|
|
import gradio as gr |
|
|
from fastapi import FastAPI, HTTPException, Body |
|
|
from fastapi.responses import StreamingResponse, JSONResponse |
|
|
from fastapi.staticfiles import StaticFiles |
|
|
from fastapi.middleware.cors import CORSMiddleware |
|
|
from pydantic import BaseModel |
|
|
from selenium import webdriver |
|
|
from selenium.webdriver.chrome.options import Options |
|
|
from selenium.webdriver.common.by import By |
|
|
from selenium.webdriver.support.ui import WebDriverWait |
|
|
from selenium.webdriver.support import expected_conditions as EC |
|
|
from PIL import Image |
|
|
from io import BytesIO |
|
|
import tempfile |
|
|
import time |
|
|
import os |
|
|
import logging |
|
|
import numpy as np |
|
|
import threading |
|
|
import queue |
|
|
import uuid |
|
|
from datetime import datetime |
|
|
from concurrent.futures import ThreadPoolExecutor |
|
|
from huggingface_hub import hf_hub_download, upload_file, login |
|
|
|
|
|
|
|
|
import google.generativeai as genai |
|
|
|
|
|
|
|
|
logging.basicConfig(level=logging.INFO) |
|
|
logger = logging.getLogger(__name__) |
|
|
|
|
|
|
|
|
class HuggingFaceUploader: |
|
|
"""HuggingFace Hubへ画像をアップロードする機能を提供するクラス""" |
|
|
def __init__(self): |
|
|
self.repo_id = os.environ.get("HF_REPO_ID", "tomo2chin2/SUPER_TENSAI_JIN") |
|
|
self.token = os.environ.get("HF_TOKEN", None) |
|
|
if self.token: |
|
|
try: |
|
|
login(token=self.token) |
|
|
logger.info(f"HuggingFace Hubにログインしました。リポジトリ: {self.repo_id}") |
|
|
except Exception as e: |
|
|
logger.error(f"HuggingFace Hubへのログインに失敗: {e}") |
|
|
else: |
|
|
logger.warning("HF_TOKEN環境変数が設定されていません。アップロード機能は制限されます。") |
|
|
|
|
|
def upload_image(self, image, prefix="generated"): |
|
|
""" |
|
|
PIL Imageをアップロードし、アクセス可能なURLを返す |
|
|
|
|
|
Args: |
|
|
image: PIL.Image - アップロードする画像 |
|
|
prefix: str - ファイル名のプレフィックス |
|
|
|
|
|
Returns: |
|
|
str - アップロードされた画像のURL |
|
|
""" |
|
|
try: |
|
|
if not self.token: |
|
|
logger.error("HF_TOKENが設定されていないため、アップロードできません") |
|
|
return None |
|
|
|
|
|
|
|
|
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S") |
|
|
unique_id = str(uuid.uuid4())[:8] |
|
|
filename = f"{prefix}_{timestamp}_{unique_id}.jpg" |
|
|
path_in_repo = f"images/{filename}" |
|
|
|
|
|
|
|
|
with tempfile.NamedTemporaryFile(suffix=".jpg", delete=False) as tmp_file: |
|
|
tmp_path = tmp_file.name |
|
|
image.save(tmp_path, format="JPEG", quality=95) |
|
|
|
|
|
logger.info(f"画像を一時ファイルに保存: {tmp_path}") |
|
|
|
|
|
|
|
|
logger.info(f"HuggingFace Hubにアップロード中: {path_in_repo}") |
|
|
upload_info = upload_file( |
|
|
path_or_fileobj=tmp_path, |
|
|
path_in_repo=path_in_repo, |
|
|
repo_id=self.repo_id, |
|
|
repo_type="dataset" |
|
|
) |
|
|
|
|
|
|
|
|
try: |
|
|
os.remove(tmp_path) |
|
|
except Exception as e: |
|
|
logger.warning(f"一時ファイル削除エラー: {e}") |
|
|
|
|
|
|
|
|
url = f"https://huggingface.co/datasets/{self.repo_id}/resolve/main/{path_in_repo}" |
|
|
logger.info(f"アップロード成功: {url}") |
|
|
return url |
|
|
|
|
|
except Exception as e: |
|
|
logger.error(f"HuggingFace Hubへのアップロード中にエラー: {e}", exc_info=True) |
|
|
return None |
|
|
|
|
|
|
|
|
hf_uploader = HuggingFaceUploader() |
|
|
|
|
|
|
|
|
class WebDriverPool: |
|
|
"""WebDriverインスタンスを再利用するためのプール""" |
|
|
def __init__(self, max_drivers=3): |
|
|
self.driver_queue = queue.Queue() |
|
|
self.max_drivers = max_drivers |
|
|
self.lock = threading.Lock() |
|
|
self.count = 0 |
|
|
logger.info(f"WebDriverプールを初期化: 最大 {max_drivers} ドライバー") |
|
|
|
|
|
def get_driver(self): |
|
|
"""プールからWebDriverを取得、なければ新規作成""" |
|
|
if not self.driver_queue.empty(): |
|
|
logger.info("既存のWebDriverをプールから取得") |
|
|
return self.driver_queue.get() |
|
|
|
|
|
with self.lock: |
|
|
if self.count < self.max_drivers: |
|
|
self.count += 1 |
|
|
logger.info(f"新しいWebDriverを作成 (合計: {self.count}/{self.max_drivers})") |
|
|
options = Options() |
|
|
options.add_argument("--headless") |
|
|
options.add_argument("--no-sandbox") |
|
|
options.add_argument("--disable-dev-shm-usage") |
|
|
options.add_argument("--force-device-scale-factor=1") |
|
|
options.add_argument("--disable-features=NetworkService") |
|
|
options.add_argument("--dns-prefetch-disable") |
|
|
|
|
|
|
|
|
webdriver_path = os.environ.get("CHROMEDRIVER_PATH") |
|
|
if webdriver_path and os.path.exists(webdriver_path): |
|
|
logger.info(f"CHROMEDRIVER_PATH使用: {webdriver_path}") |
|
|
service = webdriver.ChromeService(executable_path=webdriver_path) |
|
|
return webdriver.Chrome(service=service, options=options) |
|
|
else: |
|
|
logger.info("デフォルトのChromeDriverを使用") |
|
|
return webdriver.Chrome(options=options) |
|
|
|
|
|
|
|
|
logger.info("WebDriverプールがいっぱいです。利用可能なドライバーを待機中...") |
|
|
return self.driver_queue.get() |
|
|
|
|
|
def release_driver(self, driver): |
|
|
"""ドライバーをプールに戻す""" |
|
|
if driver: |
|
|
try: |
|
|
|
|
|
driver.get("about:blank") |
|
|
driver.execute_script(""" |
|
|
document.documentElement.style.overflow = ''; |
|
|
document.body.style.overflow = ''; |
|
|
""") |
|
|
self.driver_queue.put(driver) |
|
|
logger.info("WebDriverをプールに戻しました") |
|
|
except Exception as e: |
|
|
logger.error(f"ドライバーをプールに戻す際にエラー: {e}") |
|
|
driver.quit() |
|
|
with self.lock: |
|
|
self.count -= 1 |
|
|
|
|
|
def close_all(self): |
|
|
"""全てのドライバーを終了""" |
|
|
logger.info("WebDriverプールを終了します") |
|
|
closed = 0 |
|
|
while not self.driver_queue.empty(): |
|
|
try: |
|
|
driver = self.driver_queue.get(block=False) |
|
|
driver.quit() |
|
|
closed += 1 |
|
|
except queue.Empty: |
|
|
break |
|
|
except Exception as e: |
|
|
logger.error(f"ドライバー終了中にエラー: {e}") |
|
|
|
|
|
logger.info(f"{closed}個のWebDriverを終了しました") |
|
|
with self.lock: |
|
|
self.count = 0 |
|
|
|
|
|
|
|
|
|
|
|
driver_pool = WebDriverPool(max_drivers=int(os.environ.get("MAX_WEBDRIVERS", "3"))) |
|
|
|
|
|
|
|
|
class GeminiRequest(BaseModel): |
|
|
"""Geminiへのリクエストデータモデル""" |
|
|
text: str |
|
|
extension_percentage: float = 10.0 |
|
|
temperature: float = 0.5 |
|
|
trim_whitespace: bool = True |
|
|
style: str = "standard" |
|
|
|
|
|
class ScreenshotRequest(BaseModel): |
|
|
"""スクリーンショットリクエストモデル""" |
|
|
html_code: str |
|
|
extension_percentage: float = 10.0 |
|
|
trim_whitespace: bool = True |
|
|
style: str = "standard" |
|
|
|
|
|
|
|
|
class ImageUrlResponse(BaseModel): |
|
|
"""画像URLのレスポンスモデル""" |
|
|
url: str |
|
|
status: str = "success" |
|
|
message: str = "画像が正常に生成されました" |
|
|
|
|
|
|
|
|
def enhance_font_awesome_layout(html_code): |
|
|
"""Font Awesomeレイアウトを改善し、プリロードタグを追加""" |
|
|
|
|
|
fa_preload = """ |
|
|
<link rel="preload" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/5.15.4/webfonts/fa-solid-900.woff2" as="font" type="font/woff2" crossorigin> |
|
|
<link rel="preload" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/5.15.4/webfonts/fa-regular-400.woff2" as="font" type="font/woff2" crossorigin> |
|
|
<link rel="preload" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/5.15.4/webfonts/fa-brands-400.woff2" as="font" type="font/woff2" crossorigin> |
|
|
""" |
|
|
|
|
|
|
|
|
fa_fix_css = """ |
|
|
<style> |
|
|
/* Font Awesomeアイコンのレイアウト修正 */ |
|
|
[class*="fa-"] { |
|
|
display: inline-block !important; |
|
|
margin-right: 8px !important; |
|
|
vertical-align: middle !important; |
|
|
} |
|
|
|
|
|
/* テキスト内のアイコン位置調整 */ |
|
|
h1 [class*="fa-"], h2 [class*="fa-"], h3 [class*="fa-"], |
|
|
h4 [class*="fa-"], h5 [class*="fa-"], h6 [class*="fa-"] { |
|
|
vertical-align: middle !important; |
|
|
margin-right: 10px !important; |
|
|
} |
|
|
|
|
|
/* 特定パターンの修正 */ |
|
|
.fa + span, .fas + span, .far + span, .fab + span, |
|
|
span + .fa, span + .fas, span + .far + span { |
|
|
display: inline-block !important; |
|
|
margin-left: 5px !important; |
|
|
} |
|
|
|
|
|
/* カード内アイコン修正 */ |
|
|
.card [class*="fa-"], .card-body [class*="fa-"] { |
|
|
float: none !important; |
|
|
clear: none !important; |
|
|
position: relative !important; |
|
|
} |
|
|
|
|
|
/* アイコンと文字が重なる場合の調整 */ |
|
|
li [class*="fa-"], p [class*="fa-"] { |
|
|
margin-right: 10px !important; |
|
|
} |
|
|
|
|
|
/* インラインアイコンのスペーシング */ |
|
|
.inline-icon { |
|
|
display: inline-flex !important; |
|
|
align-items: center !important; |
|
|
justify-content: flex-start !important; |
|
|
} |
|
|
|
|
|
/* アイコン後のテキスト */ |
|
|
[class*="fa-"] + span { |
|
|
display: inline-block !important; |
|
|
vertical-align: middle !important; |
|
|
} |
|
|
</style> |
|
|
""" |
|
|
|
|
|
|
|
|
if '<head>' in html_code: |
|
|
return html_code.replace('</head>', f'{fa_preload}{fa_fix_css}</head>') |
|
|
|
|
|
elif '<html' in html_code: |
|
|
head_end = html_code.find('</head>') |
|
|
if head_end > 0: |
|
|
return html_code[:head_end] + fa_preload + fa_fix_css + html_code[head_end:] |
|
|
else: |
|
|
body_start = html_code.find('<body') |
|
|
if body_start > 0: |
|
|
return html_code[:body_start] + f'<head>{fa_preload}{fa_fix_css}</head>' + html_code[body_start:] |
|
|
|
|
|
|
|
|
return f'<html><head>{fa_preload}{fa_fix_css}</head>' + html_code + '</html>' |
|
|
|
|
|
def load_system_instruction(style="standard"): |
|
|
""" |
|
|
指定されたスタイルのシステムインストラクションを読み込む |
|
|
|
|
|
Args: |
|
|
style: 使用するスタイル名 (standard, cute, resort, cool, dental, school) |
|
|
|
|
|
Returns: |
|
|
読み込まれたシステムインストラクション |
|
|
""" |
|
|
try: |
|
|
|
|
|
valid_styles = ["standard", "cute", "resort", "cool", "dental", "school"] |
|
|
|
|
|
|
|
|
if style not in valid_styles: |
|
|
logger.warning(f"無効なスタイル '{style}' が指定されました。デフォルトの 'standard' を使用します。") |
|
|
style = "standard" |
|
|
|
|
|
logger.info(f"スタイル '{style}' のシステムインストラクションを読み込みます") |
|
|
|
|
|
|
|
|
local_path = os.path.join(os.path.dirname(__file__), style, "prompt.txt") |
|
|
|
|
|
|
|
|
if os.path.exists(local_path): |
|
|
logger.info(f"ローカルファイルを使用: {local_path}") |
|
|
with open(local_path, 'r', encoding='utf-8') as file: |
|
|
instruction = file.read() |
|
|
return instruction |
|
|
|
|
|
|
|
|
try: |
|
|
|
|
|
file_path = hf_hub_download( |
|
|
repo_id="tomo2chin2/GURAREKOstlyle", |
|
|
filename=f"{style}/prompt.txt", |
|
|
repo_type="dataset" |
|
|
) |
|
|
|
|
|
logger.info(f"スタイル '{style}' のプロンプトをHuggingFaceから読み込みました: {file_path}") |
|
|
with open(file_path, 'r', encoding='utf-8') as file: |
|
|
instruction = file.read() |
|
|
return instruction |
|
|
|
|
|
except Exception as style_error: |
|
|
|
|
|
logger.warning(f"スタイル '{style}' のプロンプト読み込みに失敗: {str(style_error)}") |
|
|
logger.info("デフォルトのprompt.txtを読み込みます") |
|
|
|
|
|
file_path = hf_hub_download( |
|
|
repo_id="tomo2chin2/GURAREKOstlyle", |
|
|
filename="prompt.txt", |
|
|
repo_type="dataset" |
|
|
) |
|
|
|
|
|
with open(file_path, 'r', encoding='utf-8') as file: |
|
|
instruction = file.read() |
|
|
|
|
|
logger.info("デフォルトのシステムインストラクションを読み込みました") |
|
|
return instruction |
|
|
|
|
|
except Exception as e: |
|
|
error_msg = f"システムインストラクションの読み込みに失敗: {str(e)}" |
|
|
logger.error(error_msg) |
|
|
raise ValueError(error_msg) |
|
|
|
|
|
def generate_html_from_text(text, temperature=0.3, style="standard"): |
|
|
"""テキストからHTMLを生成する""" |
|
|
try: |
|
|
|
|
|
api_key = os.environ.get("GEMINI_API_KEY") |
|
|
if not api_key: |
|
|
logger.error("GEMINI_API_KEY 環境変数が設定されていません") |
|
|
raise ValueError("GEMINI_API_KEY 環境変数が設定されていません") |
|
|
|
|
|
|
|
|
model_name = os.environ.get("GEMINI_MODEL", "gemini-1.5-pro") |
|
|
logger.info(f"使用するGeminiモデル: {model_name}") |
|
|
|
|
|
|
|
|
genai.configure(api_key=api_key) |
|
|
|
|
|
|
|
|
system_instruction = load_system_instruction(style) |
|
|
|
|
|
|
|
|
logger.info(f"Gemini APIにリクエストを送信: テキスト長さ = {len(text)}, 温度 = {temperature}, スタイル = {style}") |
|
|
|
|
|
|
|
|
model = genai.GenerativeModel(model_name) |
|
|
|
|
|
|
|
|
generation_config = { |
|
|
"temperature": temperature, |
|
|
"top_p": 0.7, |
|
|
"top_k": 20, |
|
|
"max_output_tokens": 8192, |
|
|
"candidate_count": 1 |
|
|
} |
|
|
|
|
|
|
|
|
safety_settings = [ |
|
|
{ |
|
|
"category": "HARM_CATEGORY_HARASSMENT", |
|
|
"threshold": "BLOCK_MEDIUM_AND_ABOVE" |
|
|
}, |
|
|
{ |
|
|
"category": "HARM_CATEGORY_HATE_SPEECH", |
|
|
"threshold": "BLOCK_MEDIUM_AND_ABOVE" |
|
|
}, |
|
|
{ |
|
|
"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT", |
|
|
"threshold": "BLOCK_MEDIUM_AND_ABOVE" |
|
|
}, |
|
|
{ |
|
|
"category": "HARM_CATEGORY_DANGEROUS_CONTENT", |
|
|
"threshold": "BLOCK_MEDIUM_AND_ABOVE" |
|
|
} |
|
|
] |
|
|
|
|
|
|
|
|
prompt = f"{system_instruction}\n\n{text}" |
|
|
|
|
|
|
|
|
response = model.generate_content( |
|
|
prompt, |
|
|
generation_config=generation_config, |
|
|
safety_settings=safety_settings |
|
|
) |
|
|
|
|
|
|
|
|
raw_response = response.text |
|
|
|
|
|
|
|
|
html_start = raw_response.find("```html") |
|
|
html_end = raw_response.rfind("```") |
|
|
|
|
|
if html_start != -1 and html_end != -1 and html_start < html_end: |
|
|
html_start += 7 |
|
|
html_code = raw_response[html_start:html_end].strip() |
|
|
logger.info(f"HTMLの生成に成功: 長さ = {len(html_code)}") |
|
|
|
|
|
|
|
|
html_code = enhance_font_awesome_layout(html_code) |
|
|
logger.info("Font Awesomeレイアウトの最適化を適用しました") |
|
|
|
|
|
return html_code |
|
|
else: |
|
|
|
|
|
logger.warning("レスポンスから ```html ``` タグが見つかりませんでした。全テキストを返します。") |
|
|
return raw_response |
|
|
|
|
|
except Exception as e: |
|
|
logger.error(f"HTML生成中にエラーが発生: {e}", exc_info=True) |
|
|
raise Exception(f"Gemini APIでのHTML生成に失敗しました: {e}") |
|
|
|
|
|
|
|
|
def trim_image_whitespace(image, threshold=250, padding=10): |
|
|
""" |
|
|
NumPyを使用して最適化された画像トリミング関数 |
|
|
|
|
|
Args: |
|
|
image: PIL.Image - 入力画像 |
|
|
threshold: int - どの明るさ以上を空白と判断するか (0-255) |
|
|
padding: int - トリミング後に残す余白のピクセル数 |
|
|
|
|
|
Returns: |
|
|
トリミングされたPIL.Image |
|
|
""" |
|
|
try: |
|
|
|
|
|
gray = image.convert('L') |
|
|
|
|
|
|
|
|
np_image = np.array(gray) |
|
|
|
|
|
|
|
|
mask = np_image < threshold |
|
|
|
|
|
|
|
|
rows = np.any(mask, axis=1) |
|
|
cols = np.any(mask, axis=0) |
|
|
|
|
|
|
|
|
if np.any(rows) and np.any(cols): |
|
|
row_indices = np.where(rows)[0] |
|
|
col_indices = np.where(cols)[0] |
|
|
|
|
|
|
|
|
min_y, max_y = row_indices[0], row_indices[-1] |
|
|
min_x, max_x = col_indices[0], col_indices[-1] |
|
|
|
|
|
|
|
|
min_x = max(0, min_x - padding) |
|
|
min_y = max(0, min_y - padding) |
|
|
max_x = min(image.width - 1, max_x + padding) |
|
|
max_y = min(image.height - 1, max_y + padding) |
|
|
|
|
|
|
|
|
trimmed = image.crop((min_x, min_y, max_x + 1, max_y + 1)) |
|
|
|
|
|
logger.info(f"画像をトリミングしました: 元サイズ {image.width}x{image.height} → トリミング後 {trimmed.width}x{trimmed.height}") |
|
|
return trimmed |
|
|
|
|
|
logger.warning("トリミング領域が見つかりません。元の画像を返します。") |
|
|
return image |
|
|
|
|
|
except Exception as e: |
|
|
logger.error(f"画像トリミング中にエラー: {e}", exc_info=True) |
|
|
return image |
|
|
|
|
|
|
|
|
def render_fullpage_screenshot(html_code: str, extension_percentage: float = 6.0, |
|
|
trim_whitespace: bool = True, driver=None) -> Image.Image: |
|
|
""" |
|
|
Renders HTML code to a full-page screenshot using Selenium. |
|
|
Optimized to accept an external driver or get one from the pool. |
|
|
|
|
|
Args: |
|
|
html_code: The HTML source code string. |
|
|
extension_percentage: Percentage of extra space to add vertically. |
|
|
trim_whitespace: Whether to trim excess whitespace from the image. |
|
|
driver: An optional pre-initialized WebDriver instance. |
|
|
|
|
|
Returns: |
|
|
A PIL Image object of the screenshot. |
|
|
""" |
|
|
tmp_path = None |
|
|
driver_from_pool = False |
|
|
|
|
|
|
|
|
if driver is None: |
|
|
driver = driver_pool.get_driver() |
|
|
driver_from_pool = True |
|
|
logger.info("WebDriverプールからドライバーを取得しました") |
|
|
|
|
|
|
|
|
try: |
|
|
with tempfile.NamedTemporaryFile(suffix=".html", delete=False, mode='w', encoding='utf-8') as tmp_file: |
|
|
tmp_path = tmp_file.name |
|
|
tmp_file.write(html_code) |
|
|
logger.info(f"HTML saved to temporary file: {tmp_path}") |
|
|
except Exception as e: |
|
|
logger.error(f"Error writing temporary HTML file: {e}") |
|
|
if driver_from_pool: |
|
|
driver_pool.release_driver(driver) |
|
|
return Image.new('RGB', (1, 1), color=(0, 0, 0)) |
|
|
|
|
|
try: |
|
|
|
|
|
initial_width = 1200 |
|
|
initial_height = 1000 |
|
|
driver.set_window_size(initial_width, initial_height) |
|
|
file_url = "file://" + tmp_path |
|
|
logger.info(f"Navigating to {file_url}") |
|
|
driver.get(file_url) |
|
|
|
|
|
|
|
|
logger.info("Waiting for body element...") |
|
|
WebDriverWait(driver, 10).until( |
|
|
EC.presence_of_element_located((By.TAG_NAME, "body")) |
|
|
) |
|
|
logger.info("Body element found. Waiting for resource loading...") |
|
|
|
|
|
|
|
|
max_wait = 5 |
|
|
wait_increment = 0.2 |
|
|
wait_time = 0 |
|
|
|
|
|
while wait_time < max_wait: |
|
|
resource_state = driver.execute_script(""" |
|
|
return { |
|
|
complete: document.readyState === 'complete', |
|
|
imgCount: document.images.length, |
|
|
imgLoaded: Array.from(document.images).filter(img => img.complete).length, |
|
|
faElements: document.querySelectorAll('.fa, .fas, .far, .fab, [class*="fa-"]').length |
|
|
}; |
|
|
""") |
|
|
|
|
|
|
|
|
if resource_state['complete'] and (resource_state['imgCount'] == 0 or |
|
|
resource_state['imgLoaded'] == resource_state['imgCount']): |
|
|
logger.info(f"リソース読み込み完了: {resource_state}") |
|
|
break |
|
|
|
|
|
time.sleep(wait_increment) |
|
|
wait_time += wait_increment |
|
|
logger.info(f"リソース待機中... {wait_time:.1f}秒経過, 状態: {resource_state}") |
|
|
|
|
|
|
|
|
fa_count = resource_state.get('faElements', 0) |
|
|
if fa_count > 30: |
|
|
logger.info(f"{fa_count}個のFont Awesome要素があるため、追加待機...") |
|
|
time.sleep(min(1.0, fa_count / 100)) |
|
|
|
|
|
|
|
|
logger.info("Performing content rendering scroll...") |
|
|
total_height = driver.execute_script("return Math.max(document.body.scrollHeight, document.documentElement.scrollHeight);") |
|
|
viewport_height = driver.execute_script("return window.innerHeight;") |
|
|
scrolls_needed = max(1, min(5, total_height // viewport_height)) |
|
|
|
|
|
|
|
|
for i in range(scrolls_needed): |
|
|
scroll_pos = i * (viewport_height - 100) |
|
|
driver.execute_script(f"window.scrollTo(0, {scroll_pos});") |
|
|
time.sleep(0.1) |
|
|
|
|
|
|
|
|
driver.execute_script("window.scrollTo(0, 0);") |
|
|
time.sleep(0.2) |
|
|
logger.info("Scroll rendering completed") |
|
|
|
|
|
|
|
|
driver.execute_script(""" |
|
|
document.documentElement.style.overflow = 'hidden'; |
|
|
document.body.style.overflow = 'hidden'; |
|
|
""") |
|
|
|
|
|
|
|
|
dimensions = driver.execute_script(""" |
|
|
return { |
|
|
width: Math.max( |
|
|
document.documentElement.scrollWidth, |
|
|
document.documentElement.offsetWidth, |
|
|
document.documentElement.clientWidth, |
|
|
document.body ? document.body.scrollWidth : 0, |
|
|
document.body ? document.body.offsetWidth : 0, |
|
|
document.body ? document.body.clientWidth : 0 |
|
|
), |
|
|
height: Math.max( |
|
|
document.documentElement.scrollHeight, |
|
|
document.documentElement.offsetHeight, |
|
|
document.documentElement.clientHeight, |
|
|
document.body ? document.body.scrollHeight : 0, |
|
|
document.body ? document.body.offsetHeight : 0, |
|
|
document.body ? document.body.clientHeight : 0 |
|
|
) |
|
|
}; |
|
|
""") |
|
|
scroll_width = dimensions['width'] |
|
|
scroll_height = dimensions['height'] |
|
|
logger.info(f"Detected dimensions: width={scroll_width}, height={scroll_height}") |
|
|
|
|
|
|
|
|
scroll_width = max(scroll_width, 100) |
|
|
scroll_height = max(scroll_height, 100) |
|
|
scroll_width = min(scroll_width, 2000) |
|
|
scroll_height = min(scroll_height, 4000) |
|
|
|
|
|
|
|
|
time.sleep(2.0) |
|
|
|
|
|
|
|
|
adjusted_height = int(scroll_height * (1 + extension_percentage / 100.0)) |
|
|
adjusted_height = max(adjusted_height, scroll_height, 100) |
|
|
|
|
|
|
|
|
adjusted_width = scroll_width |
|
|
logger.info(f"Resizing window to: width={adjusted_width}, height={adjusted_height}") |
|
|
driver.set_window_size(adjusted_width, adjusted_height) |
|
|
time.sleep(0.5) |
|
|
|
|
|
|
|
|
logger.info("Taking screenshot...") |
|
|
png = driver.get_screenshot_as_png() |
|
|
logger.info("Screenshot taken successfully.") |
|
|
|
|
|
|
|
|
img = Image.open(BytesIO(png)) |
|
|
logger.info(f"Screenshot dimensions: {img.width}x{img.height}") |
|
|
|
|
|
|
|
|
if trim_whitespace: |
|
|
img = trim_image_whitespace(img, threshold=248, padding=20) |
|
|
logger.info(f"Trimmed dimensions: {img.width}x{img.height}") |
|
|
|
|
|
return img |
|
|
|
|
|
except Exception as e: |
|
|
logger.error(f"Error during screenshot generation: {e}", exc_info=True) |
|
|
|
|
|
return Image.new('RGB', (1, 1), color=(0, 0, 0)) |
|
|
finally: |
|
|
logger.info("Cleaning up...") |
|
|
|
|
|
if driver_from_pool: |
|
|
driver_pool.release_driver(driver) |
|
|
logger.info("Returned driver to pool") |
|
|
|
|
|
if tmp_path and os.path.exists(tmp_path): |
|
|
try: |
|
|
os.remove(tmp_path) |
|
|
logger.info(f"Temporary file {tmp_path} removed.") |
|
|
except Exception as e: |
|
|
logger.error(f"Error removing temporary file {tmp_path}: {e}") |
|
|
|
|
|
|
|
|
def text_to_screenshot_parallel(text: str, extension_percentage: float, temperature: float = 0.3, |
|
|
trim_whitespace: bool = True, style: str = "standard") -> tuple: |
|
|
""" |
|
|
テキストをGemini APIでHTMLに変換し、並列処理でスクリーンショットを生成する関数 |
|
|
|
|
|
Returns: |
|
|
tuple - (PIL.Image, URL) - 生成された画像とHuggingFaceのURL |
|
|
""" |
|
|
start_time = time.time() |
|
|
logger.info("並列処理によるテキスト→スクリーンショット生成を開始") |
|
|
|
|
|
try: |
|
|
|
|
|
with ThreadPoolExecutor(max_workers=2) as executor: |
|
|
|
|
|
html_future = executor.submit( |
|
|
generate_html_from_text, |
|
|
text=text, |
|
|
temperature=temperature, |
|
|
style=style |
|
|
) |
|
|
|
|
|
|
|
|
driver_future = executor.submit(driver_pool.get_driver) |
|
|
|
|
|
|
|
|
html_code = html_future.result() |
|
|
driver = driver_future.result() |
|
|
|
|
|
|
|
|
driver_from_pool = True |
|
|
|
|
|
|
|
|
logger.info(f"HTML生成完了:{len(html_code)}文字。スクリーンショット生成開始。") |
|
|
|
|
|
|
|
|
tmp_path = None |
|
|
try: |
|
|
|
|
|
with tempfile.NamedTemporaryFile(suffix=".html", delete=False, mode='w', encoding='utf-8') as tmp_file: |
|
|
tmp_path = tmp_file.name |
|
|
tmp_file.write(html_code) |
|
|
logger.info(f"HTMLを一時ファイルに保存: {tmp_path}") |
|
|
|
|
|
|
|
|
initial_width = 1200 |
|
|
initial_height = 1000 |
|
|
driver.set_window_size(initial_width, initial_height) |
|
|
file_url = "file://" + tmp_path |
|
|
logger.info(f"ページに移動: {file_url}") |
|
|
driver.get(file_url) |
|
|
|
|
|
|
|
|
|
|
|
logger.info("body要素を待機...") |
|
|
WebDriverWait(driver, 10).until( |
|
|
EC.presence_of_element_located((By.TAG_NAME, "body")) |
|
|
) |
|
|
logger.info("body要素を検出。リソース読み込みを待機...") |
|
|
|
|
|
|
|
|
max_wait = 3 |
|
|
wait_increment = 0.2 |
|
|
wait_time = 0 |
|
|
|
|
|
while wait_time < max_wait: |
|
|
resource_state = driver.execute_script(""" |
|
|
return { |
|
|
complete: document.readyState === 'complete', |
|
|
imgCount: document.images.length, |
|
|
imgLoaded: Array.from(document.images).filter(img => img.complete).length, |
|
|
faElements: document.querySelectorAll('.fa, .fas, .far, .fab, [class*="fa-"]').length |
|
|
}; |
|
|
""") |
|
|
|
|
|
|
|
|
if resource_state['complete'] and (resource_state['imgCount'] == 0 or |
|
|
resource_state['imgLoaded'] == resource_state['imgCount']): |
|
|
logger.info(f"リソース読み込み完了: {resource_state}") |
|
|
break |
|
|
|
|
|
time.sleep(wait_increment) |
|
|
wait_time += wait_increment |
|
|
|
|
|
|
|
|
fa_count = resource_state.get('faElements', 0) |
|
|
if fa_count > 30: |
|
|
logger.info(f"{fa_count}個のFont Awesome要素があるため、追加待機...") |
|
|
time.sleep(min(1.0, fa_count / 100)) |
|
|
|
|
|
|
|
|
driver.execute_script("window.scrollTo(0, document.body.scrollHeight);") |
|
|
time.sleep(0.2) |
|
|
driver.execute_script("window.scrollTo(0, 0);") |
|
|
time.sleep(0.2) |
|
|
|
|
|
|
|
|
driver.execute_script(""" |
|
|
document.documentElement.style.overflow = 'hidden'; |
|
|
document.body.style.overflow = 'hidden'; |
|
|
""") |
|
|
|
|
|
|
|
|
dimensions = driver.execute_script(""" |
|
|
return { |
|
|
width: Math.max( |
|
|
document.documentElement.scrollWidth, |
|
|
document.documentElement.offsetWidth, |
|
|
document.documentElement.clientWidth, |
|
|
document.body ? document.body.scrollWidth : 0, |
|
|
document.body ? document.body.offsetWidth : 0, |
|
|
document.body ? document.body.clientWidth : 0 |
|
|
), |
|
|
height: Math.max( |
|
|
document.documentElement.scrollHeight, |
|
|
document.documentElement.offsetHeight, |
|
|
document.documentElement.clientHeight, |
|
|
document.body ? document.body.scrollHeight : 0, |
|
|
document.body ? document.body.offsetHeight : 0, |
|
|
document.body ? document.body.clientHeight : 0 |
|
|
) |
|
|
}; |
|
|
""") |
|
|
scroll_width = dimensions['width'] |
|
|
scroll_height = dimensions['height'] |
|
|
|
|
|
|
|
|
scroll_width = max(scroll_width, 100) |
|
|
scroll_height = max(scroll_height, 100) |
|
|
scroll_width = min(scroll_width, 2000) |
|
|
scroll_height = min(scroll_height, 4000) |
|
|
|
|
|
|
|
|
adjusted_height = int(scroll_height * (1 + extension_percentage / 100.0)) |
|
|
adjusted_height = max(adjusted_height, scroll_height, 100) |
|
|
|
|
|
|
|
|
driver.set_window_size(scroll_width, adjusted_height) |
|
|
time.sleep(0.2) |
|
|
|
|
|
|
|
|
logger.info("スクリーンショットを撮影...") |
|
|
png = driver.get_screenshot_as_png() |
|
|
|
|
|
|
|
|
img = Image.open(BytesIO(png)) |
|
|
logger.info(f"スクリーンショットサイズ: {img.width}x{img.height}") |
|
|
|
|
|
|
|
|
if trim_whitespace: |
|
|
img = trim_image_whitespace(img, threshold=248, padding=20) |
|
|
logger.info(f"トリミング後のサイズ: {img.width}x{img.height}") |
|
|
|
|
|
|
|
|
prefix = f"infographic_{style}" |
|
|
image_url = hf_uploader.upload_image(img, prefix=prefix) |
|
|
|
|
|
elapsed = time.time() - start_time |
|
|
logger.info(f"並列処理による生成完了。所要時間: {elapsed:.2f}秒、URL: {image_url}") |
|
|
return img, image_url |
|
|
|
|
|
except Exception as e: |
|
|
logger.error(f"スクリーンショット生成中にエラー: {e}", exc_info=True) |
|
|
return Image.new('RGB', (1, 1), color=(0, 0, 0)), None |
|
|
finally: |
|
|
|
|
|
if driver_from_pool: |
|
|
driver_pool.release_driver(driver) |
|
|
|
|
|
if tmp_path and os.path.exists(tmp_path): |
|
|
try: |
|
|
os.remove(tmp_path) |
|
|
except Exception as e: |
|
|
logger.error(f"一時ファイル削除エラー: {e}") |
|
|
|
|
|
except Exception as e: |
|
|
logger.error(f"並列処理中のエラー: {e}", exc_info=True) |
|
|
return Image.new('RGB', (1, 1), color=(0, 0, 0)), None |
|
|
|
|
|
|
|
|
def text_to_screenshot(text: str, extension_percentage: float, temperature: float = 0.3, |
|
|
trim_whitespace: bool = True, style: str = "standard") -> tuple: |
|
|
"""テキストをGemini APIでHTMLに変換し、スクリーンショットを生成する統合関数(レガシー版)""" |
|
|
|
|
|
return text_to_screenshot_parallel(text, extension_percentage, temperature, trim_whitespace, style) |
|
|
|
|
|
|
|
|
def render_and_upload_screenshot(html_code: str, extension_percentage: float = 10.0, |
|
|
trim_whitespace: bool = True, prefix: str = "screenshot") -> tuple: |
|
|
""" |
|
|
HTMLコードからスクリーンショットを生成し、HuggingFaceにアップロードする |
|
|
|
|
|
Returns: |
|
|
tuple - (PIL.Image, URL) - 生成された画像とHuggingFaceのURL |
|
|
""" |
|
|
try: |
|
|
|
|
|
img = render_fullpage_screenshot(html_code, extension_percentage, trim_whitespace) |
|
|
|
|
|
|
|
|
image_url = hf_uploader.upload_image(img, prefix=prefix) |
|
|
|
|
|
return img, image_url |
|
|
except Exception as e: |
|
|
logger.error(f"スクリーンショット生成とアップロード中にエラー: {e}", exc_info=True) |
|
|
return Image.new('RGB', (1, 1), color=(0, 0, 0)), None |
|
|
|
|
|
|
|
|
app = FastAPI() |
|
|
|
|
|
|
|
|
app.add_middleware( |
|
|
CORSMiddleware, |
|
|
allow_origins=["*"], |
|
|
allow_credentials=True, |
|
|
allow_methods=["*"], |
|
|
allow_headers=["*"], |
|
|
) |
|
|
|
|
|
|
|
|
|
|
|
gradio_dir = os.path.dirname(gr.__file__) |
|
|
logger.info(f"Gradio version: {gr.__version__}") |
|
|
logger.info(f"Gradio directory: {gradio_dir}") |
|
|
|
|
|
|
|
|
static_dir = os.path.join(gradio_dir, "templates", "frontend", "static") |
|
|
if os.path.exists(static_dir): |
|
|
logger.info(f"Mounting static directory: {static_dir}") |
|
|
app.mount("/static", StaticFiles(directory=static_dir), name="static") |
|
|
|
|
|
|
|
|
app_dir = os.path.join(gradio_dir, "templates", "frontend", "_app") |
|
|
if os.path.exists(app_dir): |
|
|
logger.info(f"Mounting _app directory: {app_dir}") |
|
|
app.mount("/_app", StaticFiles(directory=app_dir), name="_app") |
|
|
|
|
|
|
|
|
assets_dir = os.path.join(gradio_dir, "templates", "frontend", "assets") |
|
|
if os.path.exists(assets_dir): |
|
|
logger.info(f"Mounting assets directory: {assets_dir}") |
|
|
app.mount("/assets", StaticFiles(directory=assets_dir), name="assets") |
|
|
|
|
|
|
|
|
cdn_dir = os.path.join(gradio_dir, "templates", "cdn") |
|
|
if os.path.exists(cdn_dir): |
|
|
logger.info(f"Mounting cdn directory: {cdn_dir}") |
|
|
app.mount("/cdn", StaticFiles(directory=cdn_dir), name="cdn") |
|
|
|
|
|
|
|
|
|
|
|
@app.post("/api/screenshot", |
|
|
response_model=ImageUrlResponse, |
|
|
tags=["Screenshot"], |
|
|
summary="Render HTML to Full Page Screenshot and Upload to HuggingFace", |
|
|
description="Takes HTML code and an optional vertical extension percentage, renders it using a headless browser, uploads to HuggingFace, and returns the URL.") |
|
|
async def api_render_screenshot(request: ScreenshotRequest): |
|
|
""" |
|
|
API endpoint to render HTML, upload to HuggingFace, and return the URL. |
|
|
""" |
|
|
try: |
|
|
logger.info(f"API request received. Extension: {request.extension_percentage}%") |
|
|
|
|
|
|
|
|
pil_image, image_url = render_and_upload_screenshot( |
|
|
request.html_code, |
|
|
request.extension_percentage, |
|
|
request.trim_whitespace, |
|
|
prefix="screenshot" |
|
|
) |
|
|
|
|
|
if pil_image.size == (1, 1) or not image_url: |
|
|
logger.error("Screenshot generation failed, or upload failed.") |
|
|
raise HTTPException(status_code=500, detail="Failed to generate or upload screenshot") |
|
|
|
|
|
|
|
|
logger.info(f"返却URL: {image_url}") |
|
|
return ImageUrlResponse(url=image_url) |
|
|
|
|
|
except Exception as e: |
|
|
logger.error(f"API Error: {e}", exc_info=True) |
|
|
raise HTTPException(status_code=500, detail=f"Internal Server Error: {e}") |
|
|
|
|
|
|
|
|
@app.post("/api/text-to-screenshot", |
|
|
response_model=ImageUrlResponse, |
|
|
tags=["Screenshot", "Gemini"], |
|
|
summary="テキストからインフォグラフィックを生成しHuggingFaceにアップロード", |
|
|
description="テキストをGemini APIを使ってHTMLインフォグラフィックに変換し、HuggingFaceにアップロードしたURLを返します。") |
|
|
async def api_text_to_screenshot(request: GeminiRequest): |
|
|
""" |
|
|
テキストからHTMLインフォグラフィックを生成してアップロードし、URLを返すAPIエンドポイント |
|
|
""" |
|
|
try: |
|
|
logger.info(f"テキスト→スクリーンショットAPIリクエスト受信。テキスト長さ: {len(request.text)}, " |
|
|
f"拡張率: {request.extension_percentage}%, 温度: {request.temperature}, " |
|
|
f"スタイル: {request.style}") |
|
|
|
|
|
|
|
|
pil_image, image_url = text_to_screenshot_parallel( |
|
|
request.text, |
|
|
request.extension_percentage, |
|
|
request.temperature, |
|
|
request.trim_whitespace, |
|
|
request.style |
|
|
) |
|
|
|
|
|
if pil_image.size == (1, 1) or not image_url: |
|
|
logger.error("スクリーンショット生成に失敗したか、アップロードに失敗しました。") |
|
|
raise HTTPException(status_code=500, detail="Failed to generate or upload screenshot") |
|
|
|
|
|
|
|
|
logger.info(f"返却URL: {image_url}") |
|
|
return ImageUrlResponse(url=image_url) |
|
|
|
|
|
except Exception as e: |
|
|
logger.error(f"API Error: {e}", exc_info=True) |
|
|
raise HTTPException(status_code=500, detail=f"Internal Server Error: {e}") |
|
|
|
|
|
|
|
|
|
|
|
def process_input(input_mode, input_text, extension_percentage, temperature, trim_whitespace, style): |
|
|
"""入力モードに応じて適切な処理を行う""" |
|
|
if input_mode == "HTML入力": |
|
|
|
|
|
img, url = render_and_upload_screenshot( |
|
|
input_text, |
|
|
extension_percentage, |
|
|
trim_whitespace, |
|
|
prefix="html_screenshot" |
|
|
) |
|
|
return img, url if url else "アップロード失敗またはURL取得できませんでした" |
|
|
else: |
|
|
|
|
|
img, url = text_to_screenshot_parallel( |
|
|
input_text, |
|
|
extension_percentage, |
|
|
temperature, |
|
|
trim_whitespace, |
|
|
style |
|
|
) |
|
|
return img, url if url else "アップロード失敗またはURL取得できませんでした" |
|
|
|
|
|
|
|
|
with gr.Blocks(title="Full Page Screenshot (テキスト変換対応)", theme=gr.themes.Base()) as iface: |
|
|
gr.Markdown("# HTMLビューア & テキスト→インフォグラフィック変換") |
|
|
gr.Markdown("HTMLコードをレンダリングするか、テキストをGemini APIでインフォグラフィックに変換して画像として取得します。") |
|
|
gr.Markdown("**パフォーマンス向上版**: 並列処理と最適化により処理時間を短縮しています") |
|
|
|
|
|
with gr.Row(): |
|
|
input_mode = gr.Radio( |
|
|
["HTML入力", "テキスト入力"], |
|
|
label="入力モード", |
|
|
value="HTML入力" |
|
|
) |
|
|
|
|
|
|
|
|
input_text = gr.Textbox( |
|
|
lines=15, |
|
|
label="入力", |
|
|
placeholder="HTMLコードまたはテキストを入力してください。入力モードに応じて処理されます。" |
|
|
) |
|
|
|
|
|
with gr.Row(): |
|
|
with gr.Column(scale=1): |
|
|
|
|
|
style_dropdown = gr.Dropdown( |
|
|
choices=["standard", "cute", "resort", "cool", "dental", "school"], |
|
|
value="standard", |
|
|
label="デザインスタイル", |
|
|
info="テキスト→HTML変換時のデザインテーマを選択します", |
|
|
visible=False |
|
|
) |
|
|
|
|
|
with gr.Column(scale=2): |
|
|
extension_percentage = gr.Slider( |
|
|
minimum=0, |
|
|
maximum=30, |
|
|
step=1.0, |
|
|
value=10, |
|
|
label="上下高さ拡張率(%)" |
|
|
) |
|
|
|
|
|
|
|
|
temperature = gr.Slider( |
|
|
minimum=0.0, |
|
|
maximum=1.0, |
|
|
step=0.1, |
|
|
value=0.5, |
|
|
label="生成時の温度(低い=一貫性高、高い=創造性高)", |
|
|
visible=False |
|
|
) |
|
|
|
|
|
|
|
|
trim_whitespace = gr.Checkbox( |
|
|
label="余白を自動トリミング", |
|
|
value=True, |
|
|
info="生成される画像から余分な空白領域を自動的に削除します" |
|
|
) |
|
|
|
|
|
submit_btn = gr.Button("生成") |
|
|
|
|
|
|
|
|
with gr.Row(): |
|
|
with gr.Column(scale=1): |
|
|
output_image = gr.Image(type="pil", label="ページ全体のスクリーンショット") |
|
|
with gr.Column(scale=1): |
|
|
output_url = gr.Textbox(label="画像URL(HuggingFace)", info="生成された画像のURLです。このURLを使用して画像にアクセスできます。") |
|
|
|
|
|
|
|
|
def update_controls_visibility(mode): |
|
|
|
|
|
is_text_mode = mode == "テキスト入力" |
|
|
return [ |
|
|
gr.update(visible=is_text_mode), |
|
|
gr.update(visible=is_text_mode), |
|
|
] |
|
|
|
|
|
input_mode.change( |
|
|
fn=update_controls_visibility, |
|
|
inputs=input_mode, |
|
|
outputs=[temperature, style_dropdown] |
|
|
) |
|
|
|
|
|
|
|
|
submit_btn.click( |
|
|
fn=process_input, |
|
|
inputs=[input_mode, input_text, extension_percentage, temperature, trim_whitespace, style_dropdown], |
|
|
outputs=[output_image, output_url] |
|
|
) |
|
|
|
|
|
|
|
|
gemini_model = os.environ.get("GEMINI_MODEL", "gemini-1.5-pro") |
|
|
hf_repo = os.environ.get("HF_REPO_ID", "tomo2chin2/SUPER_TENSAI_JIN") |
|
|
gr.Markdown(f""" |
|
|
## APIエンドポイント |
|
|
- `/api/screenshot` - HTMLコードからスクリーンショットを生成し、URLを返します |
|
|
- `/api/text-to-screenshot` - テキストからインフォグラフィックスクリーンショットを生成し、URLを返します |
|
|
|
|
|
## 設定情報 |
|
|
- 使用モデル: {gemini_model} (環境変数 GEMINI_MODEL で変更可能) |
|
|
- HuggingFaceリポジトリ: {hf_repo} (環境変数 HF_REPO_ID で変更可能) |
|
|
- WebDriverプール最大数: {driver_pool.max_drivers} (環境変数 MAX_WEBDRIVERS で変更可能) |
|
|
- 対応スタイル: standard, cute, resort, cool, dental |
|
|
""") |
|
|
|
|
|
|
|
|
app = gr.mount_gradio_app(app, iface, path="/") |
|
|
|
|
|
|
|
|
if __name__ == "__main__": |
|
|
import uvicorn |
|
|
logger.info("Starting Uvicorn server for local development...") |
|
|
uvicorn.run(app, host="0.0.0.0", port=7860) |
|
|
|
|
|
|
|
|
import atexit |
|
|
atexit.register(driver_pool.close_all) |