@lightcone-ai/daemon 0.15.52 → 0.15.54

package/src/_vendor/video/understanding/heuristics.js

@@ -1,826 +0,0 @@
-const RECRUITMENT_KEYWORDS = /招聘|校招|春招|秋招|实习|应届|岗位|职位|网申|career|job|intern|campus/i;
-const CAMPUS_KEYWORDS = /校招|校园招聘|应届|毕业生|春招|秋招|campus/i;
-const INTERN_KEYWORDS = /实习|intern/i;
-const EXPERIENCED_KEYWORDS = /社招|社会招聘|experienced|experienced hire|社会人才/i;
-
-const JOB_DIRECTION_HINTS = Object.freeze([
-  '算法',
-  '研发',
-  '开发',
-  '前端',
-  '后端',
-  '测试',
-  '运维',
-  '产品',
-  '运营',
-  '市场',
-  '销售',
-  '设计',
-  '财务',
-  '人力',
-  'hr',
-  '数据',
-  '供应链',
-  '法务',
-  '风控',
-  '投研',
-  '量化',
-  '内容',
-  '客服',
-]);
-
-const LOCATION_HINTS = Object.freeze([
-  '北京',
-  '上海',
-  '广州',
-  '深圳',
-  '杭州',
-  '成都',
-  '南京',
-  '武汉',
-  '西安',
-  '苏州',
-  '天津',
-  '重庆',
-  '长沙',
-  '郑州',
-  '青岛',
-  '厦门',
-  '合肥',
-  '宁波',
-  '佛山',
-  '东莞',
-  '珠海',
-  '无锡',
-  'remote',
-  '远程',
-]);
-
-export const RECRUITMENT_SLOT_KEYS = Object.freeze([
-  'company',
-  'published_at',
-  'recruitment_type',
-  'cohort',
-  'job_directions',
-  'locations',
-  'target_or_requirements',
-  'process',
-  'entry_or_cta',
-]);
-
-const MODE_MIN_CONFIDENCE = Object.freeze({
-  strong: 0.6,
-  support: 0.52,
-  weakSignal: 0.42,
-  floor: 0.36,
-});
-
-function cleanText(value) {
-  return String(value ?? '').replace(/\s+/g, ' ').trim();
-}
-
-function midpoint(range = [0, 0]) {
-  const start = Number(range[0] ?? 0);
-  const end = Number(range[1] ?? start);
-  return Math.round((start + end) / 2);
-}
-
-function clampNumber(value, min, max, fallback) {
-  const num = Number(value);
-  if (!Number.isFinite(num)) return fallback;
-  return Math.max(min, Math.min(max, num));
-}
-
-function toSourceType(value) {
-  const normalized = cleanText(value).toLowerCase();
-  if (normalized === 'title' || normalized === 'meta' || normalized === 'body_text' || normalized === 'image_ocr' || normalized === 'heuristic') {
-    return normalized;
-  }
-  return 'heuristic';
-}
-
-function baseConfidenceForSourceType(sourceType) {
-  const normalized = toSourceType(sourceType);
-  if (normalized === 'title') return 0.9;
-  if (normalized === 'meta') return 0.84;
-  if (normalized === 'body_text') return 0.74;
-  if (normalized === 'image_ocr') return 0.7;
-  return 0.45;
-}
-
-function hasValue(value) {
-  if (Array.isArray(value)) return value.length > 0;
-  return cleanText(value).length > 0;
-}
-
-function uniqueStrings(values = []) {
-  const seen = new Set();
-  const output = [];
-  for (const row of values) {
-    const text = cleanText(row);
-    if (!text) continue;
-    const key = text.toLowerCase();
-    if (seen.has(key)) continue;
-    seen.add(key);
-    output.push(text);
-  }
-  return output;
-}
-
-function normalizeFocusRegion(range) {
-  if (!Array.isArray(range) || range.length < 2) return null;
-  const start = Number(range[0]);
-  const end = Number(range[1]);
-  if (!Number.isFinite(start) || !Number.isFinite(end)) return null;
-  const low = Math.round(Math.min(start, end));
-  const high = Math.round(Math.max(start, end));
-  if (high <= low) return null;
-  return [low, high];
-}
-
-function resolveTopFocusRegion({ coreRange, focusRange }) {
-  const normalizedCore = normalizeFocusRegion(coreRange);
-  const normalizedFocus = normalizeFocusRegion(focusRange);
-  const base = normalizedFocus ?? normalizedCore;
-  if (!base) return null;
-  const start = base[0];
-  const span = Math.max(180, Math.round((base[1] - base[0]) * 0.22));
-  const end = Math.min(base[1], start + span);
-  return normalizeFocusRegion([start, end]);
-}
-
-function createMissingSlot() {
-  return {
-    value: null,
-    source_type: 'heuristic',
-    confidence: 0,
-    focus_region: null,
-    status: 'missing',
-  };
-}
-
-function createSlot({
-  value,
-  sourceType = 'heuristic',
-  confidence = 0.45,
-  focusRegion = null,
-  forceMissing = false,
-} = {}) {
-  const source_type = toSourceType(sourceType);
-  let normalizedValue = value;
-  if (Array.isArray(value)) {
-    normalizedValue = uniqueStrings(value);
-  } else {
-    normalizedValue = cleanText(value) || null;
-  }
-
-  const valuePresent = !forceMissing && hasValue(normalizedValue);
-  return {
-    value: valuePresent ? normalizedValue : null,
-    source_type,
-    confidence: Number(clampNumber(confidence, 0, 1, 0).toFixed(2)),
-    focus_region: normalizeFocusRegion(focusRegion),
-    status: valuePresent ? 'present' : 'missing',
-  };
-}
-
-function toNonGlobalRegex(regex) {
-  const flags = String(regex.flags ?? '').replace(/g/g, '');
-  return new RegExp(regex.source, flags);
-}
-
-function focusRegionFromBin(bin = {}) {
-  const yStart = Number(bin?.y_start);
-  if (Number.isFinite(yStart)) {
-    return normalizeFocusRegion([yStart, yStart + 400]);
-  }
-
-  const yCenter = Number(bin?.y_center);
-  if (!Number.isFinite(yCenter)) return null;
-  return normalizeFocusRegion([yCenter - 180, yCenter + 180]);
-}
-
-function buildSources({ meta, bins, hotspots, coreRange, focusRange }) {
-  const topRegion = resolveTopFocusRegion({ coreRange, focusRange });
-  const sources = [];
-  const title = cleanText(meta?.og_title || meta?.title || '');
-  const description = cleanText(meta?.og_description || meta?.description || '');
-
-  if (title) {
-    sources.push({
-      source_type: 'title',
-      text: title,
-      focus_region: topRegion,
-    });
-  }
-  if (description) {
-    sources.push({
-      source_type: 'meta',
-      text: description,
-      focus_region: topRegion,
-    });
-  }
-
-  for (const bin of (Array.isArray(bins) ? bins : [])) {
-    const text = cleanText(bin?.text);
-    if (!text) continue;
-    sources.push({
-      source_type: 'body_text',
-      text,
-      focus_region: focusRegionFromBin(bin),
-    });
-  }
-
-  for (const hotspot of (Array.isArray(hotspots) ? hotspots : [])) {
-    const text = cleanText(hotspot?.reason);
-    if (!text) continue;
-    const y = Number(hotspot?.y);
-    const focusRegion = Number.isFinite(y) ? normalizeFocusRegion([y - 160, y + 160]) : null;
-    sources.push({
-      source_type: 'body_text',
-      text,
-      focus_region: focusRegion,
-    });
-  }
-
-  return sources;
-}
-
-function findRegexCandidate({ sources, patterns }) {
-  for (const rawPattern of patterns) {
-    const pattern = rawPattern instanceof RegExp ? { regex: rawPattern } : rawPattern;
-    if (!(pattern.regex instanceof RegExp)) continue;
-    const regex = toNonGlobalRegex(pattern.regex);
-    const allowedTypes = Array.isArray(pattern.sourceTypes)
-      ? new Set(pattern.sourceTypes.map(toSourceType))
-      : null;
-
-    for (const source of sources) {
-      if (allowedTypes && !allowedTypes.has(toSourceType(source.source_type))) continue;
-      const text = cleanText(source?.text);
-      if (!text) continue;
-
-      const match = regex.exec(text);
-      if (!match) continue;
-
-      const raw = cleanText(match[1] ?? match[0]);
-      const parsedValue = typeof pattern.parser === 'function'
-        ? pattern.parser(raw, match, source)
-        : raw;
-      if (!hasValue(parsedValue)) continue;
-
-      const confidenceBoost = Number(pattern.confidenceBoost ?? 0);
-      const confidence = clampNumber(
-        baseConfidenceForSourceType(source.source_type) + confidenceBoost,
-        0,
-        0.99,
-        0.45
-      );
-
-      return {
-        value: parsedValue,
-        source_type: source.source_type,
-        confidence,
-        focus_region: source.focus_region ?? null,
-      };
-    }
-  }
-
-  return null;
-}
-
-function parseListTokens(raw, { maxItems = 6, allowChineseComma = true } = {}) {
-  const text = cleanText(raw).replace(/^[:：\-\s]+/, '');
-  if (!text) return [];
-
-  const delimiter = allowChineseComma ? /[、,，/|；;]+/ : /[,/|;]+/;
-  const items = text
-    .split(delimiter)
-    .map(item => cleanText(item).replace(/^[·•\-\s]+/, '').replace(/[。；;，,\s]+$/, ''))
-    .filter(Boolean);
-
-  return uniqueStrings(items).slice(0, maxItems);
-}
-
-function normalizeDateToken(raw) {
-  const normalized = cleanText(raw)
-    .replace(/年|\/|\./g, '-')
-    .replace(/月/g, '-')
-    .replace(/日/g, '')
-    .replace(/\s+/g, '');
-
-  const fullDate = normalized.match(/(20\d{2})-(\d{1,2})-(\d{1,2})/);
-  if (fullDate) {
-    const year = Number(fullDate[1]);
-    const month = Number(fullDate[2]);
-    const day = Number(fullDate[3]);
-    if (month >= 1 && month <= 12 && day >= 1 && day <= 31) {
-      return `${year}-${String(month).padStart(2, '0')}-${String(day).padStart(2, '0')}`;
-    }
-  }
-
-  const monthDate = normalized.match(/(20\d{2})-(\d{1,2})$/);
-  if (monthDate) {
-    const year = Number(monthDate[1]);
-    const month = Number(monthDate[2]);
-    if (month >= 1 && month <= 12) {
-      return `${year}-${String(month).padStart(2, '0')}`;
-    }
-  }
-
-  return cleanText(raw).slice(0, 24);
-}
-
-function normalizeCompanyCandidate(raw) {
-  let text = cleanText(raw)
-    .replace(/^[\s:：\-—_【】\[\]()（）|｜]+/, '')
-    .replace(/[\s:：\-—_【】\[\]()（）|｜]+$/, '');
-
-  text = text
-    .replace(/(?:20\d{2}年(?:春招|秋招|暑期|寒假|实习)?|20\d{2}届)\s*$/i, '')
-    .replace(/(?:春招|秋招|校园招聘|校招|暑期实习|实习招聘|实习生招聘|实习|招聘信息|招聘公告|招聘简章|招聘|职位招聘|岗位招聘|职位|岗位)\s*$/i, '')
-    .replace(/(?:20\d{2}年?)\s*$/i, '')
-    .trim();
-
-  if (!text) return '';
-  if (text.length < 2 || text.length > 40) return '';
-  if (/^\d+$/.test(text)) return '';
-  if (/^(?:年|月|日)+$/.test(text)) return '';
-  if (/^(?:年)?(?:春招|秋招|校招|暑期|寒假|实习|招聘|岗位|职位|公告|信息|简章|海报|计划|入口|流程|网申|投递|campus|intern|career|jobs?)$/i.test(text)) {
-    return '';
-  }
-  return text;
-}
-
-function buildCombinedText(sources = []) {
-  return sources.map(item => cleanText(item?.text)).filter(Boolean).join(' ');
-}
-
-function slotHasValue(slot) {
-  if (!slot || typeof slot !== 'object') return false;
-  return hasValue(slot.value);
-}
-
-function slotConfidence(slot) {
-  const confidence = Number(slot?.confidence);
-  if (!Number.isFinite(confidence)) return 0;
-  return Math.max(0, Math.min(1, confidence));
-}
-
-function slotHasMinConfidence(slot, min = 0) {
-  return slotHasValue(slot) && slotConfidence(slot) >= min;
-}
-
-function averageConfidence(slots = []) {
-  const valid = slots
-    .map(item => Number(item?.confidence))
-    .filter(value => Number.isFinite(value) && value > 0);
-  if (valid.length === 0) return 0;
-  const sum = valid.reduce((acc, item) => acc + item, 0);
-  return Number((sum / valid.length).toFixed(2));
-}
-
-function extractCompanySlot(sources) {
-  const candidate = findRegexCandidate({
-    sources,
-    patterns: [
-      {
-        regex: /([\u4e00-\u9fa5A-Za-z0-9·()（）\-]{2,40})(?:20\d{2}届)?(?:校园招聘|校招|春招|秋招|招聘简章|招聘公告|招聘信息|招聘|实习招聘|实习生招聘|社招|社会招聘|职位招聘|岗位招聘)/i,
-        parser: value => normalizeCompanyCandidate(value),
-        sourceTypes: ['title', 'meta'],
-        confidenceBoost: 0.06,
-      },
-      {
-        regex: /(?:公司名称|招聘单位|用人单位)[:：]?\s*([\u4e00-\u9fa5A-Za-z0-9·()（）\-]{2,40})/i,
-        parser: value => normalizeCompanyCandidate(value),
-        sourceTypes: ['body_text'],
-        confidenceBoost: 0.04,
-      },
-      {
-        regex: /([\u4e00-\u9fa5A-Za-z0-9·()（）\-]{2,40}(?:公司|集团|科技|信息|网络|银行|证券|研究院|大学|股份有限公司|有限公司|Inc\.?|Ltd\.?))/i,
-        parser: value => normalizeCompanyCandidate(value),
-        confidenceBoost: 0.02,
-      },
-    ],
-  });
-
-  if (!candidate) return createMissingSlot();
-  return createSlot(candidate);
-}
-
-function extractPublishedAtSlot(sources) {
-  const candidate = findRegexCandidate({
-    sources,
-    patterns: [
-      {
-        regex: /(?:发布时间|发布日期|发布于|更新于)[:：]?\s*([12]\d{3}[./\-年]\d{1,2}(?:[./\-月]\d{1,2}日?)?)/i,
-        parser: normalizeDateToken,
-        confidenceBoost: 0.06,
-      },
-      {
-        regex: /([12]\d{3}[./\-年]\d{1,2}[./\-月]\d{1,2}日?)/,
-        parser: normalizeDateToken,
-        confidenceBoost: 0.02,
-      },
-    ],
-  });
-
-  if (!candidate) return createMissingSlot();
-  return createSlot(candidate);
-}
-
-function extractRecruitmentTypeSlot(sources, { recruitmentRelated = false } = {}) {
-  const candidate = findRegexCandidate({
-    sources,
-    patterns: [
-      {
-        regex: CAMPUS_KEYWORDS,
-        parser: () => 'campus',
-        confidenceBoost: 0.06,
-      },
-      {
-        regex: INTERN_KEYWORDS,
-        parser: () => 'intern',
-        confidenceBoost: 0.06,
-      },
-      {
-        regex: EXPERIENCED_KEYWORDS,
-        parser: () => 'experienced',
-        confidenceBoost: 0.06,
-      },
-      {
-        regex: RECRUITMENT_KEYWORDS,
-        parser: () => 'job_posting_unknown',
-        confidenceBoost: -0.08,
-      },
-    ],
-  });
-
-  if (candidate) return createSlot(candidate);
-  if (!recruitmentRelated) return createMissingSlot();
-
-  return createSlot({
-    value: 'job_posting_unknown',
-    sourceType: 'heuristic',
-    confidence: 0.35,
-  });
-}
-
-function extractCohortSlot(sources) {
-  const candidate = findRegexCandidate({
-    sources,
-    patterns: [
-      {
-        regex: /((?:20\d{2}|\d{2})届)/,
-        confidenceBoost: 0.05,
-      },
-      {
-        regex: /(20\d{2}\s*(?:春招|秋招|校招|实习))/,
-        confidenceBoost: 0.05,
-      },
-    ],
-  });
-
-  if (!candidate) return createMissingSlot();
-  return createSlot(candidate);
-}
-
-function extractJobDirectionsSlot(sources, combinedText) {
-  const candidate = findRegexCandidate({
-    sources,
-    patterns: [
-      {
-        regex: /(?:岗位方向|招聘岗位|职位方向|岗位类别|职位类别|招聘方向)[:：]?\s*([^\n。；;]{4,90})/i,
-        parser: value => parseListTokens(value, { maxItems: 8 }),
-        confidenceBoost: 0.05,
-      },
-      {
-        regex: /(?:岗位|职位)[:：]?\s*([^\n。；;]{4,90})/i,
-        parser: value => parseListTokens(value, { maxItems: 8 }),
-        confidenceBoost: 0.01,
-      },
-    ],
-  });
-
-  if (candidate) return createSlot(candidate);
-
-  const lower = cleanText(combinedText).toLowerCase();
-  const inferred = JOB_DIRECTION_HINTS.filter(item => lower.includes(item.toLowerCase())).slice(0, 6);
-  if (inferred.length === 0) return createMissingSlot();
-
-  return createSlot({
-    value: inferred,
-    sourceType: 'heuristic',
-    confidence: 0.52,
-  });
-}
-
-function extractLocationsSlot(sources, combinedText) {
-  const candidate = findRegexCandidate({
-    sources,
-    patterns: [
-      {
-        regex: /(?:工作地点|工作城市|地点|城市|base地|工作地)[:：]?\s*([^\n。；;]{2,70})/i,
-        parser: value => parseListTokens(value, { maxItems: 8 }),
-        confidenceBoost: 0.04,
-      },
-    ],
-  });
-
-  if (candidate) return createSlot(candidate);
-
-  const lower = cleanText(combinedText).toLowerCase();
-  const inferred = LOCATION_HINTS
-    .filter(item => lower.includes(item.toLowerCase()))
-    .slice(0, 8);
-  if (inferred.length === 0) return createMissingSlot();
-
-  return createSlot({
-    value: inferred,
-    sourceType: 'heuristic',
-    confidence: 0.5,
-  });
-}
-
-function extractTargetOrRequirementsSlot(sources) {
-  const candidate = findRegexCandidate({
-    sources,
-    patterns: [
-      {
-        regex: /(?:面向对象|招聘对象|任职要求|岗位要求|任职资格|申请条件|target|requirements?)[:：]?\s*([^\n。]{8,140})/i,
-        parser: value => cleanText(value).slice(0, 140),
-        confidenceBoost: 0.03,
-      },
-    ],
-  });
-
-  if (!candidate) return createMissingSlot();
-  return createSlot(candidate);
-}
-
-function extractProcessSlot(sources, combinedText) {
-  const candidate = findRegexCandidate({
-    sources,
-    patterns: [
-      {
-        regex: /(?:招聘流程|流程安排|投递流程|面试流程|笔试流程|流程)[:：]?\s*([^\n。]{8,140})/i,
-        parser: value => cleanText(value).slice(0, 140),
-        confidenceBoost: 0.03,
-      },
-    ],
-  });
-  if (candidate) return createSlot(candidate);
-
-  const lower = cleanText(combinedText).toLowerCase();
-  const signals = ['网申', '测评', '笔试', '面试', 'offer'].filter(item => lower.includes(item.toLowerCase()));
-  if (signals.length === 0) return createMissingSlot();
-
-  return createSlot({
-    value: `${signals.join(' -> ')}${signals.length > 0 ? '（流程线索）' : ''}`,
-    sourceType: 'heuristic',
-    confidence: 0.48,
-  });
-}
-
-function extractEntryOrCtaSlot(sources, combinedText) {
-  const candidate = findRegexCandidate({
-    sources,
-    patterns: [
-      {
-        regex: /(?:投递入口|网申入口|申请入口|报名入口|简历投递|投递方式|投递链接|apply|立即申请|点击申请|马上投递)[:：]?\s*([^\n。]{4,120})/i,
-        parser: value => cleanText(value).slice(0, 120),
-        confidenceBoost: 0.04,
-      },
-      {
-        regex: /((?:立即申请|点击申请|马上投递|网申入口|投递入口|简历投递))/i,
-        parser: value => cleanText(value),
-        confidenceBoost: 0.02,
-      },
-    ],
-  });
-  if (candidate) return createSlot(candidate);
-
-  if (!/(投递|网申|申请|报名|简历)/i.test(combinedText)) return createMissingSlot();
-  return createSlot({
-    value: '页面存在投递/申请线索，建议定位入口后确认截止时间',
-    sourceType: 'heuristic',
-    confidence: 0.46,
-  });
-}
-
-function computeModeHint({ slots, recruitmentRelated }) {
-  const companyStrong = slotHasMinConfidence(slots.company, MODE_MIN_CONFIDENCE.strong);
-  const publishedStrong = slotHasMinConfidence(slots.published_at, MODE_MIN_CONFIDENCE.strong);
-  const cohortSupport = slotHasMinConfidence(slots.cohort, MODE_MIN_CONFIDENCE.support);
-  const jobDirectionsSupport = slotHasMinConfidence(slots.job_directions, MODE_MIN_CONFIDENCE.support);
-  const locationsSupport = slotHasMinConfidence(slots.locations, MODE_MIN_CONFIDENCE.support);
-  const processSignal = slotHasMinConfidence(slots.process, MODE_MIN_CONFIDENCE.weakSignal);
-  const recruitmentTypeValue = cleanText(slots.recruitment_type?.value);
-  const recruitmentTypeKnown = recruitmentTypeValue === 'campus'
-    || recruitmentTypeValue === 'intern'
-    || recruitmentTypeValue === 'experienced';
-  const recruitmentTypePresent = recruitmentTypeKnown || recruitmentTypeValue === 'job_posting_unknown';
-  const recruitmentTypeStrong = recruitmentTypeKnown
-    && slotHasMinConfidence(slots.recruitment_type, MODE_MIN_CONFIDENCE.strong);
-  const recruitmentTypeSignal = recruitmentTypePresent
-    && slotHasMinConfidence(slots.recruitment_type, MODE_MIN_CONFIDENCE.weakSignal);
-
-  const keySlots = [slots.company, slots.published_at, slots.recruitment_type, slots.job_directions, slots.locations, slots.process];
-  const avgConfidence = averageConfidence(keySlots);
-  const reliableSignalCount = [
-    companyStrong,
-    publishedStrong,
-    recruitmentTypeSignal,
-    jobDirectionsSupport,
-    locationsSupport,
-    processSignal,
-  ].filter(Boolean).length;
-
-  const mainMode = companyStrong && (
-    (publishedStrong && recruitmentTypeStrong)
-    || (cohortSupport && jobDirectionsSupport)
-    || (recruitmentTypeStrong && locationsSupport)
-  );
-  if (mainMode) {
-    return {
-      value: 'job_intel_broadcast',
-      confidence: Number(Math.max(0.7, avgConfidence).toFixed(2)),
-    };
-  }
-
-  const reminderMode = companyStrong && publishedStrong && recruitmentTypeSignal;
-  if (reminderMode) {
-    return {
-      value: 'job_alert',
-      confidence: Number(Math.max(0.6, avgConfidence).toFixed(2)),
-    };
-  }
-
-  if (recruitmentRelated
-    && reliableSignalCount > 0
-    && avgConfidence >= MODE_MIN_CONFIDENCE.floor) {
-    return {
-      value: 'info_summary',
-      confidence: Number(Math.max(0.42, avgConfidence).toFixed(2)),
-    };
-  }
-
-  return {
-    value: 'refuse_auto_broadcast',
-    confidence: Number(Math.max(0.2, avgConfidence * 0.6).toFixed(2)),
-  };
-}
-
-function isRecruitmentPage({ pageType, url, combinedText }) {
-  if (String(pageType) === 'job_detail') return true;
-  if (RECRUITMENT_KEYWORDS.test(cleanText(url))) return true;
-  return RECRUITMENT_KEYWORDS.test(combinedText);
-}
-
-export function derivePageType({ hostname, meta }) {
-  const host = String(hostname ?? '').toLowerCase();
-  const title = cleanText(meta?.og_title || meta?.title || '').toLowerCase();
-
-  if (host.includes('mp.weixin.qq.com') || host.includes('zhuanlan.zhihu.com')) return 'news_article';
-  if (host.includes('jianshu.com') || host.includes('juejin.cn')) return 'article';
-  if (/招聘|校招|实习|职位|job|career/.test(title)) return 'job_detail';
-  if (/产品|feature|pricing|方案|落地页|landing/.test(title)) return 'product_landing';
-  return 'generic_article';
-}
-
-export function buildHeuristicCoreMessage({ persona, meta, bins }) {
-  const p = cleanText(persona);
-  const title = cleanText(meta?.og_title || meta?.title || '');
-  const desc = cleanText(meta?.og_description || '');
-  const firstBin = cleanText(bins?.[0]?.text || '');
-
-  if (title && desc) return `${title}：${desc.slice(0, 80)}`;
-  if (title && firstBin) return `${title}。${firstBin.slice(0, 80)}`;
-  if (title) return title;
-  if (firstBin) return firstBin.slice(0, 100);
-  if (p) return `页面围绕「${p}」的关键信息组织，建议聚焦核心段落。`;
-  return '页面包含可用于短视频介绍的核心信息。';
-}
-
-function pickSpaced(candidates, { maxCount = 3, minGap = 750 } = {}) {
-  const sorted = [...candidates].sort((a, b) => b.score - a.score);
-  const picked = [];
-
-  for (const row of sorted) {
-    const farEnough = picked.every((prev) => Math.abs(prev.y - row.y) >= minGap);
-    if (!farEnough) continue;
-    picked.push(row);
-    if (picked.length >= maxCount) break;
-  }
-
-  return picked.sort((a, b) => a.y - b.y);
-}
-
-export function buildHeuristicHighlights({
-  bins,
-  hotspots,
-  coreRange,
-  focusRange,
-  maxCount = 3,
-}) {
-  const [coreStart, coreEnd] = Array.isArray(coreRange) ? coreRange : [0, 0];
-  const [focusStart, focusEnd] = Array.isArray(focusRange) ? focusRange : [coreStart, coreEnd];
-  const candidates = [];
-
-  for (const hotspot of (Array.isArray(hotspots) ? hotspots : [])) {
-    const y = Number(hotspot?.y);
-    if (!Number.isFinite(y) || y < focusStart || y > focusEnd) continue;
-    candidates.push({
-      y: Math.round(y),
-      reason: cleanText(hotspot?.reason) || '结构锚点',
-      from_chunk: 'dom_hotspot',
-      score: Number(hotspot?.weight ?? 5) * 10,
-    });
-  }
-
-  for (const [idx, bin] of (Array.isArray(bins) ? bins : []).entries()) {
-    const y = Number(bin?.y_center);
-    if (!Number.isFinite(y) || y < focusStart || y > focusEnd) continue;
-    const text = cleanText(bin?.text);
-    if (text.length < 16) continue;
-    const score = Math.min(100, Math.max(20, text.length));
-    candidates.push({
-      y: Math.round(y),
-      reason: `正文信息密集段（bin_${idx}）`,
-      from_chunk: `bin_${idx}`,
-      score,
-    });
-  }
-
-  if (candidates.length === 0) {
-    return [{
-      y: midpoint([focusStart, focusEnd]),
-      reason: '核心区中点回退',
-      from_chunk: 'fallback_center',
-    }];
-  }
-
-  return pickSpaced(candidates, { maxCount, minGap: 750 }).map((row) => ({
-    y: row.y,
-    reason: row.reason,
-    from_chunk: row.from_chunk,
-  }));
-}
-
-export function computeFocusRange(coreRange) {
-  const [coreStart, coreEnd] = Array.isArray(coreRange) ? coreRange : [0, 0];
-  const span = Math.max(0, coreEnd - coreStart);
-  if (span <= 3000) return [coreStart, coreEnd];
-  return [coreStart, coreStart + Math.round(span * 0.7)];
-}
-
-export function buildRecruitmentSemanticSlots({
-  pageType = 'generic_article',
-  url = '',
-  meta = {},
-  bins = [],
-  hotspots = [],
-  coreRange = [0, 0],
-  focusRange = [0, 0],
-} = {}) {
-  const sources = buildSources({
-    meta,
-    bins,
-    hotspots,
-    coreRange,
-    focusRange,
-  });
-  const combinedText = buildCombinedText(sources);
-  const recruitmentRelated = isRecruitmentPage({
-    pageType,
-    url,
-    combinedText,
-  });
-
-  const semanticSlots = {
-    company: extractCompanySlot(sources),
-    published_at: extractPublishedAtSlot(sources),
-    recruitment_type: extractRecruitmentTypeSlot(sources, { recruitmentRelated }),
-    cohort: extractCohortSlot(sources),
-    job_directions: extractJobDirectionsSlot(sources, combinedText),
-    locations: extractLocationsSlot(sources, combinedText),
-    target_or_requirements: extractTargetOrRequirementsSlot(sources),
-    process: extractProcessSlot(sources, combinedText),
-    entry_or_cta: extractEntryOrCtaSlot(sources, combinedText),
-  };
-
-  for (const key of RECRUITMENT_SLOT_KEYS) {
-    if (!semanticSlots[key]) semanticSlots[key] = createMissingSlot();
-  }
-
-  const modeHint = computeModeHint({
-    slots: semanticSlots,
-    recruitmentRelated,
-  });
-
-  return {
-    semantic_slots: semanticSlots,
-    mode_hint: modeHint.value,
-    mode_hint_confidence: modeHint.confidence,
-  };
-}