feat(monitor): 抖音 Web 接口客户端 —— 绕开爬虫子进程，直接发 HTTP

移植自 mac-agent-os 的 mediacrawler_adapter：不起子进程、不开页面，用浏览器里那份
登录态直接调抖音 Web 接口。产物键名照抄 store/douyin，所以 ingest 那条链路一个字不用改。

**目前能用的（真环境实测，非推断）**：

    profile/other : 200, 7075 字节  —— 博主主页指标（粉丝/获赞/作品数/昵称）
    aweme/detail  : 200, 45425 字节 —— 单条作品详情（含点赞/评论/收藏/分享）

**目前不能用的：作品列表 `aweme/post`。** 两个互相独立的原因：
  1. 这个接口被抖音单独升级成了真校验：不带 x-tt-argus 回 403「Uifid Not Found」，
     带上 dummy 值回 200 + **空 body**。也就是说「头在不在」骗得过，「真校验」过不了。
     同一套头打 profile/other 和 aweme/detail 都是通的 —— 抖音是挑着接口加保护的，
     挑中的恰好是「批量拉作品列表」这个最敏感的动作。
  2. 改走页面截获也不行：CDP 浏览器打开博主主页会落到「验证码中间页」（当天大量探测的
     代价，过几小时要重测）。

  所以现在的边界是：**已知作品的指标刷新能做，自动发现新作品做不了**。

**排查中控住变量后得到的两条事实**（都写进注释了）：
  · `Accept` / `Accept-Language` / `Referer` 才是主页接口能返回真数据的原因 —— 只有
    UA+client hints+Cookie 时是 200 但仅 121 字节的空壳，补上这三个头变 7074 字节。
    （我先前猜的 sec-ch-ua 不是关键。）
  · 因此 UA 与 client hints 必须**成套地取自同一个浏览器**，所以 BrowserIdentity 一次
    从 CDP 取齐 cookie + UA + hints，而不是各自写死。

「200 + 空 body 必须当场报错」也是刻意写死的：放过去它会在下游变成「这个博主没作品」，
把一次失败伪装成一条正常结果 —— 爬虫那条路正是这么栽的，还被翻译成「账号被封」。

测试 +11：cookie 解析、请求头成套性（含 uifid 缺失/回退）、产物键名与 store 对齐、
以及 _get 的三条失败路径（空 body / 403 带网关原话 / 正常返回）。

mac-agent-os-main/05_tools/09_ave/scripts/material_producer/pexels/__init__.py

@@ -0,0 +1,5 @@
+"""AVE 素材搜索模块 — Pexels"""
+
+from .search import search_videos
+
+__all__ = ["search_videos"]

mac-agent-os-main/05_tools/09_ave/scripts/material_producer/pexels/search.py

@@ -0,0 +1,337 @@
+"""
+
+AVE 05_material_producer/pexels — Pexels 素材搜索 v1.1
+
+API 文档: https://www.pexels.com/api/documentation/
+版本: v1.1 | 更新: 2026-05-06
+
+改进:
+  - 中文→英文关键词自动映射 (70+ 常用词)
+  - 多关键词回落搜索 (最多 3 轮)
+  - portrait/landscape 自适应筛选
+  - 优先选取竖屏高质量视频
+
+用法:
+  search_videos("sunset beach", count=3, api_key="xxx")
+    → 返回下载后的本地文件路径列表
+
+回落策略:
+  1. 精确搜索 (按 query)
+  2. 简化关键词搜索 (去掉形容词/场景词)
+  3. 核心词搜索 (保留名词/动词)
+  4. 返回空列表 (由 caller 用 fallback 兜底)
+"""
+import sys, os; sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
+import hashlib
+import os
+import time
+from pathlib import Path
+from typing import Optional
+
+import httpx
+
+from lib.logger import get_logger
+
+logger = get_logger("pexels")
+
+PEXELS_BASE = "https://api.pexels.com/videos"
+# ⚠️ Pexels 强制要求 User-Agent：缺失会直接 403（error code 1010）
+PEXELS_UA = "AgentOS-AVE/1.0 (macOS; +https://github.com/neowoodland-art/mac-agent-os)"
+CACHE_DIR = (
+    Path(os.environ.get("AVE_CACHE_DIR",
+        str(Path.home() / "workbuddy-agent-os/agent-local/tools/ave/cache/materials")))
+)
+
+# ── 中文→英文关键词映射 ────────────────────────────────────
+ZH_TO_EN = {
+    "猫": "cat", "狗狗": "dog", "狗": "dog", "猫咪": "cat",
+    "日出": "sunrise", "日落": "sunset",
+    "大海": "ocean", "沙滩": "beach", "海滩": "beach",
+    "森林": "forest", "树木": "trees", "绿叶": "green leaves",
+    "河流": "river", "湖泊": "lake", "山川": "mountains",
+    "星空": "night sky stars", "银河": "milky way",
+    "城市": "city", "夜景": "city night", "街道": "city street",
+    "雪山": "snow mountain", "云海": "clouds mountain",
+    "瀑布": "waterfall", "彩虹": "rainbow",
+    "花海": "flower field", "樱花": "cherry blossoms",
+    "草原": "meadow", "草地": "grass field",
+    "海浪": "ocean waves", "雨滴": "rain drops",
+    "美食": "food cooking", "咖啡": "coffee cafe",
+    "书本": "books reading", "写字": "writing pen",
+    "瑜伽": "yoga meditation", "冥想": "meditation calm",
+    "跑步": "running jogging", "健身": "fitness gym",
+    "舞蹈": "dancing", "弹琴": "piano playing",
+    "茶道": "tea ceremony", "书法": "calligraphy",
+    "太极": "tai chi", "武术": "martial arts",
+    "烟花": "fireworks", "灯笼": "lanterns",
+    "古镇": "traditional town", "寺庙": "temple",
+    "海岛": "tropical island", "椰子树": "palm tree beach",
+    "汽车": "driving car", "火车": "train railway",
+    "飞机": "airplane sky", "轮船": "sailing ship",
+    "建筑": "architecture building", "桥梁": "bridge",
+    "天空": "sky clouds", "白云": "white clouds",
+    "下雨": "rain falling", "下雪": "snow falling",
+    "春天": "spring nature", "夏天": "summer sunny",
+    "秋天": "autumn leaves", "冬天": "winter snow",
+    "工作": "office work", "开会": "meeting business",
+    "微笑": "smiling happy", "牵手": "couple holding hands",
+    "小孩": "child playing", "老人": "elderly portrait",
+    "情侣": "couple romantic", "家庭": "family together",
+    "宠物": "pet dog cat", "小鸟": "bird flying",
+    "金鱼": "fish aquarium", "蝴蝶": "butterfly",
+    "月亮": "moon night", "阳光": "sunlight",
+}
+
+
+def _translate_query(query: str) -> str:
+    """将中文关键词翻译为英文（Pexels 英文搜索更精准）
+
+    策略：① 静态词典（快、零成本）→ ② 词典未完全覆盖时用 LLM 翻译（DeepSeek，成本极低）
+    """
+    q = (query or "").strip()
+    if not q:
+        return ""
+    has_cjk = any("\u4e00" <= c <= "\u9fff" for c in q)
+    if not has_cjk:
+        return q                      # 已是英文，直接用
+
+    # ① 静态词典替换
+    result = q
+    for zh, en in ZH_TO_EN.items():
+        if zh in result:
+            result = result.replace(zh, en)
+    if not any("\u4e00" <= c <= "\u9fff" for c in result):
+        return result.strip()         # 词典已完全覆盖
+
+    # ② 词典不够 → LLM 翻译（任意中文都能翻）
+    try:
+        import sys as _sys
+        from pathlib import Path as _P
+        _scripts = _P(__file__).resolve().parents[2]
+        if str(_scripts) not in _sys.path:
+            _sys.path.insert(0, str(_scripts))
+        from directors.llm import DirectorLLM
+        llm = DirectorLLM(timeout=25)
+        if llm.available:
+            en = llm.chat(
+                f"把这个中文搜索关键词翻译成最简短的英文（只输出英文，2~4 个词，不要解释）：{q}",
+                temperature=0.2, max_tokens=30)
+            en = (en or "").strip().strip('"\'').replace("\n", " ").strip()
+            if en and not any("\u4e00" <= c <= "\u9fff" for c in en):
+                logger.info(f"  🌐 LLM 翻译: {q!r} → {en!r}")
+                return en
+    except Exception as e:
+        logger.debug(f"LLM 翻译失败（回退词典结果）: {e}")
+    logger.warning(f"  ⚠️ 关键词可能仍是中文（{result!r}），Pexels 命中率会偏低")
+    return result.strip()
+
+
+def search_videos(
+    query: str,
+    count: int = 3,
+    api_key: str = "",
+    min_duration: int = 5,
+    orientation: str = "portrait",
+) -> list[dict]:
+    """
+    搜索并下载 Pexels 视频素材 v1.1
+
+    改进:
+      - 自动将中文关键词翻译为英文再搜索
+      - 3轮回落搜索策略，提升命中率
+      - portrait 模式优先竖屏高质量视频
+
+    参数:
+      query: 搜索关键词 (支持中文，自动翻译)
+      count: 需要的视频数量
+      api_key: Pexels API Key
+      min_duration: 最短视频时长(秒)
+      orientation: portrait(竖屏) | landscape(横屏) | square
+
+    返回:
+      [{"path": "/local/file.mp4", "duration": 15.0, "width": 1080, "height": 1920, "url": "原始链接"}, ...]
+    """
+    if not api_key:
+        logger.error("Pexels API Key 未配置")
+        return []
+
+    CACHE_DIR.mkdir(parents=True, exist_ok=True)
+
+    # 翻译关键词
+    en_query = _translate_query(query)
+    logger.info(f"搜索关键词: {query!r} → {en_query!r}")
+
+    # 多轮回落搜索
+    keywords_to_try = _generate_fallback_keywords(en_query)[:3]     # 最多 3 轮（避免长时间卡住）
+    for kw in keywords_to_try:
+        videos = _search_api(kw, count * 2, api_key, orientation)
+        if videos:
+            logger.info(f"  ✅ 命中关键词: {kw} ({len(videos)} 个候选)")
+            break
+        logger.warning(f"  ⏳ 关键词无结果，回落: {kw}")
+    else:
+        logger.warning(f"Pexels 所有关键词均无结果: {query}")
+        return []
+
+    # 过滤 + 选取（先选出候选，再**并行下载** —— 串行会慢 N 倍，Pexels 大文件尤其明显）
+    picks = []
+    for vid in videos:
+        if len(picks) >= count:
+            break
+        duration = vid.get("duration", 0)
+        if duration < min_duration:
+            continue
+
+        # 取最合适的视频文件
+        video_files = vid.get("video_files", [])
+        best = _pick_best_file(video_files, orientation)
+        if not best:
+            continue
+        picks.append((vid, best, duration))
+
+    results = []
+    if picks:
+        from concurrent.futures import ThreadPoolExecutor, as_completed
+        with ThreadPoolExecutor(max_workers=min(4, len(picks))) as ex:
+            futs = {ex.submit(_download, p[1]["link"], api_key, kw): p for p in picks}
+            for fut in as_completed(futs):
+                vid, best, duration = futs[fut]
+                try:
+                    local_path = fut.result()
+                except Exception as e:
+                    logger.warning("下载异常: %s", e)
+                    local_path = None
+                if local_path:
+                    results.append({
+                        "path": local_path,
+                        "duration": duration,
+                        "width": best.get("width", 0),
+                        "height": best.get("height", 0),
+                        "url": vid.get("url", ""),
+                    })
+
+    logger.info(f"Pexels 下载完成: {len(results)}/{count} 个 (原始:{query})")
+    return results
+
+
+def _generate_fallback_keywords(query: str) -> list[str]:
+    """生成回落关键词列表，最多3轮"""
+    keywords = [query]
+
+    # 第1轮回落: 去掉形容词和场景词，保留核心名词
+    # 例如 "sunset beach calm" → "beach sunset" / "beach"
+    words = query.split()
+    # 去掉常见形容词
+    adj_strip = [w for w in words if w.lower() not in (
+        "calm", "peaceful", "beautiful", "lovely", "nice", "serene",
+        "tranquil", "golden", "warm", "cool", "soft", "bright",
+        "dramatic", "mystic", "dreamy", "ethereal", "vibrant",
+    )]
+    if len(adj_strip) >= 2 and len(keywords) < 2:
+        keywords.append(" ".join(adj_strip))
+    elif len(adj_strip) >= 1 and len(keywords) < 2:
+        keywords.append(adj_strip[0])
+
+    # 第2轮回落: 只保留核心名词
+    core = [w for w in words if w.lower() not in (
+        "calm", "peaceful", "beautiful", "lovely", "nice", "serene",
+        "tranquil", "golden", "warm", "cool", "soft", "bright",
+        "dramatic", "mystic", "dreamy", "ethereal", "vibrant",
+        "sky", "view", "scene", "footage", "video", "clip",
+    )]
+    if core and len(keywords) < 3:
+        keywords.append(core[0])
+
+    return keywords[:3]
+
+
+def _search_api(
+    query: str, per_page: int, api_key: str, orientation: str
+) -> list[dict]:
+    """调用 Pexels 搜索 API"""
+    params = {
+        "query": query,
+        "per_page": min(per_page, 80),
+        "orientation": orientation,
+        "size": "large",
+    }
+    try:
+        resp = httpx.get(
+            f"{PEXELS_BASE}/search",
+            params=params,
+            headers={"Authorization": api_key, "User-Agent": PEXELS_UA},
+            timeout=15,
+            # 绕过系统代理（SOCKS5 可能导致 SSL 连接失败）
+            trust_env=False,
+        )
+        resp.raise_for_status()
+        data = resp.json()
+        return data.get("videos", [])
+    except httpx.HTTPStatusError as e:
+        if e.response.status_code == 429:
+            logger.warning("Pexels 速率限制，等待 1 秒重试")
+            time.sleep(1)
+            return _search_api(query, per_page, api_key, orientation)
+        logger.error(f"Pexels 搜索失败 [{e.response.status_code}]: {query}")
+        return []
+    except Exception as e:
+        logger.error(f"Pexels 搜索异常: {e}")
+        return []
+
+
+def _pick_best_file(video_files: list[dict], orientation: str) -> Optional[dict]:
+    """选择最佳视频文件 v1.1
+
+    优先策略:
+      portrait: 竖屏(9:16) > 接近竖屏 > 横屏中最高质量
+      竖屏判定: height > width 且 height >= 720
+    """
+    if not video_files:
+        return None
+
+    def quality_score(f):
+        w = f.get("width", 0)
+        h = f.get("height", 0)
+        file_size = f.get("file_size", 0)
+        fps = f.get("fps", 30)
+
+        # 竖屏优先 (height > width 且 h >= 720)
+        if orientation == "portrait" and h > w and h >= 720:
+            # **优选 ~1920 高（1080x1920 标准竖屏）**：够用且下载快；
+            # 4K(2160x3840) 文件 13MB+ 下载慢，扣分
+            return 100000 - abs(h - 1920) * 10 + min(fps, 30) - (file_size // 1024 // 1024) // 10
+        elif orientation == "landscape" and w > h and w >= 1280:
+            return 100000 - abs(w - 1920) * 10 + min(fps, 30)
+        elif orientation == "square" and abs(w - h) < max(w, h) * 0.2:
+            return 100000 - abs(min(w, h) - 1080) * 10
+        else:
+            # 无匹配时选最接近 1080p 的
+            return 50000 - abs(max(w, h) - 1920) * 5
+
+    video_files.sort(key=quality_score, reverse=True)
+    chosen = video_files[0]
+    logger.debug(f"  选片: {chosen.get('width')}x{chosen.get('height')} @ {chosen.get('fps')}fps")
+    return chosen
+
+
+def _download(url: str, api_key: str, query: str) -> Optional[str]:
+    """下载视频文件到本地缓存"""
+    # 缓存键: URL 的 hash
+    cache_key = hashlib.md5(url.encode()).hexdigest()[:16]
+    ext = ".mp4"
+    local_path = str(CACHE_DIR / f"{cache_key}{ext}")
+
+    if os.path.exists(local_path):
+        logger.debug(f"缓存命中: {local_path}")
+        return local_path
+
+    try:
+        resp = httpx.get(url, headers={"Authorization": api_key, "User-Agent": PEXELS_UA}, timeout=60, follow_redirects=True, trust_env=False)
+        resp.raise_for_status()
+        with open(local_path, "wb") as f:
+            f.write(resp.content)
+        logger.info(f"下载完成: {local_path} ({len(resp.content)//1024}KB)")
+        return local_path
+    except Exception as e:
+        logger.warning(f"下载失败: {e}")
+        return None