wechat-media-writer 2.2.4 → 2.2.5

package/SKILL.md

@@ -208,28 +208,21 @@ from image_downloader import download_theme_images
 urls = download_theme_images("books", "/tmp/images", count=6)
 ```
 
-### 第二步：制作公众号封面图
+### 第二步：下载公众号封面图（精确 900x500）
 
-公众号封面（`cover_900x500.jpg`）用主题贴图中的一张裁剪为 900x500 比例：
+**公众号封面必须是 900x500 精确尺寸的精美图片**，与正文主题风格统一。
+不再从主题贴图裁剪（裁剪可能损失画面关键内容）——直接从 Pexels CDN 请求 900x500 精确裁切。
+
+每个主题预存 3 张高质量 Pexels 候选图（见 `scripts/book_cover.py` 的 `THEME_COVER_CANDIDATES`），
+通过 `?w=900&h=500&fit=crop` 参数让 Pexels 服务端精确裁切为 900x500：
 
 ```python
-from PIL import Image
-img = Image.open("img_1.jpg")
-w, h = img.size
-target_ratio = 900 / 500
-if w / h > target_ratio:
-    new_w = int(h * target_ratio)
-    left = (w - new_w) // 2
-    img = img.crop((left, 0, left + new_w, h))
-else:
-    new_h = int(w / target_ratio)
-    top = (h - new_h) // 2
-    img = img.crop((0, top, w, top + new_h))
-img = img.resize((900, 500), Image.LANCZOS)
-img.save("cover_900x500.jpg", "JPEG", quality=85)
+from book_cover import download_cover_900x500
+# 下载一张 books 主题的 900x500 封面（seed=0/1/2 轮换候选图）
+download_cover_900x500("books", "/tmp/wx", seed=0)
 ```
 
-`scripts/book_cover.py` 的 `download_and_process_cover()` 函数会从主题贴图（`img_1.jpg`）裁剪生成 900x500 封面。
+**支持的主题**：`abstract` / `books` / `nature` / `technology` / `business`
 
 ### 第三步：上传图片到微信
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "wechat-media-writer",
-  "version": "2.2.4",
+  "version": "2.2.5",
   "description": "微信公众号书评、影评文章自动生成与发布 - Skill",
   "main": "index.js",
   "bin": {

package/scripts/book_cover.py

@@ -124,22 +124,65 @@ def fetch_book_cover(book_title, author="", isbn=""):
     return cover_url
 
 
-def fetch_theme_cover(theme="abstract"):
-    """从 Pexels CDN 获取与主题贴图一致的主题封面（推荐使用）"""
-    theme_urls = {
-        "abstract": "https://images.pexels.com/photos/2693208/pexels-photo-2693208.jpeg?auto=compress&cs=tinysrgb&w=1080",
-        "books": "https://images.pexels.com/photos/159711/books-bookstore-book-reading-159711.jpeg?auto=compress&cs=tinysrgb&w=1080",
-        "nature": "https://images.pexels.com/photos/2387873/pexels-photo-2387873.jpeg?auto=compress&cs=tinysrgb&w=1080",
-        "technology": "https://images.pexels.com/photos/3568520/pexels-photo-3568520.jpeg?auto=compress&cs=tinysrgb&w=1080",
-        "business": "https://images.pexels.com/photos/3184292/pexels-photo-3184292.jpeg?auto=compress&cs=tinysrgb&w=1080",
-    }
-    url = theme_urls.get(theme, theme_urls["abstract"])
-    print(f"✓ 主题封面URL ({theme}): {url}")
+# 每个主题对应 3 张精选的 900x500 Pexels 封面图（fit=crop 自动精确裁切）
+# 选用宽高比接近 1.8 的高质量原图，避免裁切损失重要内容
+THEME_COVER_CANDIDATES = {
+    "abstract": [1108572, 1762851, 3109808],
+    "books": [256450, 590493, 762687],
+    "nature": [2387873, 2387874, 2387876],
+    "technology": [3568520, 1181244, 546819],
+    "business": [3184292, 210607, 3184296],
+}
+
+
+def _build_cover_url(pid, w=900, h=500):
+    """构造 Pexels CDN URL，含精确 900x500 fit=crop 裁切参数"""
+    return f"https://images.pexels.com/photos/{pid}/pexels-photo-{pid}.jpeg?auto=compress&cs=tinysrgb&w={w}&h={h}&fit=crop"
+
+
+def fetch_theme_cover(theme="abstract", seed=0):
+    """从 Pexels CDN 获取一张与主题对应的 900x500 精确封面 URL（不裁剪原图）"""
+    candidates = THEME_COVER_CANDIDATES.get(theme, THEME_COVER_CANDIDATES["abstract"])
+    pid = candidates[seed % len(candidates)]
+    url = _build_cover_url(pid)
+    print(f"✓ 主题封面URL ({theme}, pid={pid}): {url}")
     return url
 
 
+def download_cover_900x500(theme="abstract", output_dir="/tmp", seed=0):
+    """下载一张 900x500 的主题封面图（Pexels 服务端精确裁剪，无需本地处理）
+
+    返回: cover_900x500.jpg 绝对路径
+    """
+    ctx = ssl.create_default_context()
+    ctx.check_hostname = False
+    ctx.verify_mode = ssl.CERT_NONE
+
+    os.makedirs(output_dir, exist_ok=True)
+    cover_900_path = os.path.join(output_dir, "cover_900x500.jpg")
+
+    candidates = THEME_COVER_CANDIDATES.get(theme, THEME_COVER_CANDIDATES["abstract"])
+    # 顺序尝试每个候选 PID，跳过 404
+    for i, pid in enumerate(candidates):
+        idx = (seed + i) % len(candidates)
+        pid_try = candidates[idx]
+        url = _build_cover_url(pid_try)
+        try:
+            req = urllib.request.Request(url, headers={"User-Agent": "Mozilla/5.0"})
+            data = urllib.request.urlopen(req, context=ctx, timeout=15).read()
+            with open(cover_900_path, "wb") as f:
+                f.write(data)
+            print(f"✓ 公众号封面已下载 ({theme}, pid={pid_try}): {cover_900_path}")
+            return cover_900_path
+        except Exception as e:
+            print(f"  候选 pid={pid_try} 失败: {e}")
+            continue
+
+    raise RuntimeError(f"主题 {theme} 的所有候选封面均下载失败")
+
+
 def download_and_process_cover(cover_url, output_dir):
-    """从主题贴图URL下载并处理为 900x500 公众号封面 + 600px 文章封面"""
+    """从主题贴图URL下载并处理为 600px 文章封面（兼容旧调用）"""
     from PIL import Image
 
     ctx = ssl.create_default_context()
@@ -154,66 +197,21 @@ def download_and_process_cover(cover_url, output_dir):
     with open(original_path, "wb") as f:
         f.write(data)
 
-    # 文章内配图（600px 宽）
     img = Image.open(original_path)
     w, h = img.size
     if w > 600:
         ratio = 600 / w
-        new_w = 600
-        new_h = int(h * ratio)
-        img = img.resize((new_w, new_h), Image.LANCZOS)
+        img = img.resize((600, int(h * ratio)), Image.LANCZOS)
     img.save(os.path.join(output_dir, "theme_cover_600.jpg"), "JPEG", quality=90)
 
-    # 公众号封面（900x500）
-    img2 = Image.open(original_path)
-    w, h = img2.size
-    target_ratio = 900 / 500
-    if w / h > target_ratio:
-        new_w = int(h * target_ratio)
-        left = (w - new_w) // 2
-        img2 = img2.crop((left, 0, left + new_w, h))
-    else:
-        new_h = int(w / target_ratio)
-        top = (h - new_h) // 2
-        img2 = img2.crop((0, top, w, top + new_h))
-    img2 = img2.resize((900, 500), Image.LANCZOS)
-    cover_900_path = os.path.join(output_dir, "cover_900x500.jpg")
-    img2.save(cover_900_path, "JPEG", quality=90)
-
-    print(f"✓ 主题封面已保存: theme_cover_600.jpg")
-    print(f"✓ 公众号封面已保存: {cover_900_path}")
-
-    return os.path.join(output_dir, "theme_cover_600.jpg"), cover_900_path
+    print(f"✓ 主题封面(600px)已保存: theme_cover_600.jpg")
+    return os.path.join(output_dir, "theme_cover_600.jpg"), None
 
 
 def crop_cover_from_local(source_path, output_dir):
-    """从本地主题贴图裁剪为 900x500 公众号封面（用于已有主题图但缺封面图的情况）"""
-    from PIL import Image
-
-    if not os.path.exists(source_path):
-        print(f"错误: 源图不存在 {source_path}", file=sys.stderr)
-        return None
-
-    os.makedirs(output_dir, exist_ok=True)
-    img = Image.open(source_path)
-    w, h = img.size
-    target_ratio = 900 / 500
-    if w / h > target_ratio:
-        new_w = int(h * target_ratio)
-        left = (w - new_w) // 2
-        img = img.crop((left, 0, left + new_w, h))
-    else:
-        new_h = int(w / target_ratio)
-        top = (h - new_h) // 2
-        img = img.crop((0, top, w, top + new_h))
-    img = img.resize((900, 500), Image.LANCZOS)
-
-    cover_900_path = os.path.join(output_dir, "cover_900x500.jpg")
-    img.save(cover_900_path, "JPEG", quality=90)
-    print(
-        f"✓ 公众号封面已生成（基于 {os.path.basename(source_path)}）: {cover_900_path}"
-    )
-    return cover_900_path
+    """兼容旧 API：建议改用 download_cover_900x500()"""
+    print("⚠️ crop_cover_from_local 已弃用，请改用 download_cover_900x500(theme=...)")
+    return None
 
 
 if __name__ == "__main__":

package/scripts/publish.py

@@ -12,7 +12,7 @@ import sys
 # 添加当前目录到模块搜索路径
 sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
 
-from book_cover import crop_cover_from_local, fetch_theme_cover
+from book_cover import download_cover_900x500
 from image_downloader import download_theme_images
 from wechat_api import (
     load_config,
@@ -42,30 +42,19 @@ def ensure_theme_images(cover_dir, theme="abstract", count=6):
     download_theme_images(theme, cover_dir, count)
 
 
-def ensure_cover_image(cover_dir, theme="abstract"):
-    """确保公众号封面（900x500）存在，缺失则从主题图裁剪"""
+def ensure_cover_image(cover_dir, theme="abstract", seed=0):
+    """确保公众号封面（900x500）存在，缺失则下载一张对应主题的精确尺寸封面
+
+    注意：不再从主题插图裁剪（裁剪可能损失画面重要内容），
+    而是从 Pexels 直接请求 900x500 精确尺寸的对应主题图片。
+    """
     cover_900_path = os.path.join(cover_dir, "cover_900x500.jpg")
     if os.path.exists(cover_900_path):
-        print("✓ 公众号封面已存在，跳过裁剪")
+        print("✓ 公众号封面已存在，跳过下载")
         return
 
-    # 优先用本地第一张主题图裁剪；否则从主题URL下载
-    candidates = sorted(
-        [
-            f
-            for f in os.listdir(cover_dir)
-            if f.startswith("img_") and f.endswith((".jpg", ".jpeg", ".png"))
-        ]
-    )
-    if candidates:
-        crop_cover_from_local(os.path.join(cover_dir, candidates[0]), cover_dir)
-    else:
-        # 极少见：连主题图都没有的情况
-        print("主题图与封面均缺失，从网络拉取主题封面...")
-        from book_cover import download_and_process_cover
-
-        url = fetch_theme_cover(theme)
-        download_and_process_cover(url, cover_dir)
+    print(f"封面图缺失，下载一张 {theme} 主题的 900x500 封面...")
+    download_cover_900x500(theme, cover_dir, seed=seed)
 
 
 def publish_article(
@@ -76,8 +65,9 @@ def publish_article(
     cover_dir,
     theme="abstract",
     image_count=6,
+    cover_seed=0,
 ):
-    """完整发布流程：准备主题贴图 -> 裁剪公众号封面 -> 上传图片 -> 创建草稿"""
+    """完整发布流程：准备主题贴图 -> 下载公众号封面 -> 上传图片 -> 创建草稿"""
     print("=" * 50)
     print("微信公众号 - 书评发布")
     print("=" * 50)
@@ -88,9 +78,9 @@ def publish_article(
     print("\n[1/4] 准备主题贴图...")
     ensure_theme_images(cover_dir, theme, image_count)
 
-    # 2. 准备公众号封面（900x500）
+    # 2. 准备公众号封面（900x500 精确尺寸）
     print("\n[2/4] 准备公众号封面...")
-    ensure_cover_image(cover_dir, theme)
+    ensure_cover_image(cover_dir, theme, seed=cover_seed)
 
     # 3. 加载配置和获取 token
     config = load_config()
@@ -158,6 +148,12 @@ def main():
     parser.add_argument(
         "--image-count", type=int, default=6, help="主题插图数量（默认6：5正文+1结尾）"
     )
+    parser.add_argument(
+        "--cover-seed",
+        type=int,
+        default=0,
+        help="封面图候选序号（0/1/2，相同主题下可轮换）",
+    )
     args = parser.parse_args()
 
     with open(args.content_file, "r", encoding="utf-8") as f:
@@ -171,6 +167,7 @@ def main():
         args.cover_dir,
         args.theme,
         args.image_count,
+        args.cover_seed,
     )