#!/usr/bin/env python3 """Translation and localized feed runtime for the website intel server.""" from __future__ import annotations import copy import hashlib import json import os import re from datetime import datetime, timezone from pathlib import Path from threading import Lock, Thread from x_intel_core import load_environment, minimax_chat, resolve_minimax_key ROOT = Path(__file__).resolve().parents[1] DATA_ROOT = ROOT / "data" FEED_PATH = DATA_ROOT / "x_intel_feed.json" I18N_FEED_PATH = DATA_ROOT / "x_intel_feed_i18n.json" TRANSLATION_CACHE_PATH = DATA_ROOT / "i18n_text_cache.json" TRANSLATION_LOCK = Lock() I18N_LOCK = Lock() I18N_STATE_LOCK = Lock() TRANSLATION_CACHE: dict[str, str] = {} TRANSLATION_CACHE_DIRTY = False # User decision: disable translation text cache to avoid stale/mixed-language reuse. TRANSLATION_CACHE_ENABLED = False I18N_BUILD_STATE: dict[str, object] = { "status": "idle", "started_at": "", "finished_at": "", "last_error": "", "source_generated_at": "", "target_langs": [], "langs": [], "lang_progress": {}, } TRANSLATE_MAX_CHARS = 320 I18N_TARGET_LANGS = ["zh-Hant", "zh-Hans", "en", "ko"] I18N_BUILD_VERSION = 8 I18N_FEED_TEXT_KEYS = { "headline", "conclusion", "title", "summary", "glance", "detail_summary", "point", "name", "for", "requirement", "label", "where", "who", "what", "why", "impact", "join", "schedule", "location", "audience", "reward", "message", "tags", } I18N_FEED_LIST_KEYS = { "takeaways", "bullets", "detail_lines", } I18N_MAX_TARGET_TEXTS = int(os.getenv("I18N_MAX_TARGET_TEXTS", "0") or "0") I18N_MAX_LIST_ITEMS_PER_FIELD = int(os.getenv("I18N_MAX_LIST_ITEMS_PER_FIELD", "0") or "0") I18N_MIN_ACCEPTABLE_COVERAGE = float(os.getenv("I18N_MIN_ACCEPTABLE_COVERAGE", "0.98") or "0.98") I18N_FEED_CHUNK_SIZE = max(3, int(os.getenv("I18N_FEED_CHUNK_SIZE", "12") or "12")) I18N_FEED_FALLBACK_MODE = str(os.getenv("I18N_FEED_FALLBACK_MODE", "base") or "base").strip().lower() I18N_SKIP_KEYS = { "id", "url", "account", "provider", "template_id", "card_type", "layout", "type", "urgency", "published_at", "timeline_date", "generated_at", "latest_source_at", } I18N_CARD_TEXT_KEYS = ( "title", "summary", "glance", "detail_summary", "headline", ) I18N_CARD_FACT_TEXT_KEYS = ( "participation", "schedule", "location", "audience", "reward", "message", "join", "where", "who", "what", "why", "impact", ) I18N_VISIBLE_FEED_ROOT_KEYS = ( "digest", "official_overview", "intel_sections", "intel_agenda", "format_templates", "key_terms", ) def configure_i18n_runtime(data_root: Path, feed_path: Path | None = None) -> None: """Point runtime caches at the mounted/active website data directory.""" global DATA_ROOT, FEED_PATH, I18N_FEED_PATH, TRANSLATION_CACHE_PATH DATA_ROOT = Path(data_root) FEED_PATH = Path(feed_path) if feed_path else DATA_ROOT / "x_intel_feed.json" I18N_FEED_PATH = DATA_ROOT / "x_intel_feed_i18n.json" TRANSLATION_CACHE_PATH = DATA_ROOT / "i18n_text_cache.json" def _now_iso() -> str: return datetime.now(timezone.utc).isoformat() def _normalize_lang_tag(lang: str | None) -> str: tag = str(lang or "zh-Hant").strip() normalized = tag.lower().replace("_", "-") if normalized in {"zh", "zh-tw", "zh-hant", "tw", "traditional", "繁中"}: return "zh-Hant" if normalized in {"zh-cn", "zh-hans", "cn", "simplified", "简中", "簡中"}: return "zh-Hans" if normalized in {"en", "en-us", "english"}: return "en" if normalized in {"ko", "ko-kr", "kr", "korean"}: return "ko" return tag if tag in {"zh-Hant", "zh-Hans", "en", "ko"} else "zh-Hant" def _safe_int(value: object, default: int = 0) -> int: try: return int(value) # type: ignore[arg-type] except Exception: return default def _read_feed_snapshot() -> dict: if not FEED_PATH.exists(): return {} try: data = json.loads(FEED_PATH.read_text(encoding="utf-8")) except Exception: return {} return data if isinstance(data, dict) else {} def _translation_cache_key_legacy(lang: str, text: str) -> str: return f"{_normalize_lang_tag(lang)}\n{text}" def _translation_cache_key(lang: str, text: str, entry_key: str = "") -> str: tag = _normalize_lang_tag(lang) source = str(text or "") scope = str(entry_key or "").strip() if not scope: return _translation_cache_key_legacy(tag, source) digest = hashlib.sha1(source.encode("utf-8")).hexdigest()[:16] return f"{tag}\n{scope}\n{digest}" def _contains_cjk(text: str) -> bool: return bool(re.search(r"[\u3400-\u9fff]", str(text or ""))) def _script_counts(text: str) -> tuple[int, int, int]: src = str(text or "") latin = len(re.findall(r"[A-Za-z]", src)) hangul = len(re.findall(r"[\uac00-\ud7a3]", src)) cjk = len(re.findall(r"[\u3400-\u9fff]", src)) return latin, hangul, cjk def _looks_translated_for_lang(text: str, lang: str) -> bool: src = str(text or "").strip() if not src: return True tag = _normalize_lang_tag(lang) latin, hangul, cjk = _script_counts(src) total = max(1, latin + hangul + cjk) if tag == "en": latin_ratio = latin / total cjk_ratio = cjk / total if cjk_ratio >= 0.05: return False if latin >= 8: return True if latin_ratio >= 0.35: return True return False if tag == "ko": hangul_ratio = hangul / total cjk_ratio = cjk / total if cjk_ratio >= 0.06: return False if hangul >= 6: return True if hangul_ratio >= 0.30: return True return False return True def _load_translation_cache_unlocked() -> None: global TRANSLATION_CACHE if not TRANSLATION_CACHE_ENABLED: TRANSLATION_CACHE = {} try: if TRANSLATION_CACHE_PATH.exists(): TRANSLATION_CACHE_PATH.unlink() except Exception: pass return if TRANSLATION_CACHE: return if not TRANSLATION_CACHE_PATH.exists(): TRANSLATION_CACHE = {} return try: raw = json.loads(TRANSLATION_CACHE_PATH.read_text(encoding="utf-8")) except Exception: TRANSLATION_CACHE = {} return if not isinstance(raw, dict): TRANSLATION_CACHE = {} return rows = raw.get("rows") if not isinstance(rows, dict): TRANSLATION_CACHE = {} return out: dict[str, str] = {} for key, value in rows.items(): k = str(key or "").strip() v = str(value or "").strip() if k and v: out[k] = v TRANSLATION_CACHE = out def _flush_translation_cache_unlocked(force: bool = False) -> None: global TRANSLATION_CACHE_DIRTY if not TRANSLATION_CACHE_ENABLED: TRANSLATION_CACHE_DIRTY = False return if not force and not TRANSLATION_CACHE_DIRTY: return TRANSLATION_CACHE_PATH.parent.mkdir(parents=True, exist_ok=True) payload = { "updated_at": _now_iso(), "rows": TRANSLATION_CACHE, } TRANSLATION_CACHE_PATH.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8") TRANSLATION_CACHE_DIRTY = False def _lang_prompt_name(tag: str) -> str: mapping = { "zh-Hant": "繁體中文", "zh-Hans": "简体中文", "en": "English", "ko": "한국어", } return mapping.get(_normalize_lang_tag(tag), "繁體中文") def _parse_json_array(raw: str) -> list[str]: text = str(raw or "").strip() if not text: return [] candidates = [text] block = re.search(r"```(?:json)?\s*(\[.*?\])\s*```", text, re.S) if block: candidates.insert(0, block.group(1)) bare = re.search(r"(\[.*\])", text, re.S) if bare: candidates.insert(0, bare.group(1)) for candidate in candidates: try: parsed = json.loads(candidate) except Exception: continue if isinstance(parsed, list): out = [] for item in parsed: out.append(str(item if item is not None else "").strip()) return out return [] ZH_HANS_TRANS = str.maketrans({ "體": "体", "萬": "万", "與": "与", "專": "专", "業": "业", "叢": "丛", "東": "东", "絲": "丝", "丟": "丢", "兩": "两", "嚴": "严", "喪": "丧", "個": "个", "豐": "丰", "臨": "临", "為": "为", "麗": "丽", "舉": "举", "麼": "么", "義": "义", "烏": "乌", "樂": "乐", "喬": "乔", "習": "习", "鄉": "乡", "書": "书", "買": "买", "亂": "乱", "爭": "争", "於": "于", "虧": "亏", "雲": "云", "亞": "亚", "產": "产", "畝": "亩", "親": "亲", "褻": "亵", "億": "亿", "僅": "仅", "從": "从", "倉": "仓", "儀": "仪", "們": "们", "價": "价", "眾": "众", "優": "优", "會": "会", "傘": "伞", "偉": "伟", "傳": "传", "傷": "伤", "倫": "伦", "偽": "伪", "佇": "伫", "佔": "占", "餘": "余", "佛": "佛", "來": "来", "侖": "仑", "侶": "侣", "俠": "侠", "係": "系", "俁": "俣", "倆": "俩", "倖": "幸", "倣": "仿", "傑": "杰", "傖": "伧", "傘": "伞", "備": "备", "傢": "家", "傭": "佣", "傯": "偬", "債": "债", "傷": "伤", "傾": "倾", "僂": "偻", "僑": "侨", "僕": "仆", "僥": "侥", "僨": "偾", "價": "价", "儂": "侬", "億": "亿", "儈": "侩", "儉": "俭", "儐": "傧", "儔": "俦", "儕": "侪", "儘": "尽", "償": "偿", "儲": "储", "兒": "儿", "兌": "兑", "內": "内", "兩": "两", "冊": "册", "冪": "幂", "凈": "净", "凍": "冻", "凜": "凛", "幾": "几", "鳳": "凤", "凱": "凯", "別": "别", "刪": "删", "剛": "刚", "創": "创", "劃": "划", "劇": "剧", "劉": "刘", "劍": "剑", "劑": "剂", "勁": "劲", "動": "动", "務": "务", "勝": "胜", "勞": "劳", "勢": "势", "勛": "勋", "勵": "励", "勸": "劝", "區": "区", "醫": "医", "華": "华", "協": "协", "單": "单", "賣": "卖", "盧": "卢", "衛": "卫", "卻": "却", "廠": "厂", "廳": "厅", "歷": "历", "厲": "厉", "壓": "压", "參": "参", "雙": "双", "發": "发", "變": "变", "敘": "叙", "疊": "叠", "葉": "叶", "號": "号", "嘆": "叹", "後": "后", "嚇": "吓", "聽": "听", "啟": "启", "吳": "吴", "員": "员", "週": "周", "呼": "呼", "和": "和", "諮": "咨", "詠": "咏", "咼": "呙", "員": "员", "響": "响", "啞": "哑", "問": "问", "啟": "启", "啢": "唡", "喚": "唤", "喪": "丧", "喬": "乔", "單": "单", "喲": "哟", "嗆": "呛", "嗇": "啬", "嗎": "吗", "嗚": "呜", "嗩": "唢", "嗶": "哔", "嘆": "叹", "嘔": "呕", "嘖": "啧", "嘗": "尝", "嘜": "唛", "嘩": "哗", "嘮": "唠", "嘯": "啸", "嘰": "叽", "噓": "嘘", "噚": "寻", "噴": "喷", "噸": "吨", "噹": "当", "嚀": "咛", "嚇": "吓", "嚌": "哜", "嚕": "噜", "嚙": "啮", "嚥": "咽", "嚨": "咙", "嚮": "向", "嚳": "喾", "囑": "嘱", "囂": "嚣", "囅": "冁", "囈": "呓", "囉": "啰", "囌": "苏", "圍": "围", "園": "园", "國": "国", "圖": "图", "圓": "圆", "聖": "圣", "場": "场", "壞": "坏", "塊": "块", "堅": "坚", "壇": "坛", "壩": "坝", "墜": "坠", "墮": "堕", "墳": "坟", "墻": "墙", "壓": "压", "壘": "垒", "壙": "圹", "壚": "垆", "壟": "垄", "壢": "坜", "壩": "坝", "壯": "壮", "聲": "声", "壺": "壶", "壽": "寿", "夠": "够", "夢": "梦", "夥": "伙", "夾": "夹", "奐": "奂", "奧": "奥", "奩": "奁", "奪": "夺", "奮": "奋", "妝": "妆", "婦": "妇", "媽": "妈", "姍": "姗", "姦": "奸", "娛": "娱", "婁": "娄", "婦": "妇", "婭": "娅", "媧": "娲", "媯": "妫", "媼": "媪", "媽": "妈", "嫗": "妪", "嫵": "妩", "嫻": "娴", "嬋": "婵", "嬌": "娇", "嬙": "嫱", "嬡": "嫒", "嬤": "嬷", "孌": "娈", "孫": "孙", "學": "学", "孿": "孪", "宮": "宫", "寢": "寝", "實": "实", "寧": "宁", "審": "审", "寫": "写", "寬": "宽", "寵": "宠", "寶": "宝", "將": "将", "專": "专", "尋": "寻", "對": "对", "導": "导", "尷": "尴", "屆": "届", "屍": "尸", "屓": "屃", "屜": "屉", "屢": "屡", "層": "层", "屬": "属", "岡": "冈", "峴": "岘", "島": "岛", "峽": "峡", "崍": "崃", "崗": "岗", "嵐": "岚", "嶁": "嵝", "嶄": "崭", "嶇": "岖", "嶗": "崂", "嶠": "峤", "嶢": "峣", "嶧": "峄", "嶮": "崄", "嶴": "岙", "嶸": "嵘", "嶺": "岭", "嶼": "屿", "巋": "岿", "巒": "峦", "巔": "巅", "幀": "帧", "帥": "帅", "師": "师", "帳": "帐", "帶": "带", "幀": "帧", "幃": "帏", "幗": "帼", "幘": "帻", "幟": "帜", "幣": "币", "幫": "帮", "幬": "帱", "幹": "干", "庫": "库", "廁": "厕", "廂": "厢", "廈": "厦", "廚": "厨", "廟": "庙", "廠": "厂", "廣": "广", "廬": "庐", "廳": "厅", "弒": "弑", "弔": "吊", "張": "张", "強": "强", "彆": "别", "彈": "弹", "彌": "弥", "彎": "弯", "彙": "汇", "彞": "彝", "彥": "彦", "後": "后", "徑": "径", "從": "从", "徠": "徕", "復": "复", "徵": "征", "德": "德", "徹": "彻", "恆": "恒", "恥": "耻", "悅": "悦", "悞": "悮", "悵": "怅", "悶": "闷", "惡": "恶", "惱": "恼", "惲": "恽", "愛": "爱", "愜": "惬", "愴": "怆", "愷": "恺", "愾": "忾", "慄": "栗", "態": "态", "慍": "愠", "慘": "惨", "慚": "惭", "慟": "恸", "慣": "惯", "慤": "悫", "慪": "怄", "慫": "怂", "慮": "虑", "慳": "悭", "慶": "庆", "憂": "忧", "憊": "惫", "憐": "怜", "憑": "凭", "憒": "愦", "憚": "惮", "憤": "愤", "憫": "悯", "憮": "怃", "憲": "宪", "憶": "忆", "懇": "恳", "應": "应", "懌": "怿", "懍": "懔", "懞": "蒙", "懟": "怼", "懣": "懑", "懨": "恹", "懲": "惩", "懶": "懒", "懷": "怀", "懸": "悬", "懺": "忏", "懼": "惧", "懾": "慑", "戀": "恋", "戇": "戆", "戔": "戋", "戧": "戗", "戰": "战", "戲": "戏", "戶": "户", "拋": "抛", "挾": "挟", "捨": "舍", "捫": "扪", "掃": "扫", "掄": "抡", "掗": "挜", "掙": "挣", "掛": "挂", "採": "采", "揀": "拣", "揚": "扬", "換": "换", "揮": "挥", "損": "损", "搖": "摇", "搗": "捣", "搶": "抢", "摑": "掴", "摜": "掼", "摟": "搂", "摯": "挚", "摳": "抠", "摶": "抟", "摺": "折", "撈": "捞", "撐": "撑", "撓": "挠", "撥": "拨", "撫": "抚", "撲": "扑", "撳": "揿", "撻": "挞", "撾": "挝", "撿": "捡", "擁": "拥", "擄": "掳", "擇": "择", "擊": "击", "擋": "挡", "擔": "担", "據": "据", "擠": "挤", "擬": "拟", "擯": "摈", "擰": "拧", "擱": "搁", "擲": "掷", "擴": "扩", "擷": "撷", "擺": "摆", "擻": "擞", "擼": "撸", "擾": "扰", "攄": "摅", "攆": "撵", "攏": "拢", "攔": "拦", "攙": "搀", "攜": "携", "攝": "摄", "攢": "攒", "攣": "挛", "攤": "摊", "攪": "搅", "攬": "揽", "敗": "败", "敘": "叙", "敵": "敌", "數": "数", "斂": "敛", "斃": "毙", "斕": "斓", "斬": "斩", "斷": "断", "於": "于", "時": "时", "晉": "晋", "晝": "昼", "暈": "晕", "暉": "晖", "暢": "畅", "暫": "暂", "曄": "晔", "曆": "历", "曇": "昙", "曉": "晓", "曏": "向", "曖": "暧", "曠": "旷", "曬": "晒", "書": "书", "會": "会", "朧": "胧", "東": "东", "柵": "栅", "桿": "杆", "梔": "栀", "條": "条", "梟": "枭", "梲": "棁", "棄": "弃", "棖": "枨", "棗": "枣", "棟": "栋", "棧": "栈", "棲": "栖", "椏": "桠", "楊": "杨", "楓": "枫", "楨": "桢", "業": "业", "極": "极", "榪": "杩", "榮": "荣", "榲": "榅", "榿": "桤", "構": "构", "槍": "枪", "槓": "杠", "槤": "梿", "槧": "椠", "槨": "椁", "槳": "桨", "樁": "桩", "樂": "乐", "樅": "枞", "樓": "楼", "標": "标", "樞": "枢", "樣": "样", "樸": "朴", "樹": "树", "樺": "桦", "橈": "桡", "橋": "桥", "機": "机", "橢": "椭", "橫": "横", "檁": "檩", "檉": "柽", "檔": "档", "檜": "桧", "檟": "槚", "檢": "检", "檣": "樯", "檯": "台", "檳": "槟", "檸": "柠", "檻": "槛", "櫃": "柜", "櫓": "橹", "櫚": "榈", "櫛": "栉", "櫝": "椟", "櫞": "橼", "櫟": "栎", "櫥": "橱", "櫧": "槠", "櫨": "栌", "櫪": "枥", "櫫": "橥", "櫬": "榇", "櫳": "栊", "櫸": "榉", "櫻": "樱", "欄": "栏", "權": "权", "欏": "椤", "欒": "栾", "欖": "榄", "歡": "欢", "歲": "岁", "歷": "历", "歸": "归", "歿": "殁", "殘": "残", "殞": "殒", "殤": "殇", "殫": "殚", "殭": "僵", "殮": "殓", "殯": "殡", "殲": "歼", "殺": "杀", "殼": "壳", "毀": "毁", "毆": "殴", "毿": "毵", "氈": "毡", "氌": "氇", "氣": "气", "氫": "氢", "氬": "氩", "氳": "氲", "決": "决", "沒": "没", "沖": "冲", "況": "况", "洶": "汹", "浹": "浃", "涇": "泾", "涼": "凉", "淒": "凄", "淚": "泪", "淥": "渌", "淨": "净", "淪": "沦", "淵": "渊", "淶": "涞", "淺": "浅", "渙": "涣", "減": "减", "渢": "沨", "渦": "涡", "測": "测", "渾": "浑", "湊": "凑", "湞": "浈", "湯": "汤", "溈": "沩", "準": "准", "溝": "沟", "溫": "温", "滄": "沧", "滅": "灭", "滌": "涤", "滎": "荥", "滬": "沪", "滯": "滞", "滲": "渗", "滷": "卤", "滸": "浒", "滻": "浐", "滾": "滚", "滿": "满", "漁": "渔", "漚": "沤", "漢": "汉", "漣": "涟", "漬": "渍", "漲": "涨", "漵": "溆", "漸": "渐", "漿": "浆", "潁": "颍", "潑": "泼", "潔": "洁", "潙": "沩", "潛": "潜", "潤": "润", "潯": "浔", "潰": "溃", "潷": "滗", "潿": "涠", "澀": "涩", "澆": "浇", "澇": "涝", "澗": "涧", "澠": "渑", "澤": "泽", "澦": "滪", "澩": "泶", "澮": "浍", "澱": "淀", "濁": "浊", "濃": "浓", "濕": "湿", "濘": "泞", "濟": "济", "濤": "涛", "濫": "滥", "濰": "潍", "濱": "滨", "濺": "溅", "濼": "泺", "濾": "滤", "瀅": "滢", "瀆": "渎", "瀉": "泻", "瀋": "沈", "瀏": "浏", "瀕": "濒", "瀘": "泸", "瀝": "沥", "瀟": "潇", "瀠": "潆", "瀦": "潴", "瀧": "泷", "瀨": "濑", "瀰": "弥", "瀲": "潋", "瀾": "澜", "灃": "沣", "灄": "滠", "灑": "洒", "灘": "滩", "灣": "湾", "灤": "滦", "灧": "滟", "災": "灾", "為": "为", "烏": "乌", "無": "无", "煉": "炼", "煒": "炜", "煙": "烟", "煢": "茕", "煥": "焕", "煩": "烦", "煬": "炀", "熅": "煴", "熒": "荧", "熗": "炝", "熱": "热", "熲": "颎", "熾": "炽", "燁": "烨", "燈": "灯", "燉": "炖", "燒": "烧", "燙": "烫", "燜": "焖", "營": "营", "燦": "灿", "燭": "烛", "燴": "烩", "燼": "烬", "燾": "焘", "爍": "烁", "爐": "炉", "爛": "烂", "爭": "争", "爺": "爷", "爾": "尔", "牆": "墙", "牘": "牍", "牽": "牵", "犖": "荦", "犢": "犊", "犧": "牺", "狀": "状", "狹": "狭", "狼": "狼", "狽": "狈", "猙": "狰", "猶": "犹", "猻": "狲", "獁": "犸", "獄": "狱", "獅": "狮", "獎": "奖", "獨": "独", "獪": "狯", "獫": "猃", "獮": "狝", "獰": "狞", "獲": "获", "獵": "猎", "獸": "兽", "獺": "獭", "獻": "献", "獼": "猕", "玀": "猡", "現": "现", "琺": "珐", "琿": "珲", "瑋": "玮", "瑣": "琐", "瑤": "瑶", "瑩": "莹", "瑪": "玛", "瑲": "玱", "璉": "琏", "璣": "玑", "璦": "瑷", "環": "环", "璽": "玺", "瓊": "琼", "瓏": "珑", "瓔": "璎", "瓚": "瓒", "甕": "瓮", "產": "产", "畢": "毕", "異": "异", "畫": "画", "當": "当", "疇": "畴", "疊": "叠", "痙": "痉", "痠": "酸", "瘂": "痖", "瘋": "疯", "瘍": "疡", "瘓": "痪", "瘞": "瘗", "瘡": "疮", "瘧": "疟", "瘮": "瘆", "瘺": "瘘", "瘻": "瘘", "療": "疗", "癆": "痨", "癇": "痫", "癉": "瘅", "癒": "愈", "癘": "疠", "癟": "瘪", "癢": "痒", "癤": "疖", "癥": "症", "癧": "疬", "癩": "癞", "癬": "癣", "癭": "瘿", "癮": "瘾", "癰": "痈", "癱": "瘫", "癲": "癫", "發": "发", "皚": "皑", "皰": "疱", "皸": "皲", "皺": "皱", "盜": "盗", "盞": "盏", "盡": "尽", "監": "监", "盤": "盘", "盧": "卢", "眥": "眦", "眾": "众", "睏": "困", "睜": "睁", "瞞": "瞒", "瞭": "了", "矇": "蒙", "矓": "眬", "矚": "瞩", "矯": "矫", "硃": "朱", "硤": "硖", "硨": "砗", "硯": "砚", "碩": "硕", "碭": "砀", "確": "确", "碼": "码", "磑": "硙", "磚": "砖", "磣": "碜", "磧": "碛", "磯": "矶", "磽": "硗", "礎": "础", "礙": "碍", "礦": "矿", "礪": "砺", "礫": "砾", "礬": "矾", "祿": "禄", "禍": "祸", "禎": "祯", "禕": "祎", "禡": "祃", "禦": "御", "禪": "禅", "禮": "礼", "禰": "祢", "禱": "祷", "禿": "秃", "秈": "籼", "稅": "税", "稈": "秆", "稜": "棱", "種": "种", "稱": "称", "穀": "谷", "穌": "稣", "積": "积", "穎": "颖", "穠": "秾", "穡": "穑", "穢": "秽", "穩": "稳", "穫": "获", "穭": "稆", "窩": "窝", "窪": "洼", "窮": "穷", "窯": "窑", "窵": "窎", "窶": "窭", "窺": "窥", "竄": "窜", "竅": "窍", "竇": "窦", "竊": "窃", "競": "竞", "筆": "笔", "筍": "笋", "筧": "笕", "箋": "笺", "箏": "筝", "節": "节", "範": "范", "築": "筑", "篋": "箧", "篔": "筼", "篤": "笃", "篩": "筛", "篳": "筚", "簀": "箦", "簍": "篓", "簞": "箪", "簡": "简", "簣": "篑", "簫": "箫", "簹": "筜", "簽": "签", "簾": "帘", "籃": "篮", "籌": "筹", "籍": "籍", "籜": "箨", "籟": "籁", "籠": "笼", "籩": "笾", "籪": "簖", "籬": "篱", "籮": "箩", "籲": "吁", "粵": "粤", "糝": "糁", "糞": "粪", "糧": "粮", "糰": "团", "糲": "粝", "糴": "籴", "糶": "粜", "糾": "纠", "紀": "纪", "紂": "纣", "約": "约", "紅": "红", "紆": "纡", "紇": "纥", "紈": "纨", "紉": "纫", "紋": "纹", "納": "纳", "紐": "纽", "紓": "纾", "純": "纯", "紕": "纰", "紗": "纱", "紙": "纸", "級": "级", "紛": "纷", "紜": "纭", "紡": "纺", "索": "索", "緊": "紧", "紫": "紫", "累": "累", "細": "细", "紱": "绂", "紲": "绁", "紳": "绅", "紹": "绍", "紺": "绀", "紼": "绋", "絀": "绌", "終": "终", "組": "组", "絆": "绊", "結": "结", "絕": "绝", "絛": "绦", "絝": "绔", "絞": "绞", "絡": "络", "絢": "绚", "給": "给", "絨": "绒", "絰": "绖", "統": "统", "絲": "丝", "絳": "绛", "絹": "绢", "綁": "绑", "綃": "绡", "綆": "绠", "綈": "绨", "綉": "绣", "綏": "绥", "經": "经", "綜": "综", "綞": "缍", "綠": "绿", "綢": "绸", "綣": "绻", "綫": "线", "綬": "绶", "維": "维", "綰": "绾", "綱": "纲", "網": "网", "綳": "绷", "綴": "缀", "綵": "彩", "綸": "纶", "綹": "绺", "綺": "绮", "綻": "绽", "綽": "绰", "綾": "绫", "綿": "绵", "緄": "绲", "緇": "缁", "緊": "紧", "緋": "绯", "緒": "绪", "緓": "绬", "緔": "绱", "緗": "缃", "緘": "缄", "緙": "缂", "線": "线", "緝": "缉", "緞": "缎", "締": "缔", "緡": "缗", "緣": "缘", "緦": "缌", "編": "编", "緩": "缓", "緬": "缅", "緯": "纬", "緱": "缑", "緲": "缈", "練": "练", "緶": "缏", "緹": "缇", "緻": "致", "縈": "萦", "縉": "缙", "縊": "缢", "縋": "缒", "縐": "绉", "縑": "缣", "縕": "缊", "縗": "缞", "縛": "缚", "縝": "缜", "縞": "缟", "縟": "缛", "縣": "县", "縧": "绦", "縫": "缝", "縭": "缡", "縮": "缩", "縱": "纵", "縲": "缧", "縳": "缚", "縴": "纤", "縵": "缦", "縶": "絷", "縷": "缕", "縹": "缥", "總": "总", "績": "绩", "繃": "绷", "繅": "缫", "繆": "缪", "繈": "襁", "繒": "缯", "織": "织", "繕": "缮", "繚": "缭", "繞": "绕", "繡": "绣", "繢": "缋", "繩": "绳", "繪": "绘", "繫": "系", "繭": "茧", "繮": "缰", "繯": "缳", "繰": "缲", "繳": "缴", "繹": "绎", "繼": "继", "繽": "缤", "繾": "缱", "纈": "缬", "纊": "纩", "續": "续", "纍": "累", "纏": "缠", "纓": "缨", "纔": "才", "纖": "纤", "纘": "缵", "纜": "缆", "缽": "钵", "罈": "坛", "罌": "罂", "罰": "罚", "罵": "骂", "罷": "罢", "羅": "罗", "羆": "罴", "羈": "羁", "羋": "芈", "羥": "羟", "義": "义", "習": "习", "翬": "翚", "翹": "翘", "耬": "耧", "耮": "耢", "聖": "圣", "聞": "闻", "聯": "联", "聰": "聪", "聲": "声", "聳": "耸", "聵": "聩", "聶": "聂", "職": "职", "聹": "聍", "聽": "听", "聾": "聋", "肅": "肃", "脅": "胁", "脈": "脉", "脛": "胫", "脫": "脱", "脹": "胀", "腎": "肾", "腖": "胨", "腡": "脶", "腦": "脑", "腫": "肿", "腳": "脚", "腸": "肠", "膃": "腽", "膚": "肤", "膠": "胶", "膩": "腻", "膽": "胆", "膾": "脍", "膿": "脓", "臉": "脸", "臍": "脐", "臏": "膑", "臘": "腊", "臚": "胪", "臟": "脏", "臠": "脔", "臢": "臜", "臥": "卧", "臨": "临", "臺": "台", "與": "与", "興": "兴", "舉": "举", "舊": "旧", "艙": "舱", "艤": "舣", "艦": "舰", "艫": "舻", "艱": "艰", "艷": "艳", "藝": "艺", "節": "节", "芻": "刍", "莊": "庄", "莖": "茎", "莢": "荚", "莧": "苋", "華": "华", "萇": "苌", "萊": "莱", "萬": "万", "萵": "莴", "葉": "叶", "葒": "荭", "著": "著", "葤": "荮", "葦": "苇", "葷": "荤", "蒐": "搜", "蒔": "莳", "蒞": "莅", "蒼": "苍", "蓀": "荪", "蓋": "盖", "蓮": "莲", "蓯": "苁", "蓴": "莼", "蓽": "荜", "蔔": "卜", "蔞": "蒌", "蔣": "蒋", "蔥": "葱", "蔦": "茑", "蔭": "荫", "蕁": "荨", "蕆": "蒇", "蕎": "荞", "蕒": "荬", "蕓": "芸", "蕕": "莸", "蕘": "荛", "蕢": "蒉", "蕩": "荡", "蕪": "芜", "蕭": "萧", "蕷": "蓣", "薈": "荟", "薊": "蓟", "薌": "芗", "薔": "蔷", "薘": "荙", "薟": "莶", "薦": "荐", "薩": "萨", "薺": "荠", "藍": "蓝", "藎": "荩", "藝": "艺", "藥": "药", "藪": "薮", "藴": "蕴", "藶": "苈", "藹": "蔼", "藺": "蔺", "蘄": "蕲", "蘆": "芦", "蘇": "苏", "蘊": "蕴", "蘋": "苹", "蘚": "藓", "蘞": "蔹", "蘢": "茏", "蘭": "兰", "蘺": "蓠", "蘿": "萝", "處": "处", "虛": "虚", "虜": "虏", "號": "号", "虧": "亏", "蛺": "蛱", "蛻": "蜕", "蜆": "蚬", "蝕": "蚀", "蝟": "猬", "蝦": "虾", "蝸": "蜗", "螄": "蛳", "螞": "蚂", "螢": "萤", "螻": "蝼", "螿": "螀", "蟄": "蛰", "蟈": "蝈", "蟎": "螨", "蟣": "虮", "蟬": "蝉", "蟯": "蛲", "蟲": "虫", "蟶": "蛏", "蟻": "蚁", "蠅": "蝇", "蠆": "虿", "蠍": "蝎", "蠐": "蛴", "蠑": "蝾", "蠟": "蜡", "蠣": "蛎", "蠱": "蛊", "蠶": "蚕", "蠻": "蛮", "衆": "众", "衊": "蔑", "術": "术", "衕": "同", "衚": "胡", "衛": "卫", "衝": "冲", "袞": "衮", "裊": "袅", "裏": "里", "補": "补", "裝": "装", "裡": "里", "製": "制", "複": "复", "褲": "裤", "褳": "裢", "褸": "褛", "褻": "亵", "襇": "裥", "襏": "袯", "襖": "袄", "襝": "裣", "襠": "裆", "襤": "褴", "襪": "袜", "襬": "摆", "襯": "衬", "襲": "袭", "見": "见", "覎": "觃", "規": "规", "覓": "觅", "視": "视", "覘": "觇", "覡": "觋", "覥": "觍", "覦": "觎", "親": "亲", "覬": "觊", "覯": "觏", "覲": "觐", "覷": "觑", "覺": "觉", "覽": "览", "觀": "观", "觴": "觞", "觸": "触", "訂": "订", "訃": "讣", "計": "计", "訊": "讯", "訌": "讧", "討": "讨", "訐": "讦", "訒": "讱", "訓": "训", "訕": "讪", "訖": "讫", "託": "托", "記": "记", "訛": "讹", "訝": "讶", "訟": "讼", "訢": "欣", "訣": "诀", "訥": "讷", "訩": "讻", "訪": "访", "設": "设", "許": "许", "訴": "诉", "訶": "诃", "診": "诊", "註": "注", "詁": "诂", "詆": "诋", "詎": "讵", "詐": "诈", "詒": "诒", "詔": "诏", "評": "评", "詖": "诐", "詗": "诇", "詘": "诎", "詛": "诅", "詞": "词", "詠": "咏", "詡": "诩", "詢": "询", "詣": "诣", "試": "试", "詩": "诗", "詫": "诧", "詬": "诟", "詭": "诡", "詮": "诠", "詰": "诘", "話": "话", "該": "该", "詳": "详", "詵": "诜", "詼": "诙", "詿": "诖", "誄": "诔", "誅": "诛", "誆": "诓", "誇": "夸", "誌": "志", "認": "认", "誑": "诳", "誒": "诶", "誕": "诞", "誘": "诱", "誚": "诮", "語": "语", "誠": "诚", "誡": "诫", "誣": "诬", "誤": "误", "誥": "诰", "誦": "诵", "誨": "诲", "說": "说", "説": "说", "誰": "谁", "課": "课", "誶": "谇", "誹": "诽", "誼": "谊", "調": "调", "諂": "谄", "諄": "谆", "談": "谈", "諉": "诿", "請": "请", "諍": "诤", "諏": "诹", "諑": "诼", "諒": "谅", "論": "论", "諗": "谂", "諛": "谀", "諜": "谍", "諞": "谝", "諡": "谥", "諢": "诨", "諤": "谔", "諦": "谛", "諧": "谐", "諫": "谏", "諭": "谕", "諮": "咨", "諱": "讳", "諳": "谙", "諶": "谌", "諷": "讽", "諸": "诸", "諺": "谚", "諼": "谖", "諾": "诺", "謀": "谋", "謁": "谒", "謂": "谓", "謄": "誊", "謅": "诌", "謊": "谎", "謎": "谜", "謐": "谧", "謔": "谑", "謖": "谡", "謗": "谤", "謙": "谦", "謚": "谥", "講": "讲", "謝": "谢", "謠": "谣", "謡": "谣", "謨": "谟", "謫": "谪", "謬": "谬", "謭": "谫", "謳": "讴", "謹": "谨", "謾": "谩", "譁": "哗", "證": "证", "譎": "谲", "譏": "讥", "譖": "谮", "識": "识", "譙": "谯", "譚": "谭", "譜": "谱", "譫": "谵", "譯": "译", "議": "议", "譴": "谴", "護": "护", "譸": "诪", "譽": "誉", "讀": "读", "變": "变", "讎": "雠", "讒": "谗", "讓": "让", "讕": "谰", "讖": "谶", "讚": "赞", "讞": "谳", "豈": "岂", "豎": "竖", "豐": "丰", "豬": "猪", "貓": "猫", "貝": "贝", "貞": "贞", "負": "负", "財": "财", "貢": "贡", "貧": "贫", "貨": "货", "販": "贩", "貪": "贪", "貫": "贯", "責": "责", "貯": "贮", "貰": "贳", "貲": "赀", "貳": "贰", "貴": "贵", "貶": "贬", "買": "买", "貸": "贷", "貺": "贶", "費": "费", "貼": "贴", "貽": "贻", "貿": "贸", "賀": "贺", "賁": "贲", "賂": "赂", "賃": "赁", "賄": "贿", "資": "资", "賈": "贾", "賊": "贼", "賑": "赈", "賒": "赊", "賓": "宾", "賕": "赇", "賙": "赒", "賚": "赉", "賜": "赐", "賞": "赏", "賠": "赔", "賡": "赓", "賢": "贤", "賣": "卖", "賤": "贱", "賦": "赋", "質": "质", "賬": "账", "賭": "赌", "賴": "赖", "賵": "赗", "賺": "赚", "購": "购", "賽": "赛", "贄": "贽", "贅": "赘", "贈": "赠", "贊": "赞", "贍": "赡", "贏": "赢", "贐": "赆", "贓": "赃", "贔": "赑", "贖": "赎", "贗": "赝", "贛": "赣", "趙": "赵", "趕": "赶", "趨": "趋", "跡": "迹", "踐": "践", "踴": "踊", "蹌": "跄", "蹕": "跸", "蹟": "迹", "蹣": "蹒", "蹤": "踪", "蹺": "跷", "躂": "跶", "躉": "趸", "躊": "踌", "躋": "跻", "躍": "跃", "躑": "踯", "躒": "跞", "躓": "踬", "躕": "蹰", "躚": "跹", "躡": "蹑", "躥": "蹿", "躦": "躜", "躪": "躏", "軀": "躯", "車": "车", "軋": "轧", "軌": "轨", "軍": "军", "軒": "轩", "軔": "轫", "軛": "轭", "軟": "软", "軤": "轷", "軫": "轸", "軲": "轱", "軸": "轴", "軹": "轵", "軺": "轺", "軻": "轲", "軼": "轶", "軾": "轼", "較": "较", "輅": "辂", "輇": "辁", "載": "载", "輊": "轾", "輒": "辄", "輓": "挽", "輔": "辅", "輕": "轻", "輛": "辆", "輜": "辎", "輝": "辉", "輞": "辋", "輟": "辍", "輥": "辊", "輦": "辇", "輩": "辈", "輪": "轮", "輯": "辑", "輳": "辏", "輸": "输", "輻": "辐", "輾": "辗", "輿": "舆", "轀": "辒", "轂": "毂", "轄": "辖", "轅": "辕", "轆": "辘", "轉": "转", "轍": "辙", "轎": "轿", "轔": "辚", "轟": "轰", "轡": "辔", "轢": "轹", "轤": "轳", "辦": "办", "辭": "辞", "辮": "辫", "辯": "辩", "農": "农", "迴": "回", "逕": "迳", "這": "这", "連": "连", "週": "周", "進": "进", "遊": "游", "運": "运", "過": "过", "達": "达", "違": "违", "遙": "遥", "遜": "逊", "遞": "递", "遠": "远", "適": "适", "遲": "迟", "遷": "迁", "選": "选", "遺": "遗", "遼": "辽", "邁": "迈", "還": "还", "邇": "迩", "邊": "边", "邏": "逻", "邐": "逦", "郟": "郏", "郵": "邮", "鄆": "郓", "鄉": "乡", "鄒": "邹", "鄔": "邬", "鄖": "郧", "鄧": "邓", "鄭": "郑", "鄰": "邻", "鄲": "郸", "鄴": "邺", "鄶": "郐", "鄺": "邝", "酇": "酂", "酈": "郦", "醞": "酝", "醬": "酱", "醫": "医", "醱": "酦", "釀": "酿", "釁": "衅", "釃": "酾", "釅": "酽", "釋": "释", "釐": "厘", "釒": "钅", "釓": "钆", "釔": "钇", "釕": "钌", "釗": "钊", "釘": "钉", "釙": "钋", "針": "针", "釣": "钓", "釤": "钐", "釦": "扣", "釧": "钏", "釩": "钒", "釵": "钗", "釷": "钍", "釹": "钕", "釺": "钎", "鈀": "钯", "鈁": "钫", "鈃": "钘", "鈄": "钭", "鈈": "钚", "鈉": "钠", "鈍": "钝", "鈎": "钩", "鈐": "钤", "鈑": "钣", "鈒": "钑", "鈔": "钞", "鈕": "钮", "鈞": "钧", "鈣": "钙", "鈥": "钬", "鈦": "钛", "鈧": "钪", "鈮": "铌", "鈰": "铈", "鈳": "钶", "鈴": "铃", "鈷": "钴", "鈸": "钹", "鈹": "铍", "鈺": "钰", "鈽": "钸", "鈾": "铀", "鈿": "钿", "鉀": "钾", "鉅": "钜", "鉈": "铊", "鉉": "铉", "鉋": "刨", "鉍": "铋", "鉑": "铂", "鉕": "钷", "鉗": "钳", "鉚": "铆", "鉛": "铅", "鉞": "钺", "鉢": "钵", "鉤": "钩", "鉦": "钲", "鉬": "钼", "鉭": "钽", "鉶": "铏", "鉸": "铰", "鉺": "铒", "鉻": "铬", "鉿": "铪", "銀": "银", "銃": "铳", "銅": "铜", "銍": "铚", "銑": "铣", "銓": "铨", "銖": "铢", "銘": "铭", "銚": "铫", "銛": "铦", "銜": "衔", "銠": "铑", "銣": "铷", "銥": "铱", "銦": "铟", "銨": "铵", "銩": "铥", "銪": "铕", "銫": "铯", "銬": "铐", "銱": "铞", "銳": "锐", "銷": "销", "銹": "锈", "銻": "锑", "銼": "锉", "鋁": "铝", "鋃": "锒", "鋅": "锌", "鋇": "钡", "鋌": "铤", "鋏": "铗", "鋒": "锋", "鋝": "锊", "鋟": "锓", "鋣": "铘", "鋤": "锄", "鋥": "锃", "鋦": "锔", "鋨": "锇", "鋩": "铓", "鋪": "铺", "鋭": "锐", "鋮": "铖", "鋯": "锆", "鋰": "锂", "鋱": "铽", "鋶": "锍", "鋸": "锯", "鋼": "钢", "錄": "录", "錆": "锖", "錇": "锫", "錈": "锩", "錏": "铔", "錐": "锥", "錒": "锕", "錕": "锟", "錘": "锤", "錙": "锱", "錚": "铮", "錛": "锛", "錟": "锬", "錠": "锭", "錡": "锜", "錢": "钱", "錦": "锦", "錨": "锚", "錫": "锡", "錮": "锢", "錯": "错", "錳": "锰", "錶": "表", "鍀": "锝", "鍁": "锨", "鍃": "锪", "鍆": "钔", "鍇": "锴", "鍈": "锳", "鍋": "锅", "鍍": "镀", "鍔": "锷", "鍘": "铡", "鍚": "钖", "鍛": "锻", "鍤": "锸", "鍥": "锲", "鍩": "锘", "鍬": "锹", "鍰": "锾", "鍵": "键", "鍶": "锶", "鍺": "锗", "鍾": "钟", "鎂": "镁", "鎄": "锿", "鎇": "镅", "鎊": "镑", "鎌": "镰", "鎔": "镕", "鎖": "锁", "鎘": "镉", "鎚": "锤", "鎛": "镈", "鎝": "𨱏", "鎡": "镃", "鎢": "钨", "鎣": "蓥", "鎦": "镏", "鎧": "铠", "鎩": "铩", "鎪": "锼", "鎬": "镐", "鎮": "镇", "鎰": "镒", "鎵": "镓", "鎶": "𫓹", "鎸": "镌", "鎿": "镎", "鏃": "镞", "鏈": "链", "鏇": "镟", "鏌": "镆", "鏍": "镙", "鏐": "镠", "鏑": "镝", "鏗": "铿", "鏘": "锵", "鏜": "镗", "鏝": "镘", "鏞": "镛", "鏟": "铲", "鏡": "镜", "鏢": "镖", "鏤": "镂", "鏨": "錾", "鏰": "镚", "鏵": "铧", "鏷": "镤", "鏹": "镪", "鏽": "锈", "鐃": "铙", "鐋": "铴", "鐐": "镣", "鐒": "铹", "鐓": "镦", "鐔": "镡", "鐘": "钟", "鐙": "镫", "鐠": "镨", "鐦": "锎", "鐧": "锏", "鐨": "镄", "鐫": "镌", "鐮": "镰", "鐲": "镯", "鐳": "镭", "鐵": "铁", "鐶": "镮", "鐸": "铎", "鐺": "铛", "鐿": "镱", "鑄": "铸", "鑊": "镬", "鑌": "镔", "鑒": "鉴", "鑔": "镲", "鑕": "锧", "鑞": "镴", "鑠": "铄", "鑣": "镳", "鑥": "镥", "鑭": "镧", "鑰": "钥", "鑱": "镵", "鑲": "镶", "鑷": "镊", "鑹": "镩", "鑼": "锣", "鑽": "钻", "鑾": "銮", "鑿": "凿", "長": "长", "門": "门", "閂": "闩", "閃": "闪", "閆": "闫", "閉": "闭", "開": "开", "閌": "闶", "閎": "闳", "閏": "闰", "閑": "闲", "間": "间", "閔": "闵", "閘": "闸", "閡": "阂", "閣": "阁", "閤": "合", "閥": "阀", "閨": "闺", "閩": "闽", "閫": "阃", "閬": "阆", "閭": "闾", "閱": "阅", "閲": "阅", "閶": "阊", "閹": "阉", "閻": "阎", "閼": "阏", "閽": "阍", "閾": "阈", "闃": "阒", "闆": "板", "闈": "闱", "闊": "阔", "闋": "阕", "闌": "阑", "闍": "阇", "闐": "阗", "闔": "阖", "闕": "阙", "闖": "闯", "關": "关", "闞": "阚", "闡": "阐", "闢": "辟", "闤": "阛", "阨": "厄", "陘": "陉", "陝": "陕", "陣": "阵", "陰": "阴", "陳": "陈", "陸": "陆", "陽": "阳", "隉": "陧", "隊": "队", "階": "阶", "隕": "陨", "際": "际", "隨": "随", "險": "险", "隱": "隐", "隴": "陇", "隸": "隶", "隻": "只", "雋": "隽", "雖": "虽", "雙": "双", "雛": "雏", "雜": "杂", "雞": "鸡", "離": "离", "難": "难", "雲": "云", "電": "电", "霧": "雾", "霽": "霁", "靂": "雳", "靄": "霭", "靈": "灵", "靚": "靓", "靜": "静", "靦": "腼", "鞏": "巩", "鞝": "绱", "鞦": "秋", "韁": "缰", "韃": "鞑", "韆": "千", "韉": "鞯", "韋": "韦", "韌": "韧", "韓": "韩", "韙": "韪", "韜": "韬", "韞": "韫", "韻": "韵", "響": "响", "頁": "页", "頂": "顶", "頃": "顷", "項": "项", "順": "顺", "須": "须", "頊": "顼", "頌": "颂", "預": "预", "頑": "顽", "頒": "颁", "頓": "顿", "頗": "颇", "領": "领", "頜": "颌", "頡": "颉", "頤": "颐", "頦": "颏", "頭": "头", "頮": "颒", "頰": "颊", "頲": "颋", "頴": "颕", "頷": "颔", "頸": "颈", "頹": "颓", "頻": "频", "顆": "颗", "題": "题", "額": "额", "顎": "颚", "顏": "颜", "顒": "颙", "顓": "颛", "願": "愿", "顙": "颡", "顛": "颠", "類": "类", "顢": "颟", "顥": "颢", "顧": "顾", "顫": "颤", "顬": "颥", "顯": "显", "顰": "颦", "顱": "颅", "顳": "颞", "顴": "颧", "風": "风", "颮": "飑", "颯": "飒", "颱": "台", "颳": "刮", "颶": "飓", "颺": "飏", "颻": "飖", "飄": "飘", "飆": "飙", "飛": "飞", "飢": "饥", "飯": "饭", "飲": "饮", "飴": "饴", "飼": "饲", "飽": "饱", "飾": "饰", "餃": "饺", "餅": "饼", "餉": "饷", "養": "养", "餌": "饵", "餎": "饹", "餏": "饻", "餑": "饽", "餒": "馁", "餓": "饿", "餘": "余", "餚": "肴", "餛": "馄", "餜": "馃", "餞": "饯", "餡": "馅", "館": "馆", "餱": "糇", "餳": "饧", "餵": "喂", "餶": "馉", "餷": "馇", "餺": "馎", "餼": "饩", "餾": "馏", "餿": "馊", "饁": "馌", "饃": "馍", "饅": "馒", "饈": "馐", "饉": "馑", "饊": "馓", "饋": "馈", "饌": "馔", "饎": "饩", "饑": "饥", "饒": "饶", "饗": "飨", "饜": "餍", "饞": "馋", "饢": "馕", "馬": "马", "馭": "驭", "馮": "冯", "馱": "驮", "馳": "驰", "馴": "驯", "駁": "驳", "駐": "驻", "駑": "驽", "駒": "驹", "駔": "驵", "駕": "驾", "駘": "骀", "駙": "驸", "駛": "驶", "駝": "驼", "駟": "驷", "駡": "骂", "駢": "骈", "駭": "骇", "駰": "骃", "駱": "骆", "駸": "骎", "駿": "骏", "騁": "骋", "騂": "骍", "騅": "骓", "騌": "骔", "騍": "骒", "騎": "骑", "騏": "骐", "騖": "骛", "騙": "骗", "騫": "骞", "騭": "骘", "騮": "骝", "騰": "腾", "騶": "驺", "騷": "骚", "騸": "骟", "騾": "骡", "驀": "蓦", "驁": "骜", "驂": "骖", "驃": "骠", "驄": "骢", "驅": "驱", "驊": "骅", "驌": "骕", "驍": "骁", "驏": "骣", "驕": "骄", "驗": "验", "驚": "惊", "驛": "驿", "驟": "骤", "驢": "驴", "驤": "骧", "驥": "骥", "驦": "骦", "驪": "骊", "骯": "肮", "髏": "髅", "髒": "脏", "體": "体", "髕": "髌", "髖": "髋", "鬆": "松", "鬍": "胡", "鬚": "须", "鬥": "斗", "鬧": "闹", "鬩": "阋", "鬱": "郁", "魎": "魉", "魘": "魇", "魚": "鱼", "魛": "鱽", "魟": "𫚉", "魢": "鱾", "魨": "鲀", "魯": "鲁", "魴": "鲂", "鮁": "鲅", "鮃": "鲆", "鮊": "鲌", "鮋": "鲉", "鮍": "鲏", "鮎": "鲇", "鮐": "鲐", "鮑": "鲍", "鮒": "鲋", "鮓": "鲊", "鮚": "鲒", "鮜": "鲘", "鮝": "鲞", "鮞": "鲕", "鮦": "鲖", "鮪": "鲔", "鮫": "鲛", "鮭": "鲑", "鮮": "鲜", "鮳": "鲓", "鮶": "鲪", "鮺": "鲝", "鯀": "鲧", "鯁": "鲠", "鯇": "鲩", "鯉": "鲤", "鯊": "鲨", "鯒": "鲬", "鯔": "鲻", "鯕": "鲯", "鯖": "鲭", "鯗": "鲞", "鯛": "鲷", "鯝": "鲴", "鯡": "鲱", "鯢": "鲵", "鯤": "鲲", "鯧": "鲳", "鯨": "鲸", "鯪": "鲮", "鯫": "鲰", "鯰": "鲶", "鯴": "鲺", "鯷": "鳀", "鯽": "鲫", "鯿": "鳊", "鰁": "鳈", "鰂": "鲗", "鰃": "鳂", "鰈": "鲽", "鰉": "鳇", "鰍": "鳅", "鰏": "鲾", "鰐": "鳄", "鰒": "鳆", "鰓": "鳃", "鰜": "鳒", "鰟": "鳑", "鰠": "鳋", "鰣": "鲥", "鰥": "鳏", "鰨": "鳎", "鰩": "鳐", "鰭": "鳍", "鰱": "鲢", "鰲": "鳌", "鰳": "鳓", "鰵": "鳘", "鰷": "鲦", "鰹": "鲣", "鰻": "鳗", "鰼": "鳛", "鰾": "鳔", "鱂": "鳉", "鱈": "鳕", "鱉": "鳖", "鱒": "鳟", "鱔": "鳝", "鱖": "鳜", "鱗": "鳞", "鱘": "鲟", "鱝": "鲼", "鱟": "鲎", "鱠": "鲙", "鱣": "鳣", "鱧": "鳢", "鱨": "鲿", "鱭": "鲚", "鱷": "鳄", "鱸": "鲈", "鱺": "鲡", "鳥": "鸟", "鳧": "凫", "鳩": "鸠", "鳲": "鸤", "鳳": "凤", "鳴": "鸣", "鳶": "鸢", "鴆": "鸩", "鴇": "鸨", "鴉": "鸦", "鴒": "鸰", "鴕": "鸵", "鴛": "鸳", "鴝": "鸲", "鴞": "鸮", "鴟": "鸱", "鴣": "鸪", "鴦": "鸯", "鴨": "鸭", "鴯": "鸸", "鴰": "鸹", "鴴": "鸻", "鴻": "鸿", "鴿": "鸽", "鵂": "鸺", "鵃": "鸼", "鵑": "鹃", "鵒": "鹆", "鵓": "鹁", "鵜": "鹈", "鵝": "鹅", "鵠": "鹄", "鵡": "鹉", "鵪": "鹌", "鵬": "鹏", "鵮": "鹐", "鵯": "鹎", "鵲": "鹊", "鶇": "鸫", "鶉": "鹑", "鶊": "鹒", "鶓": "鹋", "鶖": "鹙", "鶘": "鹕", "鶚": "鹗", "鶡": "鹖", "鶥": "鹛", "鶩": "鹜", "鶪": "䴗", "鶯": "莺", "鶲": "鹟", "鶴": "鹤", "鶹": "鹠", "鶺": "鹡", "鶻": "鹘", "鶼": "鹣", "鶿": "鹚", "鷁": "鹢", "鷂": "鹞", "鷈": "䴘", "鷓": "鹧", "鷖": "鹥", "鷗": "鸥", "鷙": "鸷", "鷚": "鹨", "鷥": "鸶", "鷦": "鹪", "鷯": "鹩", "鷲": "鹫", "鷳": "鹇", "鷸": "鹬", "鷹": "鹰", "鷺": "鹭", "鷽": "鸴", "鸇": "鹯", "鸊": "䴙", "鸌": "鹱", "鸕": "鸬", "鸚": "鹦", "鸛": "鹳", "鸝": "鹂", "鹵": "卤", "鹹": "咸", "鹺": "鹾", "鹼": "碱", "鹽": "盐", "麗": "丽", "麥": "麦", "麩": "麸", "麼": "么", "黃": "黄", "黌": "黉", "點": "点", "黨": "党", "黲": "黪", "黴": "霉", "黷": "黩", "黽": "黾", "鼇": "鳌", "鼉": "鼍", "鼴": "鼹", "齊": "齐", "齋": "斋", "齎": "赍", "齏": "齑", "齒": "齿", "齔": "龀", "齕": "龁", "齟": "龃", "齡": "龄", "齣": "出", "齦": "龈", "齧": "啮", "齪": "龊", "齬": "龉", "齲": "龋", "齶": "腭", "齷": "龌", "龍": "龙", "龐": "庞", "龔": "龚", "龕": "龛", "龜": "龟", }) def _to_zh_hans(text: str) -> str: return str(text or "").translate(ZH_HANS_TRANS) def _translate_chunk_with_minimax(texts: list[str], lang: str, api_key: str) -> list[str]: if not texts: return [] lang_name = _lang_prompt_name(lang) prompt = ( f"你是專業翻譯。把輸入 JSON 陣列每一項翻成 {lang_name}。\n" "規則：\n" "1) 不可遺漏項目，順序必須完全一致。\n" "2) 保留 URL、@帳號、#標籤、數字、時間、貨幣單位（如 USD/USDT/NTD）、專有名詞（Renaiss/SBT/BNB/Discord）。\n" "3) 不要加註解，不要加前後文。\n" "4) 只輸出 JSON 字串陣列。\n\n" f"source={json.dumps(texts, ensure_ascii=False)}" ) raw = minimax_chat(prompt, api_key) parsed = _parse_json_array(raw) if len(parsed) != len(texts): return list(texts) return [str(x or "").strip() or texts[idx] for idx, x in enumerate(parsed)] def _translate_texts(texts: list[str], lang: str) -> tuple[list[str], str]: tag = _normalize_lang_tag(lang) rows = [str(x or "").replace("\n", " ").strip()[:TRANSLATE_MAX_CHARS] for x in (texts or [])] if tag == "zh-Hant": return rows, "base" if tag == "zh-Hans": return [_to_zh_hans(x) for x in rows], "local" load_environment() api_key = resolve_minimax_key() out = list(rows) need_texts: list[str] = [] if TRANSLATION_CACHE_ENABLED: with TRANSLATION_LOCK: _load_translation_cache_unlocked() cached = TRANSLATION_CACHE need_indices: list[int] = [] for idx, text in enumerate(rows): if not text: continue key = _translation_cache_key(tag, text) got = cached.get(key) if got and not (tag != "zh-Hant" and _contains_cjk(text) and str(got) == text): out[idx] = got else: if got and tag != "zh-Hant" and _contains_cjk(text) and str(got) == text: cached.pop(key, None) need_indices.append(idx) need_texts.append(text) if not need_texts: return out, "cache" else: need_texts = [text for text in rows if text] if not api_key: return rows, "no-key" unique_texts: list[str] = [] seen: set[str] = set() for text in need_texts: if text in seen: continue seen.add(text) unique_texts.append(text) translated_map: dict[str, str] = {} chunk_size = 3 for start in range(0, len(unique_texts), chunk_size): chunk = unique_texts[start:start + chunk_size] try: translated = _translate_chunk_with_minimax(chunk, tag, api_key) except Exception: translated = list(chunk) for idx, original in enumerate(chunk): candidate = str(translated[idx] if idx < len(translated) else original).strip() translated_map[original] = candidate or original if TRANSLATION_CACHE_ENABLED: with TRANSLATION_LOCK: _load_translation_cache_unlocked() for original, translated in translated_map.items(): if not original: continue if tag != "zh-Hant" and _contains_cjk(original) and str(translated) == original: continue key = _translation_cache_key(tag, original) TRANSLATION_CACHE[key] = translated global TRANSLATION_CACHE_DIRTY TRANSLATION_CACHE_DIRTY = True _flush_translation_cache_unlocked() final_rows = [] for text in rows: if not text: final_rows.append("") continue final_rows.append(translated_map.get(text, text)) return final_rows, "live" def _is_feed_translatable_text(key: str, text: str) -> bool: name = str(key or "").strip() if not name or name in I18N_SKIP_KEYS: return False value = str(text or "") stripped = value.strip() if not stripped or len(stripped) < 2: return False if re.match(r"^https?://", stripped, re.I): return False if re.fullmatch(r"[@#:/\-\d\s.,()%+]+", stripped): return False if name in I18N_FEED_TEXT_KEYS: return True return _contains_cjk(stripped) def _join_entry_path(parent: str, key: str) -> str: k = str(key or "").strip() if not parent: return k if not k: return parent return f"{parent}.{k}" def _collect_feed_i18n_entries(node: object) -> list[tuple[str, str]]: out: list[tuple[str, str]] = [] seen_keys: set[str] = set() max_targets = max(0, I18N_MAX_TARGET_TEXTS) max_list_items = max(0, I18N_MAX_LIST_ITEMS_PER_FIELD) def _is_full() -> bool: return bool(max_targets and len(out) >= max_targets) def _iter_limited(rows: list[object]) -> list[object]: return rows[:max_list_items] if max_list_items else rows def _push(entry_key: str, value: str, field_key: str = "") -> None: if _is_full(): return key = str(entry_key or "").strip() text = str(value or "") if not key or key in seen_keys: return if not _is_feed_translatable_text(field_key, text): return seen_keys.add(key) out.append((key, text)) def _collect_section(value: object, path: str, field_key: str = "") -> None: if _is_full(): return if isinstance(value, str): _push(path, value, field_key) return if isinstance(value, list): for idx, item in enumerate(_iter_limited(value)): if _is_full(): break item_path = f"{path}[{idx}]" if isinstance(item, str): _push(item_path, item, field_key) elif isinstance(item, (dict, list)): _collect_section(item, item_path, field_key) return if isinstance(value, dict): for child_key, child in value.items(): if _is_full(): break child_name = str(child_key or "").strip() if not child_name: continue child_path = _join_entry_path(path, child_name) _collect_section(child, child_path, child_name) def _collect_card(card: dict[str, object], index: int) -> None: if _is_full(): return card_key = _card_lookup_key(card, index) card_path = f"cards[{card_key}]" for key in I18N_CARD_TEXT_KEYS: value = card.get(key) if isinstance(value, str): _push(f"{card_path}.{key}", value, key) for key in I18N_FEED_LIST_KEYS: value = card.get(key) if not isinstance(value, list): continue for idx, item in enumerate(_iter_limited(value)): if isinstance(item, str): _push(f"{card_path}.{key}[{idx}]", item, key) facts = card.get("event_facts") if isinstance(facts, dict): for fact_key in I18N_CARD_FACT_TEXT_KEYS: raw = facts.get(fact_key) if isinstance(raw, str): _push(f"{card_path}.event_facts.{fact_key}", raw, fact_key) if not isinstance(node, dict): return out for root_key in I18N_VISIBLE_FEED_ROOT_KEYS: if _is_full(): break raw = node.get(root_key) if raw is None: continue _collect_section(raw, root_key, root_key) cards = node.get("cards") if not isinstance(cards, list): return out for idx, item in enumerate(_iter_limited(cards)): if not isinstance(item, dict): continue _collect_card(item, idx) return out def _collect_feed_i18n_strings(node: object) -> list[str]: return [text for _, text in _collect_feed_i18n_entries(node)] def _entry_map_from_node(node: object) -> dict[str, str]: out: dict[str, str] = {} for entry_key, value in _collect_feed_i18n_entries(node): key = str(entry_key or "").strip() if not key: continue out[key] = str(value or "") return out def _apply_feed_translation(node: object, mapping: dict[str, str]) -> object: def _walk(value: object, parent_key: str = "", path: str = "") -> object: if isinstance(value, dict): out: dict[str, object] = {} for key, child in value.items(): k = str(key or "") next_path = _join_entry_path(path, k) if isinstance(child, str): if _is_feed_translatable_text(k, child): out[key] = mapping.get(next_path, child) else: out[key] = child elif isinstance(child, list): if k == "cards": rows: list[object] = [] for idx, item in enumerate(child): item_key = str(idx) if isinstance(item, dict): item_key = _card_lookup_key(item, idx) card_path = f"{next_path}[{item_key}]" if isinstance(item, str): if _is_feed_translatable_text(k, item): rows.append(mapping.get(card_path, item)) else: rows.append(item) elif isinstance(item, (dict, list)): rows.append(_walk(item, k, card_path)) else: rows.append(item) out[key] = rows elif k in I18N_FEED_LIST_KEYS: rows: list[object] = [] for idx, item in enumerate(child): item_path = f"{next_path}[{idx}]" if isinstance(item, str): if _is_feed_translatable_text(k, item): rows.append(mapping.get(item_path, item)) else: rows.append(item) elif isinstance(item, (dict, list)): rows.append(_walk(item, k, item_path)) else: rows.append(item) out[key] = rows else: out[key] = _walk(child, k, next_path) elif isinstance(child, dict): out[key] = _walk(child, k, next_path) else: out[key] = child return out if isinstance(value, list): rows: list[object] = [] for idx, item in enumerate(value): item_path = f"{path}[{idx}]" if path else f"[{idx}]" if isinstance(item, str): if _is_feed_translatable_text(parent_key, item): rows.append(mapping.get(item_path, item)) else: rows.append(item) elif isinstance(item, (dict, list)): rows.append(_walk(item, parent_key, item_path)) else: rows.append(item) return rows return value return _walk(node) def _build_best_effort_mapping(feed: dict[str, object], lang: str) -> tuple[dict[str, str], dict[str, object]]: tag = _normalize_lang_tag(lang) entries = _collect_feed_i18n_entries(feed) mapping: dict[str, str] = {} total = len(entries) translated = 0 if tag == "zh-Hant": for entry_key, text in entries: mapping[entry_key] = text return mapping, {"lang": tag, "total": total, "translated": total, "coverage": 1.0, "mode": "base-cache"} if tag == "zh-Hans": for entry_key, text in entries: out = _to_zh_hans(text) mapping[entry_key] = out return mapping, {"lang": tag, "total": total, "translated": total, "coverage": 1.0, "mode": "local"} if not TRANSLATION_CACHE_ENABLED: for entry_key, text in entries: mapping[entry_key] = text return mapping, { "lang": tag, "total": total, "translated": 0, "coverage": 0.0 if total else 1.0, "mode": "no-cache-best-effort", } with TRANSLATION_LOCK: _load_translation_cache_unlocked() cache = TRANSLATION_CACHE dirty = False for entry_key, text in entries: key = _translation_cache_key(tag, text, entry_key=entry_key) out = str(cache.get(key) or "").strip() if not out: out = str(cache.get(_translation_cache_key_legacy(tag, text)) or "").strip() if out and _is_untranslated_for_lang(text, out, tag): cache.pop(key, None) cache.pop(_translation_cache_key_legacy(tag, text), None) out = "" dirty = True if out: mapping[entry_key] = out translated += 1 else: mapping[entry_key] = text if dirty: global TRANSLATION_CACHE_DIRTY TRANSLATION_CACHE_DIRTY = True _flush_translation_cache_unlocked() coverage = round((translated / total), 4) if total else 1.0 return mapping, { "lang": tag, "total": total, "translated": translated, "coverage": coverage, "mode": "cache-best-effort", } def _best_effort_localized_feed(feed: dict[str, object], lang: str) -> tuple[dict[str, object], dict[str, object]]: tag = _normalize_lang_tag(lang) mapping, qa = _build_best_effort_mapping(feed, tag) localized = _apply_feed_translation(copy.deepcopy(feed), mapping) if isinstance(localized, dict): localized["lang"] = tag return localized, qa out = copy.deepcopy(feed) out["lang"] = tag return out, qa def _card_lookup_key(card: dict[str, object], index: int) -> str: for key in ("id", "url", "_card_key"): value = str(card.get(key) or "").strip() if value: return value account = str(card.get("account") or "").strip().lower() published = str(card.get("published_at") or "").strip() title = str(card.get("title") or "").strip() return f"{account}|{published}|{title}|{index}" def _iter_card_translatable_pairs( base_card: dict[str, object] | None, localized_card: dict[str, object], ) -> list[tuple[str, str]]: pairs: list[tuple[str, str]] = [] src_card = base_card if isinstance(base_card, dict) else localized_card dst_card = localized_card if isinstance(localized_card, dict) else {} for key in I18N_CARD_TEXT_KEYS: source = str(src_card.get(key) or "").strip() translated = str(dst_card.get(key) or "").strip() if not source and not translated: continue pairs.append((source, translated)) for key in I18N_FEED_LIST_KEYS: src_rows = src_card.get(key) if isinstance(src_card.get(key), list) else [] dst_rows = dst_card.get(key) if isinstance(dst_card.get(key), list) else [] total = max(len(src_rows), len(dst_rows)) for idx in range(total): source = str(src_rows[idx] or "").strip() if idx < len(src_rows) else "" translated = str(dst_rows[idx] or "").strip() if idx < len(dst_rows) else "" if not source and not translated: continue pairs.append((source, translated)) src_facts = src_card.get("event_facts") if isinstance(src_card.get("event_facts"), dict) else {} dst_facts = dst_card.get("event_facts") if isinstance(dst_card.get("event_facts"), dict) else {} for fact_key in I18N_CARD_FACT_TEXT_KEYS: source = str(src_facts.get(fact_key) or "").strip() translated = str(dst_facts.get(fact_key) or "").strip() if not source and not translated: continue pairs.append((source, translated)) return pairs def _card_has_target_text( card: dict[str, object], lang: str, base_card: dict[str, object] | None = None, ) -> bool: tag = _normalize_lang_tag(lang) if tag in {"en", "ko"}: src_card = base_card if isinstance(base_card, dict) else card # EN/KO publish gating is based on core card text only. # Secondary lists/facts may lag and should not hide the entire card. for key in I18N_CARD_TEXT_KEYS: source = str(src_card.get(key) or "").strip() translated = str(card.get(key) or "").strip() if not source and not translated: continue if _is_untranslated_for_lang(source, translated, tag): return False return True if tag in {"zh-Hant", "zh-Hans"}: if tag == "zh-Hans": sample = " ".join(str(card.get(k) or "") for k in I18N_CARD_TEXT_KEYS).strip() if not sample: return True return _to_zh_hans(sample) == sample return True sample = " ".join(str(card.get(k) or "") for k in I18N_CARD_TEXT_KEYS).strip() if not sample: return True return _looks_translated_for_lang(sample, tag) def _apply_card_level_fallback( *, base_feed: dict[str, object], localized_feed: dict[str, object], lang: str, mode: str, ) -> tuple[dict[str, object], dict[str, int]]: base_cards = base_feed.get("cards") localized_cards = localized_feed.get("cards") if not isinstance(base_cards, list) or not isinstance(localized_cards, list): return localized_feed, {"total": 0, "ready": 0, "fallback": 0, "hidden": 0} normalized_mode = "hide" if str(mode or "").strip().lower() == "hide" else "base" base_index: dict[str, dict[str, object]] = {} localized_index: dict[str, dict[str, object]] = {} for idx, raw in enumerate(base_cards): if not isinstance(raw, dict): continue base_index[_card_lookup_key(raw, idx)] = raw for idx, raw in enumerate(localized_cards): if not isinstance(raw, dict): continue key = _card_lookup_key(raw, idx) if key not in localized_index: localized_index[key] = raw out_cards: list[dict[str, object]] = [] total = 0 ready = 0 fallback = 0 hidden = 0 for idx, base_raw in enumerate(base_cards): if not isinstance(base_raw, dict): continue key = _card_lookup_key(base_raw, idx) localized_raw = localized_index.get(key) total += 1 if isinstance(localized_raw, dict) and _card_has_target_text(localized_raw, lang, base_card=base_raw): ready += 1 out_cards.append(localized_raw) continue if normalized_mode == "hide": hidden += 1 continue base_card = base_index.get(key) if isinstance(base_card, dict): fallback += 1 out_cards.append(copy.deepcopy(base_card)) else: hidden += 1 out = copy.deepcopy(localized_feed) out["cards"] = out_cards out["total_cards"] = len(out_cards) return out, {"total": total, "ready": ready, "fallback": fallback, "hidden": hidden} def _resolve_card_fallback_mode(lang: str) -> str: # Card-level publishing mode: # - hide: only show cards already translated for target language # - base: show translated cards first, fallback to base-language card per item if _normalize_lang_tag(lang) in {"en", "ko"}: return "hide" raw_mode = str(I18N_FEED_FALLBACK_MODE or "base").strip().lower() if raw_mode in {"hide", "strict", "translated-only"}: return "hide" return "base" def _fallback_feed_from_base(feed: dict[str, object], lang: str, reason: str) -> dict[str, object]: out = copy.deepcopy(feed) out["lang"] = _normalize_lang_tag(lang) out["_i18n"] = { "mode": "base-fallback", "source_generated_at": str(feed.get("generated_at") or ""), "qa": {"coverage": 0.0, "reason": reason}, "state": _i18n_state_snapshot(), } return out def _is_untranslated_for_lang(source: str, translated: str, lang: str) -> bool: src = str(source or "").strip() dst = str(translated or "").strip() if _normalize_lang_tag(lang) in {"en", "ko"}: if not dst: return True # For EN/KO, reject CJK echo outputs so they don't poison cache. if _contains_cjk(dst): return True # Source has CJK but output still doesn't look like target language. if _contains_cjk(src) and not _looks_translated_for_lang(dst, _normalize_lang_tag(lang)): return True return False return False def _translate_chunk_agent(texts: list[str], lang: str, api_key: str, strict: bool = False) -> list[str]: if not texts: return [] lang_name = _lang_prompt_name(lang) strict_rules = ( "5) 若原文是中文，目標語言是 English/한국어 時，不可原樣保留中文，必須翻譯。\n" if strict else "" ) prompt = ( f"你是高精度翻譯代理。把輸入 JSON 陣列每一項翻成 {lang_name}。\n" "規則：\n" "1) 順序完全一致，不可遺漏。\n" "2) 保留 URL、@帳號、#標籤、數字、時間、貨幣與專有名詞（Renaiss/SBT/BNB/Discord）。\n" "3) 只輸出 JSON 字串陣列，不要任何說明。\n" "4) 語意必須完整，不可只翻一半。\n" f"{strict_rules}\n" f"source={json.dumps(texts, ensure_ascii=False)}" ) max_attempts = int(str(os.getenv("I18N_TRANSLATE_RETRY_MAX_ATTEMPTS") or "0") or 0) retry_forever = str(os.getenv("I18N_TRANSLATE_RETRY_FOREVER", "1") or "1").strip().lower() in {"1", "true", "yes", "on"} attempt = 0 while True: attempt += 1 raw = minimax_chat(prompt, api_key) parsed = _parse_json_array(raw) if len(parsed) == len(texts): return [str(x or "").strip() or texts[idx] for idx, x in enumerate(parsed)] if not retry_forever and max_attempts > 0 and attempt >= max_attempts: raise RuntimeError(f"translate chunk parse mismatch: expected={len(texts)} got={len(parsed)}") time.sleep(min(12.0, 1.2 * (1.5 ** max(0, attempt - 1)))) def _translate_feed_text_map( entries: list[tuple[str, str]] | list[str], lang: str, previous_source_map: dict[str, str] | None = None, previous_translation_map: dict[str, str] | None = None, ) -> tuple[dict[str, str], dict[str, object]]: tag = _normalize_lang_tag(lang) normalized_entries: list[tuple[str, str]] = [] seen_entry_keys: set[str] = set() if entries and isinstance(entries[0], str): # backward-compatible input for idx, text in enumerate(entries): # type: ignore[arg-type] source = str(text or "") if not source: continue entry_key = f"legacy[{idx}]" if entry_key in seen_entry_keys: continue seen_entry_keys.add(entry_key) normalized_entries.append((entry_key, source)) else: for row in entries: # type: ignore[assignment] if not isinstance(row, (list, tuple)) or len(row) < 2: continue entry_key = str(row[0] or "").strip() source = str(row[1] or "") if not entry_key or not source or entry_key in seen_entry_keys: continue seen_entry_keys.add(entry_key) normalized_entries.append((entry_key, source)) mapping = {entry_key: source for entry_key, source in normalized_entries} qa: dict[str, object] = { "lang": tag, "total": len(normalized_entries), "cached_hits": 0, "reused_hits": 0, "pending_count": 0, "translated": 0, "coverage": 0.0, "mode": "base" if tag == "zh-Hant" else "pending", "sample_pending": [], } if tag == "zh-Hant" or not normalized_entries: qa["translated"] = len(normalized_entries) qa["coverage"] = 1.0 _set_i18n_lang_progress( tag, total=len(normalized_entries), done=len(normalized_entries), status="ok", mode=str(qa["mode"]), sample_pending=[], ) return mapping, qa if tag == "zh-Hans": mapping = {entry_key: _to_zh_hans(source) for entry_key, source in normalized_entries} qa["mode"] = "local" qa["translated"] = len(normalized_entries) qa["coverage"] = 1.0 _set_i18n_lang_progress( tag, total=len(normalized_entries), done=len(normalized_entries), status="ok", mode="local", sample_pending=[], ) return mapping, qa prev_sources = previous_source_map if isinstance(previous_source_map, dict) else {} prev_translations = previous_translation_map if isinstance(previous_translation_map, dict) else {} pending_entries: list[tuple[str, str]] = [] reused_hits = 0 for entry_key, source in normalized_entries: old_source = str(prev_sources.get(entry_key) or "") old_translated = str(prev_translations.get(entry_key) or "").strip() if old_translated and not _is_untranslated_for_lang(source, old_translated, tag): # Normal incremental reuse: source text unchanged. if old_source and old_source == source: mapping[entry_key] = old_translated reused_hits += 1 continue # Legacy migration path: old bundles may not have `sources`. if not old_source: mapping[entry_key] = old_translated reused_hits += 1 continue pending_entries.append((entry_key, source)) qa["reused_hits"] = reused_hits qa["cached_hits"] = reused_hits if TRANSLATION_CACHE_ENABLED: with TRANSLATION_LOCK: _load_translation_cache_unlocked() migrated = False pending_next: list[tuple[str, str]] = [] for entry_key, source in pending_entries: key = _translation_cache_key(tag, source, entry_key=entry_key) legacy_key = _translation_cache_key_legacy(tag, source) cached = str(TRANSLATION_CACHE.get(key) or "").strip() if not cached: legacy = str(TRANSLATION_CACHE.get(legacy_key) or "").strip() if legacy: cached = legacy TRANSLATION_CACHE[key] = legacy migrated = True if cached and _is_untranslated_for_lang(source, cached, tag): TRANSLATION_CACHE.pop(key, None) TRANSLATION_CACHE.pop(legacy_key, None) cached = "" migrated = True if cached: mapping[entry_key] = cached qa["cached_hits"] = int(qa.get("cached_hits") or 0) + 1 continue pending_next.append((entry_key, source)) if migrated: global TRANSLATION_CACHE_DIRTY TRANSLATION_CACHE_DIRTY = True _flush_translation_cache_unlocked() pending_entries = pending_next translated_count = len(normalized_entries) - len(pending_entries) qa["translated"] = translated_count qa["pending_count"] = len(pending_entries) pending_entry_keys = [entry_key for entry_key, _ in pending_entries] if not pending_entries: qa["mode"] = "reused" if int(qa.get("reused_hits") or 0) > 0 else "cache" qa["coverage"] = 1.0 _set_i18n_lang_progress( tag, total=len(normalized_entries), done=len(normalized_entries), status="ok", mode=str(qa.get("mode") or "cache"), sample_pending=[], ) return mapping, qa load_environment() api_key = resolve_minimax_key() if not api_key: still_pending: list[tuple[str, str]] = [] reused_without_key = 0 for entry_key, source in pending_entries: old_translated = str(prev_translations.get(entry_key) or "").strip() if old_translated and not _is_untranslated_for_lang(source, old_translated, tag): mapping[entry_key] = old_translated reused_without_key += 1 else: still_pending.append((entry_key, source)) if reused_without_key: qa["reused_hits"] = int(qa.get("reused_hits") or 0) + reused_without_key qa["cached_hits"] = int(qa.get("cached_hits") or 0) + reused_without_key translated_count += reused_without_key pending_entries = still_pending qa["translated"] = translated_count qa["pending_count"] = len(pending_entries) qa["mode"] = "no-key" qa["coverage"] = round((translated_count / len(normalized_entries)), 4) if normalized_entries else 1.0 qa["sample_pending"] = [entry_key for entry_key, _ in pending_entries[:8]] _set_i18n_lang_progress( tag, total=len(normalized_entries), done=translated_count, status="pending", mode="no-key", error="missing MiniMax key", sample_pending=[str(x) for x in qa["sample_pending"] if str(x).strip()][:8], ) return mapping, qa entry_keys_by_source: dict[str, list[str]] = {} for entry_key, source in pending_entries: entry_keys_by_source.setdefault(source, []).append(entry_key) pending_texts = list(entry_keys_by_source.keys()) chunk_size = I18N_FEED_CHUNK_SIZE done_count = translated_count unresolved_keys: set[str] = set() _set_i18n_lang_progress( tag, total=len(normalized_entries), done=done_count, status="running", mode="live", sample_pending=pending_entry_keys[:8], ) for start in range(0, len(pending_texts), chunk_size): chunk = pending_texts[start:start + chunk_size] translated = _translate_chunk_agent(chunk, tag, api_key, strict=False) for idx, source in enumerate(chunk): candidate = str(translated[idx] if idx < len(translated) else source).strip() or source source_entry_keys = entry_keys_by_source.get(source, []) if _is_untranslated_for_lang(source, candidate, tag): for entry_key in source_entry_keys: old_translated = str(prev_translations.get(entry_key) or "").strip() if old_translated and not _is_untranslated_for_lang(source, old_translated, tag): mapping[entry_key] = old_translated done_count += 1 qa["reused_hits"] = int(qa.get("reused_hits") or 0) + 1 qa["cached_hits"] = int(qa.get("cached_hits") or 0) + 1 else: mapping[entry_key] = source unresolved_keys.add(entry_key) continue for entry_key in source_entry_keys: mapping[entry_key] = candidate done_count += 1 _set_i18n_lang_progress( tag, total=len(normalized_entries), done=done_count, status="running", mode="live", sample_pending=pending_entry_keys[:8], ) if TRANSLATION_CACHE_ENABLED: with TRANSLATION_LOCK: _load_translation_cache_unlocked() cache_updated = False for entry_key, source in pending_entries: if entry_key in unresolved_keys: continue translated = str(mapping.get(entry_key, source) or "").strip() or source if _is_untranslated_for_lang(source, translated, tag): continue key = _translation_cache_key(tag, source, entry_key=entry_key) if TRANSLATION_CACHE.get(key) != translated: TRANSLATION_CACHE[key] = translated cache_updated = True if cache_updated: TRANSLATION_CACHE_DIRTY = True _flush_translation_cache_unlocked() if unresolved_keys: translated_final = max(0, len(normalized_entries) - len(unresolved_keys)) pending_final = len(unresolved_keys) qa["mode"] = "live-partial" qa["translated"] = translated_final qa["pending_count"] = pending_final qa["coverage"] = round((translated_final / len(normalized_entries)), 4) if normalized_entries else 1.0 qa["sample_pending"] = sorted(list(unresolved_keys))[:8] _set_i18n_lang_progress( tag, total=len(normalized_entries), done=translated_final, status="pending", mode="live-partial", sample_pending=sorted(list(unresolved_keys))[:8], ) return mapping, qa qa["mode"] = "live" qa["translated"] = len(normalized_entries) qa["pending_count"] = 0 qa["coverage"] = 1.0 _set_i18n_lang_progress( tag, total=len(normalized_entries), done=len(normalized_entries), status="ok", mode="live", sample_pending=[], ) return mapping, qa def _write_i18n_feed_bundle(bundle: dict[str, object]) -> None: I18N_FEED_PATH.parent.mkdir(parents=True, exist_ok=True) I18N_FEED_PATH.write_text(json.dumps(bundle, ensure_ascii=False, indent=2), encoding="utf-8") def _load_i18n_feed_bundle() -> dict[str, object]: if not I18N_FEED_PATH.exists(): return {} try: raw = json.loads(I18N_FEED_PATH.read_text(encoding="utf-8")) except Exception: return {} return raw if isinstance(raw, dict) else {} def _i18n_state_snapshot() -> dict[str, object]: def _progress_from_qa(tag: str, qa: dict[str, object], total_hint: int) -> dict[str, object]: total = _safe_int(qa.get("total"), total_hint) if total <= 0: total = total_hint translated = _safe_int(qa.get("translated"), _safe_int(qa.get("done"), 0)) if translated <= 0 and total > 0: coverage = float(qa.get("coverage") or 0.0) if qa else 0.0 translated = int(round(total * max(0.0, min(1.0, coverage)))) done = total if tag == "zh-Hant" else min(total, max(0, translated)) status = "ok" if done >= total else "pending" mode = str(qa.get("mode") or ("base" if tag == "zh-Hant" else "")) percent = 100 if total == 0 and status == "ok" else round((done / total) * 100, 1) if total else 0 safe_done = min(done, total) return { "lang": tag, "status": status, "mode": mode, "total": total, "done": safe_done, "remaining": max(0, total - safe_done), "cached_hits": _safe_int(qa.get("cached_hits"), 0), "pending_count": _safe_int(qa.get("pending_count"), max(0, total - safe_done)), "sample_pending": [str(x) for x in (qa.get("sample_pending") or []) if str(x).strip()][:8], "percent": percent, "error": "", "updated_at": str(qa.get("updated_at") or ""), } with I18N_STATE_LOCK: state = copy.deepcopy(I18N_BUILD_STATE) progress = state.get("lang_progress") if not isinstance(progress, dict): progress = {} else: progress = copy.deepcopy(progress) bundle = _load_i18n_feed_bundle() qa_rows = bundle.get("qa") if isinstance(bundle.get("qa"), dict) else {} total_hint = _safe_int(bundle.get("targets_count"), 0) if total_hint <= 0: try: feed = _read_feed_snapshot() total_hint = len(_collect_feed_i18n_strings(feed)) if isinstance(feed, dict) else 0 except Exception: total_hint = 0 for tag in I18N_TARGET_LANGS: if tag == "zh-Hant": progress[tag] = { "lang": tag, "status": "ok", "mode": "base", "total": total_hint, "done": total_hint, "remaining": 0, "cached_hits": total_hint, "pending_count": 0, "sample_pending": [], "percent": 100 if total_hint else 100, "error": "", "updated_at": "", } continue qa = qa_rows.get(tag) if isinstance(qa_rows.get(tag), dict) else {} if qa and tag not in progress: progress[tag] = _progress_from_qa(tag, qa, total_hint) elif qa and str(progress.get(tag, {}).get("status") or "").lower() not in {"running", "queued"}: progress[tag] = _progress_from_qa(tag, qa, total_hint) elif tag not in progress: progress[tag] = { "lang": tag, "status": "pending", "mode": "", "total": total_hint, "done": 0, "remaining": total_hint, "cached_hits": 0, "pending_count": total_hint, "sample_pending": [], "percent": 0, "error": "", "updated_at": "", } state["lang_progress"] = progress state["target_langs"] = list(I18N_TARGET_LANGS) if isinstance(bundle.get("langs"), dict): state["langs"] = sorted([str(x) for x in bundle["langs"].keys() if str(x).strip()]) current_status = str(state.get("status") or "").strip().lower() if current_status not in {"running", "queued", "failed"}: non_base_rows = [ progress.get(tag) for tag in I18N_TARGET_LANGS if tag != "zh-Hant" and isinstance(progress.get(tag), dict) ] non_base_statuses = [str(row.get("status") or "").strip().lower() for row in non_base_rows] if non_base_statuses: if any(st in {"running", "queued"} for st in non_base_statuses): state["status"] = "running" elif any(st == "failed" for st in non_base_statuses): state["status"] = "failed" elif any(st in {"pending"} for st in non_base_statuses): state["status"] = "pending" elif all(st == "ok" for st in non_base_statuses): state["status"] = "ok" return state def _set_i18n_lang_progress( lang: str, *, total: int, done: int, status: str, mode: str = "", error: str = "", sample_pending: list[str] | None = None, ) -> None: tag = _normalize_lang_tag(lang) safe_total = max(0, int(total or 0)) safe_done = min(max(0, int(done or 0)), safe_total) percent = 100 if safe_total == 0 and status == "ok" else round((safe_done / safe_total) * 100, 1) if safe_total else 0 with I18N_STATE_LOCK: progress = I18N_BUILD_STATE.get("lang_progress") if not isinstance(progress, dict): progress = {} I18N_BUILD_STATE["lang_progress"] = progress current = progress.get(tag) if isinstance(progress.get(tag), dict) else {} existing_pending = [str(x) for x in (current.get("sample_pending") or []) if str(x).strip()] if isinstance(current, dict) else [] sample_rows = [str(x) for x in (sample_pending or []) if str(x).strip()][:8] if sample_pending is not None else existing_pending[:8] progress[tag] = { "lang": tag, "status": status, "mode": mode, "total": safe_total, "done": safe_done, "remaining": max(0, safe_total - safe_done), "cached_hits": _safe_int(progress.get(tag, {}).get("cached_hits"), 0) if isinstance(progress.get(tag), dict) else 0, "pending_count": max(0, safe_total - safe_done), "sample_pending": sample_rows, "percent": percent, "error": str(error or ""), "updated_at": _now_iso(), } def _is_active_i18n_source(source_generated_at: str) -> bool: expected = str(source_generated_at or "").strip() if not expected: return True with I18N_STATE_LOCK: status = str(I18N_BUILD_STATE.get("status") or "").strip().lower() current = str(I18N_BUILD_STATE.get("source_generated_at") or "").strip() if status != "running": return True return current == expected def _build_i18n_feed_bundle( feed: dict[str, object], force: bool = False, target_langs: list[str] | tuple[str, ...] | None = None, ) -> dict[str, object]: src_generated = str(feed.get("generated_at") or "").strip() request_langs = [ _normalize_lang_tag(x) for x in (target_langs or I18N_TARGET_LANGS) if str(x or "").strip() ] seen_langs: set[str] = set() normalized_targets: list[str] = [] for tag in request_langs: if tag in seen_langs: continue seen_langs.add(tag) normalized_targets.append(tag) if not normalized_targets: normalized_targets = list(I18N_TARGET_LANGS) if "zh-Hant" not in normalized_targets: normalized_targets.insert(0, "zh-Hant") with I18N_LOCK: cached = _load_i18n_feed_bundle() cache_version_ok = int(cached.get("version") or 0) == I18N_BUILD_VERSION if isinstance(cached, dict) else False cached_source_generated = str(cached.get("source_generated_at") or "").strip() if isinstance(cached, dict) else "" cache_source_ok = bool(cache_version_ok and cached and cached_source_generated == src_generated) cached_langs = cached.get("langs") if isinstance(cached.get("langs"), dict) else {} cached_qa = cached.get("qa") if isinstance(cached.get("qa"), dict) else {} cached_sources = cached.get("sources") if isinstance(cached.get("sources"), dict) else {} if not isinstance(cached_langs, dict): cached_langs = {} if not isinstance(cached_qa, dict): cached_qa = {} if not isinstance(cached_sources, dict): cached_sources = {} has_all_targets = all(isinstance(cached_langs.get(tag), dict) for tag in normalized_targets) if not force and cache_source_ok and has_all_targets: return cached entries = _collect_feed_i18n_entries(feed) entry_source_map = {str(entry_key): str(source or "") for entry_key, source in entries if str(entry_key or "").strip()} langs_payload: dict[str, object] = dict(cached_langs) qa_payload: dict[str, object] = dict(cached_qa) sources_payload: dict[str, object] = dict(cached_sources) bundle: dict[str, object] = { "version": I18N_BUILD_VERSION, "generated_at": _now_iso(), "source_generated_at": src_generated, "langs": langs_payload, "qa": qa_payload, "sources": sources_payload, "targets_count": len(entries), } for tag in normalized_targets: if not _is_active_i18n_source(src_generated): return _load_i18n_feed_bundle() previous_source_map = sources_payload.get(tag) if isinstance(sources_payload.get(tag), dict) else {} if not previous_source_map and cached_source_generated == src_generated: previous_source_map = dict(entry_source_map) previous_translation_map = _entry_map_from_node(langs_payload.get(tag)) if isinstance(langs_payload.get(tag), dict) else {} mapping, qa = _translate_feed_text_map( entries, tag, previous_source_map=previous_source_map, previous_translation_map=previous_translation_map, ) if not _is_active_i18n_source(src_generated): return _load_i18n_feed_bundle() localized = _apply_feed_translation(copy.deepcopy(feed), mapping) if isinstance(localized, dict): localized["lang"] = tag langs_payload[tag] = localized qa_payload[tag] = qa sources_payload[tag] = dict(entry_source_map) # Persist per-language progress so completed langs can be served # immediately while other langs are still translating. bundle = { "version": I18N_BUILD_VERSION, "generated_at": _now_iso(), "source_generated_at": src_generated, "langs": dict(langs_payload), "qa": dict(qa_payload), "sources": dict(sources_payload), "targets_count": len(entries), } if not _is_active_i18n_source(src_generated): return _load_i18n_feed_bundle() _write_i18n_feed_bundle(bundle) bundle = { "version": I18N_BUILD_VERSION, "generated_at": _now_iso(), "source_generated_at": src_generated, "langs": dict(langs_payload), "qa": dict(qa_payload), "sources": dict(sources_payload), "targets_count": len(entries), } if not _is_active_i18n_source(src_generated): return _load_i18n_feed_bundle() _write_i18n_feed_bundle(bundle) return bundle def _build_i18n_feed_bundle_async( feed: dict[str, object], force: bool = False, target_langs: list[str] | tuple[str, ...] | None = None, ) -> None: if not isinstance(feed, dict): return src_generated = str(feed.get("generated_at") or "").strip() requested = [ _normalize_lang_tag(x) for x in (target_langs or I18N_TARGET_LANGS) if str(x or "").strip() ] seen_tags: set[str] = set() target_tags: list[str] = [] for tag in requested: if tag in seen_tags: continue seen_tags.add(tag) target_tags.append(tag) if not target_tags: target_tags = list(I18N_TARGET_LANGS) with I18N_STATE_LOCK: status = str(I18N_BUILD_STATE.get("status") or "idle").strip().lower() running_for = str(I18N_BUILD_STATE.get("source_generated_at") or "").strip() if status == "running" and running_for == src_generated: current_targets = I18N_BUILD_STATE.get("target_langs") if not isinstance(current_targets, list): current_targets = [] merged_targets = list(current_targets) for tag in target_tags: if tag not in merged_targets: merged_targets.append(tag) progress = I18N_BUILD_STATE.get("lang_progress") if not isinstance(progress, dict): progress = {} I18N_BUILD_STATE["lang_progress"] = progress progress[tag] = { "lang": tag, "status": "queued", "mode": "", "total": 0, "done": 0, "remaining": 0, "percent": 0, "error": "", "updated_at": _now_iso(), } I18N_BUILD_STATE["target_langs"] = merged_targets return I18N_BUILD_STATE.update( { "status": "running", "started_at": _now_iso(), "finished_at": "", "last_error": "", "source_generated_at": src_generated, "target_langs": list(target_tags), } ) def _worker() -> None: current_targets = list(target_tags) try: while current_targets: with I18N_STATE_LOCK: if str(I18N_BUILD_STATE.get("source_generated_at") or "").strip() != src_generated: return bundle = _build_i18n_feed_bundle(feed, force=force, target_langs=current_targets) langs = [] raw_langs = bundle.get("langs") if isinstance(raw_langs, dict): langs = sorted([str(x) for x in raw_langs.keys() if str(x).strip()]) with I18N_STATE_LOCK: if str(I18N_BUILD_STATE.get("source_generated_at") or "").strip() != src_generated: return requested = [ str(x) for x in (I18N_BUILD_STATE.get("target_langs") or []) if str(x).strip() ] pending = [x for x in requested if x not in langs] if pending: current_targets = pending continue I18N_BUILD_STATE.update( { "status": "ok", "finished_at": _now_iso(), "last_error": "", "source_generated_at": src_generated, "langs": langs, } ) current_targets = [] except Exception as exc: with I18N_STATE_LOCK: I18N_BUILD_STATE.update( { "status": "failed", "finished_at": _now_iso(), "last_error": str(exc), "source_generated_at": src_generated, } ) Thread(target=_worker, daemon=True).start() def _localized_feed_from_bundle(feed: dict[str, object], lang: str) -> dict[str, object]: tag = _normalize_lang_tag(lang) if tag == "zh-Hant": out = copy.deepcopy(feed) if isinstance(out, dict): out["lang"] = "zh-Hant" out["_i18n"] = {"mode": "base", "qa": {"coverage": 1.0}} return out if isinstance(out, dict) else dict(feed) def _base_building_response(reason: str) -> dict[str, object]: out = copy.deepcopy(feed) if not isinstance(out, dict): out = dict(feed) cards = out.get("cards") total_cards = len(cards) if isinstance(cards, list) else 0 if tag in {"en", "ko"}: out["cards"] = [] out["total_cards"] = 0 card_state = {"total": total_cards, "ready": 0, "fallback": 0, "hidden": total_cards} else: card_state = {"total": total_cards, "ready": total_cards, "fallback": 0, "hidden": 0} out["lang"] = tag out["_i18n"] = { "mode": "building", "source_generated_at": str(feed.get("generated_at") or ""), "qa": { "lang": tag, "coverage": 0.0, "reason": reason, "card_state": card_state, }, "state": _i18n_state_snapshot(), } return out def _stale_building_response(bundle_obj: dict[str, object], reason: str) -> dict[str, object] | None: if not isinstance(bundle_obj, dict): return None langs = bundle_obj.get("langs") if isinstance(bundle_obj.get("langs"), dict) else {} stale = langs.get(tag) if isinstance(langs, dict) else None if not isinstance(stale, dict): return None out = copy.deepcopy(stale) if not isinstance(out, dict): return None fallback_mode = _resolve_card_fallback_mode(tag) out, card_state = _apply_card_level_fallback( base_feed=feed, localized_feed=out, lang=tag, mode=fallback_mode, ) qa_rows = bundle_obj.get("qa") if isinstance(bundle_obj.get("qa"), dict) else {} qa_for_tag = qa_rows.get(tag) if isinstance(qa_rows.get(tag), dict) else {} qa_with_cards = dict(qa_for_tag or {}) qa_with_cards["reason"] = reason qa_with_cards["card_state"] = card_state out["lang"] = tag out["_i18n"] = { "mode": "building-stale", "source_generated_at": str(feed.get("generated_at") or ""), "qa": qa_with_cards, "state": _i18n_state_snapshot(), } return out bundle = _load_i18n_feed_bundle() src_generated = str(feed.get("generated_at") or "").strip() bundle_generated = str(bundle.get("source_generated_at") or "").strip() if isinstance(bundle, dict) else "" bundle_version_ok = int(bundle.get("version") or 0) == I18N_BUILD_VERSION if isinstance(bundle, dict) else False if not bundle or bundle_generated != src_generated or not bundle_version_ok: _build_i18n_feed_bundle_async(feed, force=False, target_langs=[tag]) stale = _stale_building_response(bundle if isinstance(bundle, dict) else {}, "bundle_missing_or_outdated") if isinstance(stale, dict): return stale return _base_building_response("bundle_missing_or_outdated") langs = bundle.get("langs") if isinstance(bundle.get("langs"), dict) else {} qa_rows = bundle.get("qa") if isinstance(bundle.get("qa"), dict) else {} localized = langs.get(tag) qa_for_tag = qa_rows.get(tag) if isinstance(qa_rows.get(tag), dict) else {} coverage = float(qa_for_tag.get("coverage") or 0.0) if isinstance(qa_for_tag, dict) else 0.0 acceptable = tag in {"zh-Hant", "zh-Hans"} or coverage >= I18N_MIN_ACCEPTABLE_COVERAGE if not isinstance(localized, dict): _build_i18n_feed_bundle_async(feed, force=False, target_langs=[tag]) stale = _stale_building_response(bundle, "lang_not_ready") if isinstance(stale, dict): return stale return _base_building_response("lang_not_ready") out = copy.deepcopy(localized) if not isinstance(out, dict): return _base_building_response("lang_payload_invalid") out["lang"] = tag base_cards = feed.get("cards") out_cards = out.get("cards") if not isinstance(out_cards, list) and isinstance(base_cards, list): out_cards = copy.deepcopy(base_cards) out["cards"] = out_cards fallback_mode = _resolve_card_fallback_mode(tag) out, card_state = _apply_card_level_fallback( base_feed=feed, localized_feed=out, lang=tag, mode=fallback_mode, ) qa_with_cards = dict(qa_for_tag or {}) qa_with_cards["card_state"] = card_state out["_i18n"] = { "mode": ( "pretranslated" if acceptable and card_state.get("fallback", 0) == 0 and card_state.get("hidden", 0) == 0 else "pretranslated-partial" ), "source_generated_at": str(bundle.get("source_generated_at") or ""), "qa": qa_with_cards, "state": _i18n_state_snapshot(), } return out def _normalize_retranslate_langs(langs: list[str] | tuple[str, ...] | str | None) -> list[str]: raw_rows: list[str] if isinstance(langs, str): raw_rows = [x.strip() for x in langs.split(",") if x.strip()] elif isinstance(langs, (list, tuple)): raw_rows = [str(x).strip() for x in langs if str(x).strip()] else: raw_rows = [] if not raw_rows: raw_rows = ["en", "ko", "zh-Hans"] out: list[str] = [] seen: set[str] = set() for item in raw_rows: tag = _normalize_lang_tag(item) if tag == "zh-Hant": continue if tag in seen: continue seen.add(tag) out.append(tag) return out def _clear_translation_cache_for_langs(target_langs: list[str]) -> int: tags = set(_normalize_retranslate_langs(target_langs)) if not tags: return 0 removed = 0 with TRANSLATION_LOCK: _load_translation_cache_unlocked() keys = list(TRANSLATION_CACHE.keys()) for key in keys: row = str(key or "") if any(row.startswith(f"{tag}\n") for tag in tags): TRANSLATION_CACHE.pop(key, None) removed += 1 if removed: global TRANSLATION_CACHE_DIRTY TRANSLATION_CACHE_DIRTY = True _flush_translation_cache_unlocked(force=True) return removed def _drop_bundle_langs(target_langs: list[str]) -> None: tags = set(_normalize_retranslate_langs(target_langs)) if not tags: return with I18N_LOCK: bundle = _load_i18n_feed_bundle() if not bundle: return langs = bundle.get("langs") qa = bundle.get("qa") sources = bundle.get("sources") changed = False if isinstance(langs, dict): for tag in tags: if tag in langs: langs.pop(tag, None) changed = True if isinstance(qa, dict): for tag in tags: if tag in qa: qa.pop(tag, None) changed = True if isinstance(sources, dict): for tag in tags: if tag in sources: sources.pop(tag, None) changed = True if changed: bundle["generated_at"] = _now_iso() _write_i18n_feed_bundle(bundle) def _queue_i18n_retranslate( feed: dict[str, object], target_langs: list[str] | tuple[str, ...] | str | None = None, ) -> dict[str, object]: tags = _normalize_retranslate_langs(target_langs) if not tags: return {"langs": [], "cache_removed": 0, "queued": False} removed = _clear_translation_cache_for_langs(tags) for tag in tags: _set_i18n_lang_progress(tag, total=0, done=0, status="pending", mode="manual", sample_pending=[]) _build_i18n_feed_bundle_async(feed, force=True, target_langs=tags) return { "langs": tags, "cache_removed": removed, "queued": True, "source_generated_at": str(feed.get("generated_at") or ""), } def _rebuild_i18n_bundle_sync( feed: dict[str, object], target_langs: list[str] | tuple[str, ...] | str | None = None, force: bool = True, ) -> dict[str, object]: requested = target_langs if isinstance(target_langs, str): requested = [x.strip() for x in target_langs.split(",") if x.strip()] rows = [str(x).strip() for x in (requested or I18N_TARGET_LANGS) if str(x).strip()] if "zh-Hant" not in rows: rows = ["zh-Hant", *rows] return _build_i18n_feed_bundle(feed, force=bool(force), target_langs=rows) # Public aliases used by ai_intel_server.py. Keep the internal helper names # stable so old cached jobs/status payloads continue to match existing keys. translate_texts = _translate_texts build_i18n_feed_bundle_async = _build_i18n_feed_bundle_async localized_feed_from_bundle = _localized_feed_from_bundle i18n_state_snapshot = _i18n_state_snapshot queue_i18n_retranslate = _queue_i18n_retranslate rebuild_i18n_bundle_sync = _rebuild_i18n_bundle_sync