = 2**31 - 1

def encode_image_file_to_b64_payload(image_path):
    """
    返回符合API要求的图片Payload对象(list，不是json字符串)。
    """
    try:
        with open(image_path, "rb") as f:
            image_bytes = f.read()
        # 如非jpeg，转码
        try:
            img = Image.open(BytesIO(image_bytes))
            buffered = BytesIO()
            img.save(buffered, format="JPEG")
            image_bytes = buffered.getvalue()
        except Exception:
            pass
        im_b64 = base64.b64encode(image_bytes).decode("utf-8")
        payload = [
            {
                "data": im_b64,
                "mime_type": "image/jpeg",
                "orig_name": os.path.basename(image_path),
            }
        ]
        return payload
    except Exception as e:
        raise RuntimeError(f"图片编码失败: {e}")

def call_named_infer(
    images_b64_payload,
    prompt,
    lora_adapter,
    seed,
    randomize_seed,
    guidance_scale,
    steps
):
    headers = {
        'Authorization': f'Bearer {HF_TOKEN}',
        'Content-Type': 'application/json'
    }
    payload = {
        "images_b64_json": images_b64_payload, # 注意此处改为直接传 list
        "prompt": prompt,
        "lora_adapter": lora_adapter,
        "seed": int(seed),
        "randomize_seed": bool(randomize_seed),
        "guidance_scale": float(guidance_scale),
        "steps": int(steps),
    }
    print("准备调用/infer:", {**payload, "images_b64_json": "[payload omitted for brevity]"})
    import json
    # 注意: 直接以json.dumps(payload)传body
    resp = requests.post(NAMED_API_URL, data=json.dumps(payload), headers=headers)
    resp.raise_for_status()
    job = resp.json()
    event_id = job.get("event_id")
    return event_id

def poll_infer(event_id):
    url = f"{API_BASE}/gradio_api/call/infer/{event_id}"
    headers = {'Authorization': f'Bearer {HF_TOKEN}'}
    import time
    for i in range(60):
        resp = requests.get(url, headers=headers)
        try:
            result = resp.json()
        except Exception:
            print(f"[轮询第{i+1}次] 响应无法decode，返回内容：{resp.text[:200]}")
            time.sleep(2)
            continue
        if result.get("status") == "complete":
            return result.get("data"), result.get("outputs")
        elif result.get("status") == "error":
            raise Exception(result.get("error"))
        time.sleep(2)
    raise TimeoutError("等候API返回超时")

def infer(
    image,
    prompt,
    lora_adapter,
    seed,
    randomize_seed,
    guidance_scale,
    steps,
    progress=gr.Progress(track_tqdm=True),
):
    if image is None:
        print("未上传图片")
        return None, seed

    if not os.path.exists(image):
        print(f"图片路径不存在: {image}")
        return None, seed

    if randomize_seed:
        seed = random.randint(0, MAX_SEED)

    try:
        images_b64_payload = encode_image_file_to_b64_payload(image)
    except Exception as e:
        print(f"[图片 base64编码失败] {e}")
        return None, seed

    try:
        event_id = call_named_infer(
            images_b64_payload,
            prompt,
            lora_adapter,
            seed,
            randomize_seed,
            guidance_scale,
            steps
        )
        print("API返回event_id:", event_id)
        data, outputs = poll_infer(event_id)
        print("[API 完成] data:", data, "outputs:", outputs)
        img_out = None
        seed_used = seed
        if outputs:
            if isinstance(outputs, dict):
                img_out = outputs.get("url") or outputs.get("path")
                seed_used = outputs.get("seed", seed)
            elif isinstance(outputs, str) and outputs.startswith("/"):
                img_out = API_BASE + outputs
            else:
                img_out = outputs
        elif data:
            if isinstance(data, dict):
                img_out = data.get("url") or data.get("path")
                seed_used = data.get("seed", seed)
            elif isinstance(data, str) and data.startswith("/"):
                img_out = API_BASE + data
            else:
                img_out = data
        if img_out and isinstance(img_out, str) and not img_out.startswith("http"):
            img_out = API_BASE + img_out
        return img_out, int(seed_used)
    except Exception as e:
        import traceback
        traceback.print_exc()
        print(f"[API 调用异常] {e}")
        return None, seed

css = """
#col-container {
    margin: 0 auto;
    max-width: 640px;
}
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.Markdown("# 图像编辑 Demo\n基于 prithivMLmods/Qwen-Image-Edit-2511-LoRAs-Fast (新版API)")

        image = gr.Image(
            label="上传图片",
            sources=["upload"],
            type="filepath",
        )

        prompt = gr.Text(
            label="编辑描述（Prompt）",
            placeholder="请输入图片编辑描述...",
        )

        lora_adapter = gr.Dropdown(
            label="编辑风格（Style）",
            choices=LORA_STYLES,
            value="Photo-to-Anime"
        )

        run_button = gr.Button("执行编辑", variant="primary")

        result = gr.Image(label="结果图片", show_label=True)

        with gr.Accordion("高级设置", open=False):
            seed = gr.Slider(
                label="随机种子",
                minimum=0,
                maximum=MAX_SEED,
                step=1,
                value=0,
            )
            randomize_seed = gr.Checkbox(label="随机化种子", value=True)
            guidance_scale = gr.Slider(
                label="引导强度 (Guidance Scale)",
                minimum=1.0,
                maximum=10.0,
                step=0.1,
                value=1.0,
            )
            steps = gr.Slider(
                label="推理步数 (Steps)",
                minimum=1,
                maximum=50,
                step=1,
                value=4,
            )

    gr.on(
        triggers=[run_button.click, prompt.submit],
        fn=infer,
        inputs=[image, prompt, lora_adapter, seed, randomize_seed, guidance_scale, steps],
        outputs=[result, seed],
    )

if __name__ == "__main__":
    demo.launch(ssr_mode=False, share=True)