Ship native iOS app, Wiki/TTS backend, and skeleton loading UX.

Replace the WebView shell with SwiftUI screens, add account-scoped Wiki and TTS APIs with adaptive review and photo scan support, and keep web/iOS pages usable while data loads asynchronously.

Co-authored-by: Cursor <cursoragent@cursor.com>

backend/scripts/test_photo_scan_accuracy.py

@@ -0,0 +1,158 @@
+"""测试拍照录入在课本词表照片上的识别成功率。"""
+from __future__ import annotations
+
+import os
+import re
+import subprocess
+import sys
+from pathlib import Path
+
+BACKEND_DIR = Path(__file__).resolve().parent.parent
+sys.path.insert(0, str(BACKEND_DIR))
+
+from services.word_scan_service import word_scan_service  # noqa: E402
+
+IMAGE = Path(
+    "/Users/john/.cursor/projects/Users-john-PycharmProjects-wordloop/assets/"
+    "bf2e15b4e1a36836efb4e9a72b200d8e-f357a710-6d8d-4f5b-a52c-0bea4420aada.png"
+)
+
+GROUND_TRUTH: list[tuple[str, str]] = [
+    ("feed the cows", "喂奶牛"),
+    ("see a film", "看电影"),
+    ("visit Grandma and Grandpa", "看望爷爷奶奶"),
+    ("have a good time", "过得愉快"),
+    ("do the housework", "做家务"),
+    ("tidy my room", "整理我的房间"),
+    ("make the bed", "铺床"),
+    ("dish", "碟；盘"),
+    ("happy", "高兴的"),
+    ("excited", "激动的；兴奋的"),
+    ("upset", "难过的"),
+    ("angry", "愤怒的；生气的"),
+    ("play music", "放音乐"),
+    ("sing and dance", "唱歌跳舞"),
+    ("talk with the baby", "和婴儿说话"),
+    ("play a game", "玩游戏"),
+    ("share", "分享"),
+    ("have a go", "尝试"),
+    ("nice", "令人愉快的；吸引人的"),
+    ("cook", "厨师"),
+    ("great", "伟大的"),
+    ("teacher", "教师"),
+    ("tree", "树"),
+    ("leaf (pl. leaves)", "叶子"),
+    ("flower", "花"),
+    ("grass", "草"),
+    ("duck", "鸭"),
+    ("pond", "池塘；水池"),
+    ("frog", "青蛙"),
+    ("lake", "湖"),
+    ("fish", "鱼"),
+    ("river", "河；江"),
+    ("shark", "鲨鱼"),
+    ("sea", "海；海洋"),
+    ("wash my hands", "(我)洗手"),
+    ("before eating", "吃东西前"),
+    ("after using the toilet", "上厕所后"),
+    ("dirty", "脏的"),
+    ("clean", "干净的"),
+    ("drink some milk", "喝一些牛奶"),
+    ("say good night", "道晚安"),
+    ("sleep", "睡觉；入睡"),
+    ("read a story", "读故事"),
+]
+
+
+def normalize_en(text: str) -> str:
+    return re.sub(r"\s+", " ", text.strip().lower())
+
+
+def normalize_zh(text: str) -> str:
+    return re.sub(r"\s+", "", text.strip())
+
+
+def fuzzy_zh_match(pred: str, truth: str) -> bool:
+    p = normalize_zh(pred)
+    t = normalize_zh(truth)
+    if not p or not t:
+        return False
+    if p == t:
+        return True
+    # 允许 OCR/截断导致的中文释义部分匹配（>=70% 字符重合）
+    common = sum(1 for ch in t if ch in p)
+    return common / max(len(t), 1) >= 0.7
+
+
+def fuzzy_en_match(pred: str, truth: str) -> bool:
+    p = normalize_en(pred)
+    t = normalize_en(truth)
+    if p == t:
+        return True
+    # 短语轻微 OCR 错误：去掉空格后前缀匹配
+    p2 = re.sub(r"[^a-z()\.]", "", p)
+    t2 = re.sub(r"[^a-z()\.]", "", t)
+    if not p2 or not t2:
+        return False
+    return t2 in p2 or p2 in t2 or p2.startswith(t2[: max(4, len(t2) - 2)])
+
+
+def run_tesseract(image_path: Path, psm: str = "6") -> str:
+    result = subprocess.run(
+        ["tesseract", str(image_path), "stdout", "-l", "chi_sim+eng", "--psm", psm],
+        capture_output=True,
+        text=True,
+        check=False,
+    )
+    return result.stdout or ""
+
+
+def evaluate(items: list[dict[str, str]], label: str) -> dict:
+    preds = [(i["source_text"], i["target_text"]) for i in items]
+    matched = []
+    missed = []
+    for en, zh in GROUND_TRUTH:
+        hit = any(fuzzy_en_match(p_en, en) and fuzzy_zh_match(p_zh, zh) for p_en, p_zh in preds)
+        if hit:
+            matched.append((en, zh))
+        else:
+            missed.append((en, zh))
+
+    recall = len(matched) / len(GROUND_TRUTH)
+    precision = len(matched) / max(len(preds), 1)
+    print(f"\n=== {label} ===")
+    print(f"标准词条: {len(GROUND_TRUTH)}")
+    print(f"识别词条: {len(preds)}")
+    print(f"命中: {len(matched)}  漏识: {len(missed)}")
+    print(f"召回率(成功率): {recall * 100:.1f}%")
+    print(f"精确率: {precision * 100:.1f}%")
+    if missed:
+        print("漏识样例:", ", ".join(f"{en}" for en, _ in missed[:12]))
+        if len(missed) > 12:
+            print(f"... 另有 {len(missed) - 12} 条")
+    return {"recall": recall, "matched": len(matched), "pred_count": len(preds), "missed": missed}
+
+
+def main() -> None:
+    if not IMAGE.exists():
+        print(f"图片不存在: {IMAGE}")
+        sys.exit(1)
+
+    ocr_raw = run_tesseract(IMAGE)
+    print("--- OCR 原文（前 1200 字）---")
+    print(ocr_raw[:1200])
+
+    try:
+        items = word_scan_service.parse_scan_text(ocr_raw)
+        result = evaluate(items, "当前流水线: 原始 OCR + 解析")
+    except Exception as exc:
+        print(f"\n解析失败: {exc}")
+        result = {"recall": 0.0}
+
+    passed = result["recall"] >= 0.9
+    print(f"\n结论: {'达到' if passed else '未达到'} 90% 成功率目标（召回 {result['recall']*100:.1f}%）")
+    sys.exit(0 if passed else 1)
+
+
+if __name__ == "__main__":
+    main()