Press n or j to go to the next uncovered block, b, p or k for the previous block.
| 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 | 20x 160x 20x 20x 20x 20x 20x 120x 20x 317x 317x 317x 317x 26x 317x 26x 317x 317x 317x 317x 2536x 317x 317x 317x 21x 317x 317x 75x 75x 317x 296x 296x 317x 317x 317x 317x 317x 317x 317x 2219x 317x 317x 317x 317x 317x 317x 312x 312x 317x 74x 74x 317x 76x 76x 317x 317x 317x 317x 317x 317x 317x 317x 317x 317x 317x 317x 2x 2x 317x 317x 2x 2x 317x 317x 317x 317x 317x 11x 11x 317x 317x 951x 17x 17x 17x 17x 317x 317x 317x 317x 317x 317x 317x 317x 317x 317x 317x 312x 312x 317x 313x 313x 317x 317x 317x 317x 22x 22x 295x 295x 295x 295x 295x 790x 222x 295x 295x 295x 644x 295x 317x 317x 317x 317x 317x 317x 317x 317x 317x 317x 1902x 217x 1685x 463x 1222x 466x 317x 1279x 317x 317x 1902x 1710x 1902x 317x 317x 210x 210x 210x 210x 1260x 1260x 210x 210x 210x 834x 252x 252x 252x 252x 252x 252x 252x 252x 252x 245x 7x 20x 143x 143x 143x 143x 129x 14x 14x 14x 1x 1x 13x 13x 143x 143x 143x 8x 14x 14x 14x 14x 14x 14x 20x 244x 244x 244x 244x 244x 20x 7x 7x 7x 4x 7x 7x 7x 7x 7x 7x 7x 7x 7x 281x 414074x 281x 281x 281x 281x 281x 281x 281x 281x 281x 281x 281x 281x 281x 281x 210x 281x 281x 281x 281x 281x 281x 281x 281x 281x 19x 19x 16x 19x 3x 19x 18x 19x 2x 281x 210x 210x 210x 210x 630x 281x 281x 244x 244x 244x 244x 244x 244x 244x 244x 244x 244x 244x 20x 20x 1x 1x 20x 78x 78x 78x 78x 252x 252x 108x 108x 78x 78x 262x 262x 64x 64x 78x 172x 172x 78x 78x 78x 78x 78x 51x 51x 51x 51x 27x 2x 2x 78x | /**
* @module generate-news-enhanced/helpers
* @description Article writing, quality validation, and date formatting helpers.
*
* @author Hack23 AB
* @license Apache-2.0
*/
import fs from 'fs';
import path from 'path';
import { translateSwedishContent } from '../translation-dictionary.js';
import type { Language } from '../types/language.js';
import type { DateRange, ArticleQualityScore, UrgencyLabel } from '../types/article.js';
import type { ClassificationLevel, RiskLevel, ConfidenceLabel, DemocraticHealthLabel } from '../analysis-reader.js';
import { readLatestAnalysis, deriveArticleClassificationMeta } from '../analysis-reader.js';
import {
NEWS_DIR,
METADATA_DIR,
dryRunArg,
stats,
QUALITY_THRESHOLD,
MULTIDIM_QUALITY_THRESHOLD,
toISODate,
} from './config.js';
import type { MultiDimensionalQualityAssessment, DimensionScore, QualityIssue } from '../types/article.js';
import { detectBannedPatterns } from '../data-transformers/content-generators/shared.js';
import { detectSwedishLeakage } from '../detect-swedish-leakage.js';
// ---------------------------------------------------------------------------
// Multi-dimensional article quality assessment
// ---------------------------------------------------------------------------
/** Swedish party abbreviations used to detect stakeholder coverage. */
const SWEDISH_PARTIES: readonly string[] = ['S', 'M', 'SD', 'V', 'MP', 'C', 'L', 'KD'];
/** Pre-compiled regexes for party name matching (word boundaries, case-sensitive). */
const PARTY_REGEXES: readonly RegExp[] = SWEDISH_PARTIES.map(p => new RegExp(`\\b${p}\\b`));
/** Minimum number of dok_id citations for a passing evidence quality score. */
const MIN_DOK_ID_CITATIONS = 2;
/** Number of Swedish leakage occurrences that trigger one deduction step. */
const LEAKAGE_OCCURRENCES_PER_DEDUCTION = 5;
/** Points deducted per leakage deduction step. */
const LEAKAGE_DEDUCTION_AMOUNT = 25;
/** Dimension weights for the weighted overall score (must sum to 1.0). */
const DIMENSION_WEIGHTS = {
factualAccuracy: 0.25,
stakeholderCoverage: 0.15,
analyticalDepth: 0.15,
editorialConsistency: 0.10,
evidenceQuality: 0.20,
languageQuality: 0.15,
} as const satisfies Record<keyof MultiDimensionalQualityAssessment['dimensions'], number>;
const DIMENSION_WEIGHT_SUM: number = Object.values(DIMENSION_WEIGHTS).reduce(
(sum, weight) => sum + weight,
0,
);
Iif (Math.abs(DIMENSION_WEIGHT_SUM - 1) > 1e-10) {
throw new Error(`DIMENSION_WEIGHTS must sum to 1.0, got ${DIMENSION_WEIGHT_SUM}`);
}
/**
* Assess the factual accuracy dimension.
* Detects banned boilerplate patterns — each one deducts 20 points.
*/
function scoreFactualAccuracy(html: string): DimensionScore {
const bannedLabels: string[] = detectBannedPatterns(html);
const deduction: number = bannedLabels.length * 20;
const score: number = Math.max(0, 100 - deduction);
const evidence: string[] = bannedLabels.length === 0
? ['No banned boilerplate patterns detected']
: bannedLabels.map(label => `Banned pattern found: ${label}`);
const improvements: string[] = bannedLabels.map(
label => `Replace banned pattern "${label}" with genuine, document-specific analysis`,
);
return { score, evidence, improvements };
}
/**
* Assess stakeholder coverage dimension.
* Checks for named political parties/actors — at least 2 distinct parties expected.
*/
function scoreStakeholderCoverage(html: string): DimensionScore {
// Strip script/style blocks before tag stripping to avoid JS/JSON-LD pollution
const stripped: string = html
.replace(/<script\b[^>]*>[\s\S]*?<\/script\s*>/gi, ' ')
.replace(/<style\b[^>]*>[\s\S]*?<\/style\s*>/gi, ' ')
.replace(/<[^>]+>/g, ' ');
const evidence: string[] = [];
const improvements: string[] = [];
// Count distinct Swedish parties mentioned using pre-compiled regexes
const mentionedParties: string[] = SWEDISH_PARTIES.filter((_, i) => PARTY_REGEXES[i].test(stripped));
evidence.push(`${mentionedParties.length} distinct parties mentioned: ${mentionedParties.join(', ') || 'none'}`);
// Check for Winners & Losers section
const hasWinnersLosers: boolean = /winners?\s*(and|&|\/)\s*losers?/i.test(stripped)
|| /vinnare\s*(och|&|\/)\s*förlorare/i.test(stripped);
if (hasWinnersLosers) {
evidence.push('Winners & Losers section detected');
}
let score = 100;
if (mentionedParties.length < 2) {
score -= 30;
improvements.push('Add coverage of at least 2 distinct political parties with named actors');
}
if (!hasWinnersLosers) {
score -= 15;
improvements.push('Add a Winners & Losers section identifying which parties gain or lose');
}
return { score: Math.max(0, score), evidence, improvements };
}
/**
* Assess analytical depth dimension.
* Checks for Mermaid diagrams, analytical framework mentions, and structured analysis.
*/
function scoreAnalyticalDepth(html: string): DimensionScore {
const evidence: string[] = [];
const improvements: string[] = [];
// Count Mermaid diagrams
const mermaidMatches: RegExpMatchArray | null = html.match(/class="mermaid"/g);
const mermaidCount: number = mermaidMatches ? mermaidMatches.length : 0;
evidence.push(`${mermaidCount} Mermaid diagram(s) found`);
// Count analytical framework references
const frameworkPatterns: RegExp[] = [
/SWOT/i, /cost-benefit/i, /risk\s+assessment/i, /impact\s+analysis/i,
/stakeholder\s+analysis/i, /comparative\s+analysis/i, /trend\s+analysis/i,
];
const frameworkCount: number = frameworkPatterns.filter(p => p.test(html)).length;
evidence.push(`${frameworkCount} analytical framework reference(s) found`);
// Check for h3 sub-sections (deeper analysis)
const h3Matches: RegExpMatchArray | null = html.match(/<h3[\s>]/gi);
const h3Count: number = h3Matches ? h3Matches.length : 0;
evidence.push(`${h3Count} sub-section(s) found`);
let score = 100;
if (mermaidCount === 0) {
score -= 20;
improvements.push('Add at least one Mermaid diagram visualizing key relationships or processes');
}
if (frameworkCount === 0) {
score -= 15;
improvements.push('Reference analytical frameworks (SWOT, cost-benefit, risk assessment) to deepen analysis');
}
if (h3Count < 2) {
score -= 10;
improvements.push('Add sub-sections (h3) for more granular analytical depth');
}
return { score: Math.max(0, score), evidence, improvements };
}
/**
* Assess editorial consistency dimension.
* Detects duplicate "Why It Matters" sections and other structural duplications.
*/
function scoreEditorialConsistency(html: string): DimensionScore {
const evidence: string[] = [];
const improvements: string[] = [];
// Count "Why It Matters" occurrences in heading tags only (h2/h3)
const whyItMattersMatches: RegExpMatchArray | null = html.match(/<h[23][^>]*>[^<]*Why\s+It\s+Matters[^<]*<\/h[23]>/gi);
const whyItMattersCount: number = whyItMattersMatches ? whyItMattersMatches.length : 0;
evidence.push(`${whyItMattersCount} "Why It Matters" heading(s)`);
// Count "What to Watch" occurrences in heading tags only (h2/h3)
const watchMatches: RegExpMatchArray | null = html.match(/<h[23][^>]*>[^<]*What\s+to\s+Watch[^<]*<\/h[23]>/gi);
const watchCount: number = watchMatches ? watchMatches.length : 0;
evidence.push(`${watchCount} "What to Watch" heading(s)`);
let score = 100;
const duplicateWhyItMatters: number = Math.max(0, whyItMattersCount - 1);
if (duplicateWhyItMatters > 0) {
score -= duplicateWhyItMatters * 15;
improvements.push(`Remove ${duplicateWhyItMatters} duplicate "Why It Matters" section(s) — only one per article`);
}
const duplicateWatch: number = Math.max(0, watchCount - 1);
if (duplicateWatch > 0) {
score -= duplicateWatch * 15;
improvements.push(`Remove ${duplicateWatch} duplicate "What to Watch" section(s) — only one per article`);
}
return { score: Math.max(0, score), evidence, improvements };
}
/**
* Assess evidence quality dimension.
* Checks for dok_id citations, confidence labels, and named sources.
*/
function scoreEvidenceQuality(html: string, docIds: readonly string[]): DimensionScore {
const evidence: string[] = [];
const improvements: string[] = [];
const normalizeDocId = (value: string): string => value.trim().toUpperCase();
// Deduplicate source document IDs
const sourceDocIds: Set<string> = new Set(
docIds
.map((docId: string) => normalizeDocId(docId))
.filter((docId: string) => docId.length > 0),
);
// Extract unique document IDs from HTML (deduplicated across patterns)
const htmlDocIds: Set<string> = new Set();
const addMatchesToSet = (pattern: RegExp, groupIndex: number = 0): void => {
for (const match of html.matchAll(pattern)) {
const candidate: string | undefined = match[groupIndex];
Iif (!candidate) continue;
const normalized: string = normalizeDocId(candidate);
Eif (normalized) htmlDocIds.add(normalized);
}
};
// Extract explicit document ID values from common HTML/JSON citation fields
addMatchesToSet(/data-dok-id\s*=\s*["']([^"']+)["']/gi, 1);
addMatchesToSet(/\bdok_id\b\s*[:=]\s*["']?([A-Za-z0-9]+)["']?/gi, 1);
// Riksdag document ID format: letter + 3 digits + letter(s) + alphanumeric suffix
// Examples: H901AU10, H901FiU1, GZ10259. Intentionally broad to catch variants.
addMatchesToSet(/\b([A-Z]\d{3}[A-Za-z]\w+)\b/gi, 1);
const totalDocIds: number = new Set<string>([...sourceDocIds, ...htmlDocIds]).size;
evidence.push(
`${totalDocIds} unique document ID reference(s) found (${sourceDocIds.size} source + ${htmlDocIds.size} inline, merged and deduplicated)`,
);
// Check for confidence labels
const confidencePattern = /\b(HIGH|MEDIUM|LOW)\s+confidence\b/gi;
const confidenceMatches: RegExpMatchArray | null = html.match(confidencePattern);
const confidenceCount: number = confidenceMatches ? confidenceMatches.length : 0;
evidence.push(`${confidenceCount} confidence label(s) found`);
let score = 100;
if (totalDocIds < MIN_DOK_ID_CITATIONS) {
score -= 20;
improvements.push(`Add at least ${MIN_DOK_ID_CITATIONS} document ID citations (dok_id) to support factual claims`);
}
if (confidenceCount === 0) {
score -= 15;
improvements.push('Add confidence labels (HIGH/MEDIUM/LOW confidence) to analytical claims');
}
return { score: Math.max(0, score), evidence, improvements };
}
/**
* Assess language quality dimension.
* Detects Swedish text leakage in non-Swedish articles using the dedicated detector.
*/
function scoreLanguageQuality(html: string, lang: string): DimensionScore {
const evidence: string[] = [];
const improvements: string[] = [];
if (lang === 'sv') {
evidence.push('Swedish article — language leakage check not applicable');
return { score: 100, evidence, improvements };
}
const leakageReport = detectSwedishLeakage(html, lang as Language);
const leakageScore: number = leakageReport.score;
const leakedTerms = leakageReport.leakedTerms;
evidence.push(`${leakedTerms.length} unique Swedish term(s) leaked (${leakageScore} total occurrence(s))`);
if (leakedTerms.length > 0) {
const topTerms: string = leakedTerms.slice(0, 5).map(t => `"${t.term}"`).join(', ');
evidence.push(`Top leaked terms: ${topTerms}`);
}
const deduction: number = Math.min(100, Math.floor(leakageScore / LEAKAGE_OCCURRENCES_PER_DEDUCTION) * LEAKAGE_DEDUCTION_AMOUNT);
const score: number = Math.max(0, 100 - deduction);
if (leakageScore > 0) {
improvements.push(`Translate ${leakedTerms.length} Swedish term(s) to ${lang}: ${leakedTerms.slice(0, 3).map(t => `"${t.term}"`).join(', ')}`);
}
return { score, evidence, improvements };
}
/**
* Multi-dimensional article quality assessment.
*
* Scores 6 dimensions independently, then computes a weighted overall score.
* Returns actionable improvement suggestions and detected quality issues.
*
* @param html - raw HTML of the article
* @param lang - language code (e.g. "en", "sv")
* @param docIds - source document IDs used in the article
* @param threshold - minimum overall score to pass (typically MULTIDIM_QUALITY_THRESHOLD)
* @returns MultiDimensionalQualityAssessment with per-dimension scores, issues, and suggestions
*/
export function assessArticleQuality(html: string, lang: string, docIds: readonly string[], threshold: number): MultiDimensionalQualityAssessment {
// ── Pass 1: score each dimension ────────────────────────────────────────
const factualAccuracy = scoreFactualAccuracy(html);
const stakeholderCoverage = scoreStakeholderCoverage(html);
const analyticalDepth = scoreAnalyticalDepth(html);
const editorialConsistency = scoreEditorialConsistency(html);
const evidenceQuality = scoreEvidenceQuality(html, docIds);
const languageQuality = scoreLanguageQuality(html, lang);
const dimensions = {
factualAccuracy,
stakeholderCoverage,
analyticalDepth,
editorialConsistency,
evidenceQuality,
languageQuality,
};
// ── Pass 2: aggregate weighted overall score ────────────────────────────
const overallScore: number = Math.round(
factualAccuracy.score * DIMENSION_WEIGHTS.factualAccuracy +
stakeholderCoverage.score * DIMENSION_WEIGHTS.stakeholderCoverage +
analyticalDepth.score * DIMENSION_WEIGHTS.analyticalDepth +
editorialConsistency.score * DIMENSION_WEIGHTS.editorialConsistency +
evidenceQuality.score * DIMENSION_WEIGHTS.evidenceQuality +
languageQuality.score * DIMENSION_WEIGHTS.languageQuality,
);
// ── Collect issues sorted by severity ───────────────────────────────────
const issues: QualityIssue[] = [];
for (const [dimName, dim] of Object.entries(dimensions)) {
if (dim.score < 50) {
issues.push({
severity: 'critical',
dimension: dimName,
description: `${dimName} score critically low (${dim.score}/100)`,
suggestedFix: dim.improvements[0] ?? 'Improve this dimension',
});
} else if (dim.score < 70) {
issues.push({
severity: 'major',
dimension: dimName,
description: `${dimName} score below acceptable (${dim.score}/100)`,
suggestedFix: dim.improvements[0] ?? 'Improve this dimension',
});
} else if (dim.score < 90 && dim.improvements.length > 0) {
issues.push({
severity: 'minor',
dimension: dimName,
description: `${dimName} has room for improvement (${dim.score}/100)`,
suggestedFix: dim.improvements[0] ?? 'Improve this dimension',
});
}
}
// Sort: critical → major → minor
const severityOrder: Record<string, number> = { critical: 0, major: 1, minor: 2 };
issues.sort((a, b) => (severityOrder[a.severity] ?? 3) - (severityOrder[b.severity] ?? 3));
// ── Collect top suggestions ─────────────────────────────────────────────
const suggestions: string[] = [];
for (const dim of Object.values(dimensions)) {
for (const imp of dim.improvements) {
suggestions.push(imp);
}
}
// Hard gate: any dimension scoring critically low (<50) vetoes the pass.
// This prevents a severely problematic dimension (e.g. languageQuality=0 from Swedish
// leakage in an English article) from silently slipping through when other dimensions
// compensate in the weighted average. See https://github.com/Hack23/riksdagsmonitor/pull/1858
const criticalDimensions = Object.entries(dimensions).filter(([, dim]) => dim.score < 50);
const passesThreshold: boolean = overallScore >= threshold && criticalDimensions.length === 0;
return {
overallScore,
dimensions,
issues,
suggestions,
passesThreshold,
assessmentPasses: 2,
};
}
/** Print a summary of the multi-dimensional quality assessment to console. */
function printQualityReport(assessment: MultiDimensionalQualityAssessment, filename: string): void {
const reportId: string = filename.replace(/\.html$/, '');
console.log(`\n🔍 Multi-dimensional Quality Report: ${reportId}`);
console.log(` Overall score: ${assessment.overallScore}/100 — ${assessment.passesThreshold ? 'PASSED' : 'BELOW THRESHOLD'}`);
for (const [dimName, dim] of Object.entries(assessment.dimensions)) {
const icon: string = dim.score >= 80 ? '✅' : dim.score >= 50 ? '⚠️' : '❌';
console.log(` ${dimName}: ${dim.score}/100 ${icon}`);
}
Eif (assessment.issues.length > 0) {
console.log(` Issues (${assessment.issues.length}):`);
for (const issue of assessment.issues.slice(0, 5)) {
console.log(` [${issue.severity}] ${issue.description}`);
}
}
}
/**
* Inject quality metadata as `<meta>` tags into the article `<head>`.
* Inserts tags just before the closing `</head>` element.
*
* Quality metadata tags follow the quality-criteria.md v2 specification:
* - `article:quality-score` — overall multi-dimensional score (0-100)
* - `article:quality-version` — quality assessment version
* - `article:quality-iterations` — number of assessment passes
* - `article:quality-assessed` — whether a quality assessment is present
*/
export function injectQualityMetadata(html: string, assessment?: MultiDimensionalQualityAssessment): string {
const score = assessment?.overallScore ?? 0;
const passes = assessment?.assessmentPasses ?? 0;
const version = 'v2';
const qualityAssessed = assessment ? 'true' : 'false';
const metaTags = [
` <meta name="article:quality-score" content="${score}">`,
` <meta name="article:quality-version" content="${version}">`,
` <meta name="article:quality-iterations" content="${passes}">`,
` <meta name="article:quality-assessed" content="${qualityAssessed}">`,
// Legacy bare quality-* tags for backward compatibility with existing tooling
` <meta name="quality-score" content="${score}">`,
` <meta name="quality-version" content="${version}">`,
` <meta name="quality-iterations" content="${passes}">`,
` <meta name="quality-assessed" content="${qualityAssessed}">`,
// Legacy article-quality-* (hyphenated) tags for backward compatibility
` <meta name="article-quality-score" content="${score}">`,
` <meta name="article-quality-version" content="${version}">`,
` <meta name="article-quality-iterations" content="${passes}">`,
` <meta name="article-quality-assessed" content="${qualityAssessed}">`,
].join('\n');
// Remove any existing quality meta tags first (idempotent)
// Matches: article:quality-*, article-quality-*, quality-score, quality-version, quality-iterations, quality-assessed
const qualityMetaTagPattern =
/\s*<meta\b(?=[^>]*\bname\s*=\s*["'](?:article[:-]quality-[^"']+|quality-(?:score|version|iterations|assessed))["'])[^>]*>\s*\n?/gi;
const closingHeadPattern = /<\/head>/i;
const sanitizedHtml = html.replace(qualityMetaTagPattern, '');
// Insert before </head> (case-insensitive), preserving the original closing tag casing
if (closingHeadPattern.test(sanitizedHtml)) {
return sanitizedHtml.replace(closingHeadPattern, (match) => `${metaTags}\n${match}`);
}
return sanitizedHtml;
}
// ---------------------------------------------------------------------------
// Analysis enrichment — cached daily analysis for article metadata
// ---------------------------------------------------------------------------
/** Pre-computed classification metadata derived from the daily analysis pipeline. */
export interface AnalysisEnrichment {
classificationLevel: ClassificationLevel;
riskLevel: RiskLevel;
confidenceLabel: ConfidenceLabel;
significance?: number;
urgency?: UrgencyLabel;
/** Key themes from pre-computed synthesis analysis (for article enrichment) */
synthesisKeyThemes?: string[];
/** Aggregate risk summary from pre-computed risk assessment */
riskSummary?: string;
/** Date of the analysis data (may differ from article date due to lookback) */
analysisDate?: string;
/** Full SWOT analysis from pre-computed analysis files */
swotAnalysis?: import('../analysis-reader.js').SwotAnalysisResult;
/** Stakeholder perspectives from pre-computed analysis */
stakeholderPerspectives?: import('../analysis-reader.js').StakeholderPerspectivesResult;
/** Synthesis narrative direction for lede generation */
narrativeDirection?: string;
/** Recommended article focus from pre-computed synthesis */
articleFocus?: string;
/** Forward indicators for "What to Watch Next" */
forwardIndicators?: string[];
/** Threat analysis indicators and democratic health */
threatIndicators?: string[];
/** Democratic health assessment */
democraticHealth?: DemocraticHealthLabel;
/** Top significance-ranked documents with reasons */
topDocuments?: Array<{ docId: string; score: number; reason: string }>;
}
/**
* Options controlling which analysis snapshot to load.
*
* - `maxDaysBack` controls how far back in time we search for an analysis file.
* Defaults to `3` days (preserves existing behavior).
* - `basePath` allows callers/tests to select an alternative analysis base
* directory. When omitted, the default path used by `readLatestAnalysis`
* applies.
*/
export interface AnalysisEnrichmentOptions {
maxDaysBack?: number;
basePath?: string;
}
/**
* Module-level cache so analysis is loaded at most once per process for a
* given option set (maxDaysBack/basePath).
*/
const analysisEnrichmentCache = new Map<string, AnalysisEnrichment | null>();
/**
* Attempt to load the latest pre-computed daily analysis and derive article
* classification metadata.
*
* The result is cached for the lifetime of the process so that all article
* generators that request the same (maxDaysBack, basePath) share the same
* snapshot.
*
* Returns `null` when no analysis files are available (backward-compatible —
* generators can omit classification fields).
*/
export async function getAnalysisEnrichment(
options: AnalysisEnrichmentOptions = {},
): Promise<AnalysisEnrichment | null> {
const maxDaysBack = options.maxDaysBack ?? 3;
const basePath = options.basePath;
const cacheKey = `${maxDaysBack}:${basePath ?? 'default'}`;
if (analysisEnrichmentCache.has(cacheKey)) {
return analysisEnrichmentCache.get(cacheKey) ?? null;
}
try {
const analysis = await readLatestAnalysis(maxDaysBack, basePath);
if (!analysis.hasAnalysis) {
analysisEnrichmentCache.set(cacheKey, null);
return null;
}
const meta = deriveArticleClassificationMeta(analysis);
const enrichment: AnalysisEnrichment = {
classificationLevel: meta.classificationLevel,
riskLevel: meta.riskLevel,
confidenceLabel: meta.confidenceLabel,
significance: meta.significanceScore,
urgency: meta.urgency,
// Feed pre-computed analysis content into article generation
synthesisKeyThemes: analysis.synthesis?.keyThemes ?? [],
riskSummary: analysis.riskAssessment?.summary ?? undefined,
analysisDate: analysis.date,
// Deep analysis content for article body enrichment
swotAnalysis: analysis.swot ?? undefined,
stakeholderPerspectives: analysis.stakeholderPerspectives ?? undefined,
narrativeDirection: analysis.synthesis?.narrativeDirection ?? undefined,
articleFocus: analysis.synthesis?.articleFocus ?? undefined,
forwardIndicators: analysis.synthesis?.forwardIndicators ?? [],
threatIndicators: analysis.threatAnalysis?.indicators ?? [],
democraticHealth: analysis.threatAnalysis?.democraticHealth ?? undefined,
topDocuments: analysis.significance?.topDocuments ?? [],
};
analysisEnrichmentCache.set(cacheKey, enrichment);
console.log(` 📊 Analysis enrichment loaded: classification=${meta.classificationLevel}, risk=${meta.riskLevel}, confidence=${meta.confidenceLabel}, keyThemes=${enrichment.synthesisKeyThemes?.length ?? 0}, swot=${enrichment.swotAnalysis ? 'YES' : 'NO'}, stakeholders=${enrichment.stakeholderPerspectives ? 'YES' : 'NO'}, forwardIndicators=${enrichment.forwardIndicators?.length ?? 0}, topDocs=${enrichment.topDocuments?.length ?? 0}`);
return enrichment;
} catch (error: unknown) {
if (process.env.DEBUG || process.env.LOG_LEVEL === 'debug') {
console.error(
'⚠️ Failed to load analysis enrichment (falling back to null):',
error,
);
}
analysisEnrichmentCache.set(cacheKey, null);
return null;
}
}
/**
* Reset the analysis enrichment cache. Useful in tests.
*/
export function resetAnalysisEnrichmentCache(): void {
analysisEnrichmentCache.clear();
}
/**
* Get date range for Week Ahead (next 7 days)
*/
export function getWeekAheadDateRange(): DateRange {
const today: Date = new Date();
const startDate: Date = new Date(today);
startDate.setDate(today.getDate() + 1); // Tomorrow
const endDate: Date = new Date(startDate);
endDate.setDate(startDate.getDate() + 7); // +7 days
return {
start: toISODate(startDate),
end: toISODate(endDate)
};
}
/**
* Format date for article slug
*/
export function formatDateForSlug(date: Date = new Date()): string {
return toISODate(date);
}
/**
* Write article to file
*/
export async function writeArticle(html: string, filename: string): Promise<boolean> {
Iif (dryRunArg) {
console.log(` [DRY RUN] Would write: ${filename}`);
return true;
}
const filepath: string = path.join(NEWS_DIR, filename);
fs.writeFileSync(filepath, html, 'utf-8');
console.log(` ✅ Wrote: ${filename}`);
return true;
}
// ---------------------------------------------------------------------------
// Per-article quality score persistence
// ---------------------------------------------------------------------------
/** In-memory store for per-article multi-dimensional scores (written per-run) */
const perArticleScores: Record<string, {
filename: string;
lang: string;
articleType: string;
score: number;
passed: boolean;
multidimensional: {
overallScore: number;
passesThreshold: boolean;
assessmentPasses: number;
dimensions: Record<string, number>;
issueCount: number;
};
timestamp: string;
}> = {};
/**
* Persist all collected per-article quality scores to
* `news/metadata/quality-scores.json`.
*
* Uses atomic write (write to temp file, then rename) to avoid leaving
* truncated/invalid JSON on disk if the process is interrupted mid-write.
*
* **Per-run overwrite**: Only the current run's scores are written. Previous
* runs' data is replaced so that stale/test entries never accumulate and
* Check 13's average score reflects the current generation only.
*
* Call once at the end of the overall generation run (not per-article) to
* avoid write amplification when many articles are generated.
*/
export function flushQualityScores(): void {
Iif (dryRunArg) return;
try {
if (!fs.existsSync(METADATA_DIR)) {
fs.mkdirSync(METADATA_DIR, { recursive: true });
}
const outPath = path.join(METADATA_DIR, 'quality-scores.json');
const tmpPath = outPath + '.tmp';
// Best-effort atomic write: write to temp file, then rename.
// Try rename first (atomic on POSIX). On Windows the rename may fail when
// the destination exists, so fall back to unlink-then-rename.
fs.writeFileSync(tmpPath, JSON.stringify(perArticleScores, null, 2), 'utf-8');
try {
fs.renameSync(tmpPath, outPath);
} catch (renameErr: unknown) {
// Only attempt unlink+rename for known Windows cross-device/exists codes.
// Other failures (permissions, missing parent dir, etc.) should propagate
// so we don't delete an otherwise valid quality-scores.json.
const code = (renameErr as NodeJS.ErrnoException).code;
Iif (code === 'EEXIST' || code === 'EPERM' || code === 'EACCES' || code === 'EXDEV') {
try { fs.unlinkSync(outPath); } catch (e: unknown) {
if ((e as NodeJS.ErrnoException).code !== 'ENOENT') throw e;
}
fs.renameSync(tmpPath, outPath);
} else {
throw renameErr;
}
}
// Clear in-memory map after a successful flush so that subsequent
// invocations in the same Node process don't carry stale entries.
for (const key of Object.keys(perArticleScores)) {
delete perArticleScores[key];
}
} catch (err: unknown) {
console.warn(` ⚠️ Could not persist quality scores: ${(err as Error).message}`);
}
}
// ---------------------------------------------------------------------------
// Article quality validation
// ---------------------------------------------------------------------------
/**
* Validate the quality of a generated article HTML.
*
* Performs two sequential assessments:
*
* 1. **Structural scoring** (0–100): word count (0–50), h2 sections (0–30),
* translation completeness (0–20). Used as the primary pass/fail gate
* against `QUALITY_THRESHOLD`.
*
* 2. **Multi-dimensional assessment** via `assessArticleQuality()`, which
* internally runs its own two passes (dimension computation + aggregation),
* producing a 6-dimension weighted score, issue list, and suggestions.
*
* Total: 3 assessment passes — 1 structural + 2 multi-dimensional.
*
* @param html - raw HTML of the article
* @param lang - language code of the article (e.g. "en")
* @param articleType - article type slug (e.g. "motions")
* @param filename - filename for the quality record
* @param sourceDocIds - optional list of source document IDs for factual-accuracy check
* @returns ArticleQualityScore with metrics, pass/fail, and multidimensional assessment
*/
export function validateArticleQuality(
html: string,
lang: string,
articleType: string,
filename: string,
sourceDocIds: readonly string[] = [],
): ArticleQualityScore {
// ── Pass 1: structural scoring ────────────────────────────────────────────
const stripped: string = html
.replace(/<script[\s>][\s\S]*?<\/script>/gi, ' ')
.replace(/<style[\s>][\s\S]*?<\/style>/gi, ' ')
.replace(/<[^>]+>/g, ' ');
const wordCount: number = stripped.split(/\s+/).filter(w => w.length > 0).length;
const wordScore: number = Math.min(50, Math.round((wordCount / 1000) * 50));
const h2Matches: RegExpMatchArray | null = html.match(/<h2[\s>]/gi);
const analyticalSections: number = h2Matches ? h2Matches.length : 0;
const sectionScore: number = Math.min(30, Math.round((analyticalSections / 3) * 30));
const untranslatedMatches: RegExpMatchArray | null = html.match(/data-translate="true"/g);
const untranslatedSpans: number = untranslatedMatches ? untranslatedMatches.length : 0;
const translationDeduction: number = lang === 'sv' ? 0 : Math.min(20, untranslatedSpans * 2);
const translationScore: number = 20 - translationDeduction;
const score: number = wordScore + sectionScore + translationScore;
const unknownMatches: RegExpMatchArray | null = html.match(/Unknown \(Unknown\)/g);
const unknownAuthors: number = unknownMatches ? unknownMatches.length : 0;
const passed: boolean = score >= QUALITY_THRESHOLD;
// ── Pass 2: multi-dimensional assessment ─────────────────────────────────
const multidimensional = assessArticleQuality(html, lang, sourceDocIds, MULTIDIM_QUALITY_THRESHOLD);
if (multidimensional.passesThreshold === false || process.env.NEWS_QUALITY_VERBOSE === '1') {
printQualityReport(multidimensional, filename);
}
// ----- console report (structural) -----
const scoreLabel: string = passed ? '✅' : '⚠️';
const reportId: string = filename.replace(/\.html$/, '');
console.log(`\n📊 Article Quality Report (structural): ${reportId}`);
console.log(` Word count: ${wordCount} (score: ${wordScore}/50)`);
console.log(` Analytical sections: ${analyticalSections} (score: ${sectionScore}/30)`);
console.log(` Untranslated spans: ${untranslatedSpans} (score: ${translationScore}/20)`);
console.log(` Unknown authors: ${unknownAuthors} ${unknownAuthors > 0 ? '⚠️' : '✅'}`);
console.log(` Quality Score: ${score}/100 — ${passed ? 'PASSED' : 'BELOW THRESHOLD'} ${scoreLabel}`);
if (!passed) {
console.warn(` ⚠️ Score ${score} is below threshold ${QUALITY_THRESHOLD}. Article written but flagged.`);
if (wordCount < 300) {
console.warn(' → Article under 300 words — expand with analytical sections');
}
if (untranslatedSpans > 10 && lang !== 'sv') {
console.warn(` → ${untranslatedSpans} untranslated data-translate spans — translate before committing`);
}
if (analyticalSections < 1) {
console.warn(' → No analytical h2 sections found — add thematic analysis');
}
if (unknownAuthors > 0) {
console.warn(` → ${unknownAuthors} "Unknown (Unknown)" entries — fix author/party metadata`);
}
}
if (!multidimensional.passesThreshold) {
console.warn(` ⚠️ Multi-dimensional score ${multidimensional.overallScore}/100 below threshold ${MULTIDIM_QUALITY_THRESHOLD}.`);
Eif (multidimensional.suggestions.length > 0) {
console.warn(' Top improvement suggestions:');
for (const s of multidimensional.suggestions.slice(0, 3)) {
console.warn(` → ${s}`);
}
}
}
// Accumulate per-article score (flushed at end of run via exported flushQualityScores())
perArticleScores[filename] = {
filename,
lang,
articleType,
score,
passed,
multidimensional: {
overallScore: multidimensional.overallScore,
passesThreshold: multidimensional.passesThreshold,
assessmentPasses: multidimensional.assessmentPasses,
dimensions: {
factualAccuracy: multidimensional.dimensions.factualAccuracy.score,
stakeholderCoverage: multidimensional.dimensions.stakeholderCoverage.score,
analyticalDepth: multidimensional.dimensions.analyticalDepth.score,
editorialConsistency: multidimensional.dimensions.editorialConsistency.score,
evidenceQuality: multidimensional.dimensions.evidenceQuality.score,
languageQuality: multidimensional.dimensions.languageQuality.score,
},
issueCount: multidimensional.issues.length,
},
timestamp: new Date().toISOString(),
};
return {
filename,
lang,
articleType,
wordCount,
unknownAuthors,
untranslatedSpans,
analyticalSections,
score,
passed,
multidimensional,
};
}
/**
* Write article in specified language
*/
export async function writeSingleArticle(html: string, slug: string, lang: Language, articleType?: string, sourceDocIds?: readonly string[]): Promise<string> {
const filename: string = `${slug}-${lang}.html`;
// Translate any remaining Swedish data-translate spans before writing or validating
const translatedHtml: string = translateSwedishContent(html, lang);
// Infer article type from slug (e.g. "2026-02-23-motions" → "motions",
// "2026-02-23-committee-reports" → "committee-reports"). Falls back to the
// full slug if the slug does not follow the YYYY-MM-DD-{type} pattern.
const slugParts: string[] = slug.split('-');
const inferredType: string = slugParts.length >= 4 ? slugParts.slice(3).join('-') : slug;
const qualityScore: ArticleQualityScore = validateArticleQuality(
translatedHtml,
lang,
articleType ?? inferredType,
filename,
sourceDocIds ?? [],
);
stats.qualityScores.push(qualityScore);
// Inject quality metadata (CSP-safe <meta> tag only; opt-in JSON-LD via injectJsonLd param)
const finalHtml: string = qualityScore.multidimensional
? injectQualityMetadata(translatedHtml, qualityScore.multidimensional)
: translatedHtml;
await writeArticle(finalHtml, filename);
stats.generated += 1;
stats.articles.push(filename);
return filename;
}
/**
* Install process-exit signal handlers that flush quality scores once.
* Call from the CLI entrypoint (not at module load time) to avoid side
* effects when helpers.ts is imported by tests or other tooling.
*/
let _flushGuardInstalled = false;
let _flushed = false;
/** Flush quality scores at most once per process lifetime. */
function flushOnce(): void {
if (_flushed) return;
if (Object.keys(perArticleScores).length === 0) {
_flushed = true;
return;
}
flushQualityScores();
if (Object.keys(perArticleScores).length === 0) {
_flushed = true;
}
}
export function installFlushHandlers(): void {
if (_flushGuardInstalled) return;
_flushGuardInstalled = true;
process.once('exit', () => flushOnce());
process.once('SIGINT', () => { flushOnce(); process.exit(130); });
process.once('SIGTERM', () => { flushOnce(); process.exit(143); });
}
/**
* Write EN/SV article pair — convenience wrapper over writeSingleArticle
*/
export async function writeArticlePair(htmlEN: string, htmlSV: string, slug: string): Promise<void> {
await writeSingleArticle(htmlEN, slug, 'en');
await writeSingleArticle(htmlSV, slug, 'sv');
}
// ---------------------------------------------------------------------------
// Dynamic title/description generation from content highlights
// ---------------------------------------------------------------------------
/** Extract top N most relevant highlight phrases from article content */
// v5.0: The HIGHLIGHT_REJECT_PATTERNS array and extractHighlights/extractDominantTheme
// functions were removed. Title generation is now performed by the AI agent (Copilot
// opus 4.6) during agentic workflows — see ai-driven-analysis-guide.md v5.0.
// The BANNED_PATTERNS for content quality remain in shared.ts detectBannedPatterns().
/**
* Template field labels and generic terms that MUST NOT appear in titles.
* Categories: (1) Template field labels: committee, published, what this means, why it matters, filed by
* (2) Generic UI text: read the full, thematic analysis, legislative pipeline, opposition strategy
* (3) Structural elements: report-entry, unknown, policy domain, department
*
* Hoisted to module scope to avoid recompiling the regex on every call to
* `generateDynamicTitle()`.
*/
const EXCLUDED_PATTERNS = /^(committee:?|published:?|what this means:?|why it matters:?|filed by:?|read the full|thematic analysis|legislative pipeline|opposition strategy|responses to|report-entry|unknown|policy domain|department)/i;
/**
* Generate a content-aware fallback title and subtitle from article content.
*
* The AI agent in the agentic workflow (.md prompt) SHOULD overwrite
* both the title and subtitle with genuine, analysis-driven values.
* This function provides a minimally newsworthy fallback so that articles
* never ship with bare category labels as titles.
*
* Extracts policy domains from `<strong>` tags and h3 headings in content
* to produce titles like "Committee Reports: Defence, Transport, and Climate"
* instead of the bare "Committee Reports".
*
* @param baseTitle - The generic article type title (e.g. "Government Propositions")
* @param content - HTML content of the article body
* @param docCount - Number of source documents (0 if unknown)
* @returns A `TitleSet` with content-aware fallback — AI agent should replace both.
*/
export function generateDynamicTitle(
baseTitle: string,
content: string,
docCount: number,
): { title: string; subtitle: string } {
// Extract topic hints from strong tags and h3 headings
const seen = new Set<string>();
const topics: string[] = [];
const strongMatches = content.matchAll(/<strong[^>]*>([^<]{3,50})<\/strong>/gi);
for (const m of strongMatches) {
const text = m[1]?.trim();
if (text && !seen.has(text) && topics.length < 5 && !EXCLUDED_PATTERNS.test(text)) {
seen.add(text);
topics.push(text);
}
}
const h3Matches = content.matchAll(/<h3[^>]*>([^<]{3,60})<\/h3>/gi);
for (const m of h3Matches) {
const text = m[1]?.trim();
if (text && !seen.has(text) && topics.length < 5 && !EXCLUDED_PATTERNS.test(text)) {
seen.add(text);
topics.push(text);
}
}
// Sanitize topic strings: strip newlines, collapse whitespace, drop quotes
const sanitized = topics.map(t =>
t.replace(/[\r\n]+/g, ' ').replace(/\s+/g, ' ').replace(/["']/g, '').trim()
).filter(t => t.length >= 3 && !EXCLUDED_PATTERNS.test(t));
// Build a content-aware title if we found topic hints
let title = baseTitle;
let subtitle = `${baseTitle} — AI-generated political intelligence from Sweden's Riksdag`;
const documentLabel = docCount === 1 ? 'document' : 'documents';
const documentTitleLabel = docCount === 1 ? 'Document' : 'Documents';
if (sanitized.length > 0) {
const topicList = sanitized.slice(0, 3).join(', ');
title = `${baseTitle}: ${topicList}`;
const countStr = docCount > 0 ? ` across ${docCount} ${documentLabel}` : '';
subtitle = `Analysis of ${topicList}${countStr} in Sweden's Riksdag`;
} else if (docCount > 0) {
title = `${baseTitle}: ${docCount} ${documentTitleLabel} Analyzed`;
subtitle = `Analysis of ${docCount} parliamentary ${documentLabel} from Sweden's Riksdag`;
}
return { title, subtitle };
}
|