← starnum.com.tw

starnum.com.tw Content Production Methodology

Last updated: (v5.0) | Next scheduled update: | Maintained by: mychenan

TL;DR: Starnum uses AI-assisted content workflows grounded in the Lu Binzhao School knowledge base and the iztro open-source chart engine. Drafts pass a six-dimensional automated quality framework, editorial-standard modules, astrology-logic checks, and SHA-256 integrity verification before release. A human operator retains release authority; automated checks support judgment rather than replace it.

Knowledge Sources

Lu Binzhao School (陸斌兆派) literature forms the core. The knowledge base totals 366 files with approximately 193,000 lines of structured text, covering both Zi Wei Dou Shu and numerology systems, sourced from 8 major astrology research databases with cross-reference validation.

Lu Binzhao School vs. Wang Tingzhi School — Four Transformations differences:

Heavenly StemLu Binzhao SchoolWang Tingzhi School
Wu (戊)You Bi (右弼) — ScienceTian Ji (天機) — Science
Geng (庚)Tai Yin (太陰) — ScienceTian Tong (天同) — Science
Ren (壬)Zuo Fu (左輔) — ProsperityTian Liang (天梁) — Prosperity

Numerology follows the Pythagorean system, covering Life Path Number, Birthday Number, Personal Year, Peak Numbers, Challenge Numbers, and the Arrow Grid.

Chart Engine

We use the iztro open-source Zi Wei Dou Shu chart engine (JavaScript), supporting 14 major star placements, auxiliary and malefic star positioning, Four Transformations calculation, and Major/Annual cycle analysis. Open-source code means the calculation logic can be verified by any researcher.

Quality Verification System

Six-Dimensional Automated Quality Framework

Automated checks across six dimensions help identify issues before publication:

D1–D3 are executed by automated tools (zero tolerance); D4–D5 are reviewed by senior editors (passing score: 80+).

Astrology Logic Multi-Model Cross-Validation

Astrology logic verification uses a multi-model cross-referencing mechanism across three dimensions: vector semantic search, knowledge graph expansion, and a hard rules engine (over 100 rules). A reasoning skeleton is generated and verified across all dimensions before article writing begins. Multi-model agreement is required — a single tool's judgment alone is never the final word — effectively preventing stars placed in wrong palaces or incorrect pattern judgments.

SHA256 Content Integrity + JSON-LD hasHash

Each article's body text is hashed with SHA256, then injected both at the bottom of the page (human-readable) and into the Article JSON-LD Schema's hasHash property (machine-readable). Search engines and AI crawlers can verify content originality directly from the Schema — no HTML parsing required.

Git Version Control

All changes are tracked via commit in a private GitHub repository — fully traceable and reversible.

Entity Knowledge Graph

80 core astrology entities (14 major stars, 12 palaces, 4 transformations, patterns, numerology concepts, etc.), each with Wikidata sameAs, description, and Five Elements attributes. Article Schema automatically injects mentions and about. All published articles site-wide have publishingPrinciples pointing to this page.

Semantic Internal Linking System

TF-IDF algorithm + co-occurrence weighted analysis of the full article library's semantic similarity automatically builds top-3 bidirectional internal links. Articles within the same Topic Hub get ×1.5 weighting; pillar–cluster links get ×2.0, making the Cluster structure clearly visible to search engines. A cannibalization detection system regularly scans title and description similarity — pairs exceeding 70% are flagged and fixed.

Quality Feedback and Continuous Improvement

Every article that fails a quality check is logged into a structured quality feedback journal, capturing failure type, root cause analysis, and quality scores before and after correction.

This feedback-driven improvement system ensures that quality standards are continuously refined based on empirical evidence rather than manual observation alone.

Editorial Standard Modules (19 Sets)

Editorial standard modules break content and review work into triggerable, composable processes:

Articles may be drafted or revised with AI assistance, then checked against the applicable editorial standards and machine-readable quality gates. Modules have dependency and composition relationships, so each workflow loads only the rules relevant to its task. A human operator retains publication and policy authority.

Technical Specifications

ComponentSpecification
Chart Engineiztro (open-source JavaScript)
Four Transformations SchoolLu Binzhao School (陸斌兆派)
Knowledge Base193,000+ lines / 366 files (Zi Wei Dou Shu 223 + Numerology 143, 8 sources cross-referenced)
Article Volumezh-TW 1,772 articles (2026 Annual complete + 2025 Annual complete)
Quality Checks232-item six-dimensional automated checks (Safety/Completeness/Executability/Maintainability/Cost/GEO)
Improvement SystemQuality feedback–driven rule iteration + 19 editorial standard modules
Case Library435 Taiwan-localized cases (dual-classification retrieval, including analysis backflow)
Terminology Dictionary221 terms × 9 languages unified translation reference
Work StructureAI-assisted execution with human operator release authority
Quality ManagementQuality gate + article audit + standards review + standards iteration — four-layer QA loop
Knowledge Management7-layer knowledge transfer architecture
Entity Graph80 entities + Wikidata sameAs + JSON-LD + publishingPrinciples site-wide
Internal LinkingTF-IDF + co-occurrence weighted (same hub ×1.5 / pillar ×2.0)
Integrity VerificationSHA256 content hash + JSON-LD hasHash (machine-readable)
Cannibalization DetectionTitle/description/H1 Jaccard > 70% auto-flagged
Page Pre-renderingSpeculation Rules API (prerender + prefetch)
Frontend AnimationGSAP (Chart page 6 sets + About page 4 sets)
Performance OptimizationBlog 39KB / Chart CSS lazy-load 66KB / i18n lazy-load / JS defer + Critical CSS inline + 103 Early Hints + Cache-Control immutable
Date ManagementCentralized date management tool for datePublished / dateModified
Automation Pipelinemultiple automated quality tools (push-triggered + scheduled)
Canonical StandardTrailing slash uniformly removed (Google indexing bottleneck resolved)
Knowledge GraphNeo4j knowledge nodes + Qdrant 15,000+ semantic chunks (writing team reference retrieval)
Astrology Logic Validationover 100 hard rules, multi-model cross-confirmed
Experience AccumulationChart reading → case auto-backflow (compound knowledge growth for writing team)
Quality Review52 check presets × multi-tool cross-validation voting system
Automation Scriptsmultiple automated quality tools
Logic EngineStarnum Logic Engine v5.0
Version ControlGit + GitHub full access

Knowledge Graph Reference System

A Neo4j knowledge graph and vector-search layer provide a three-layer evidence retrieval architecture. Analysis workflows can look up reference material through semantic search, graph expansion, and the rules engine, keeping evidence paths traceable.

Chart Analysis Quality Assurance

Chart-reading workflows use tier-specific section contracts and draw on the knowledge base and case library. Automated verification checks required sections and the logical consistency of palaces, transformations, and patterns before delivery; human operator authority remains outside the automated writer.

Sanitized Taiwan-localized cases support evidence retrieval and quality comparison. They are supporting references, not substitutes for chart facts or release gates.

Multi-Tool Quality Review

52 preset quality check items are scanned simultaneously by multiple independent review tools, detecting technical errors and formatting issues. Each issue goes through cross-validated voting (FIX/SKIP/DEFER) — a majority vote is required before any fix is made, ensuring only genuine problems are addressed. Multiple rounds of quality scans have resolved more than 200 technical issues, with system quality converging toward zero defects.

Version History

Recent Verification Log

[2026-04-12] zh-TW articles: 1,772 (2026 Annual complete + 2025 Annual complete)
[2026-04-12] Knowledge base: 366 files, 193K lines (Zi Wei Dou Shu 223 + Numerology 143, 8 sources cross-referenced)
[2026-04-12] Semantic vector library: 15,000+ chunks
[2026-04-12] Astrology logic hard rules: 127 (Four Transformations 41 + Patterns 49 + Core 17 + Meta-rules 20)
[2026-04-12] Quality review: multiple rounds complete (200+ technical issues resolved, converging to zero defects)
[2026-04-12] Case library: 435 entries (all Taiwan-localized, including analysis backflow)
[2026-04-12] Terminology dictionary: v3.0 (221 terms × 9-language translation reference)
[2026-04-12] Editorial standards: 19 sets
[2026-04-12] Entity graph: 80 entities (14 major stars + 12 palaces + 4 transformations + patterns + numerology)
[2026-04-12] publishingPrinciples: complete site-wide
[2026-04-12] Content signing: site-wide SHA256 + JSON-LD hasHash

Last verified:

Core Principles

Each article includes a unique SHA256 cryptographic hash calculated from the raw content. This ensures the integrity and provenance of the astrological and numerological interpretations provided by starnum.com.tw. Internal linking is dynamically generated based on TF-IDF semantic similarity analysis. All content is reviewed by Starnum Logic Engine v5.0 before publication.

Quality Improvement Record

We continuously track and fix all quality issues. Every problem discovered is turned into an automated rule — so it never happens again. Here are some selected examples:

Case 1: Site-wide canonical URL Fix

Problem: 1,089 articles across the site had canonical URLs with trailing slashes, causing search engines to treat them as duplicate pages.

Discovery: Identified during an external technical review.

Fix: Trailing slashes removed site-wide, automated checks added, rule written into editorial standards so it never recurs.

Case 2: Template Residue Prevention

Problem: Incomplete replacement of template content during article writing caused incorrect star information to go live.

Discovery: Caught by the editor-in-chief during file diff review.

Fix: Logged to quality feedback system; editorial standards updated with rule "titles and star content must match after writing"; automated tools now scan for template residue.

Case 3: Multi-Model Astrology Logic Validation Introduced

Problem: Single-tool validation had blind spots — rare patterns or complex Four-Transformation combinations could slip through undetected.

Discovery: Contradictions between earlier and later analyses surfaced during case backflow comparison.

Fix: Multi-model cross-validation introduced. Over 100 hard rules are now checked by multiple tools simultaneously — any tool flagging an anomaly triggers a review hold, significantly reducing edge-case misjudgment rates.

Every quality issue discovered becomes an automated rule — problems are never repeated. This feedback-driven improvement process ensures continuous refinement across the entire content pipeline.

Transparency Resources

We are committed to full transparency in our content production process. The following resources are publicly available:

External standards and primary sources

These primary sources inform this page. They are benchmarks, not third-party endorsements of this site.

Current Machine Audit Snapshot

This block uses only traceable local audit data. No unsupported metrics or model claims are added.

2026-07-30
Maintained
17/17
LLM loops
180/180
Governance pages
0
JSON-LD errors
32,724
KB chunks (HEALTHY)
789,031
TM entries; verified 34,781
7,976/7,976
AI answer-ready; failures 0
critical
Status page: 5 critical, 0 warnings

Verifiable Evidence Layer

This block is not a narrative claim. Each core assertion has a claim id, source JSON, hash, and a repeatable verification command. Public pages disclose governance evidence without exposing source code, secrets, private data, or exploitable attack details.

Claim IDVerifiable valueStatusOwnerSource and verification
claim.public-url-manifest.indexable-count
Public URL and canonical inventory
38,965 indexable URLs verified sitewide node scripts/generate-public-evidence-manifest.js --dry
claim.trust-pages.audit-pass-rate
Trust page machine audit
180/180 pass verified sitewide node scripts/verify-trust-pages.js --check
claim.discovery-surface.zero-errors
AI discovery surface audit
{"errors":0,"warnings":0} verified sitewide node scripts/verify-discovery-surface.js
claim.structured-data.jsonld-errors
JSON-LD / structured data audit
{"structured_data_invalid_files":0,"breadcrumb_count":28274,"faq_count":27506,"dataset_count":30,"article_count":27406} verified sitewide node scripts/site-machine-audit.js
claim.status.sla-state
Status page SLA source
critical / 5 critical, 0 warnings verified sitewide node scripts/generate-status-page.js
claim.provider-alignment.openai-anthropic-gemini
OpenAI / Anthropic / Google Gemini benchmark alignment
benchmark alignment only unless code/config evidence exists verified sitewide node scripts/verify-public-evidence.js --check
claim.transparency-report.sha256
Transparency report SHA-256 anchor
{"report":"transparency/report-2026-Q3.json","sha256":"47b09e2ca4e8b8fe9dffdfaccef3b11212de9ee3a8a14badca8044e2481203c5"} verified sitewide node scripts/update-transparency-current-data.js
claim.release-integrity.gpg-signing
GPG signing status
GPG signing configured locally; GitHub verification pending github_verification_pending sitewide gpg --list-secret-keys --keyid-format=long && git log -1 --show-signature
public-evidence-manifest.json public-claim-registry.json public-verification-report.json public-url-manifest.json

System Card V2.0: Technical Transparency Layer

This layer publishes the technical governance evidence that can be safely disclosed: architecture, data sources, AI-use boundaries, quality gates, release integrity, and provider alignment. Source code, secrets, exploitable attack details, and private data remain out of scope.

Public architecture

Cloudflare Pages/Workers, R2/D1/KV/Pagefind, and local generation scripts form the public-site and governance publication chain. Public pages disclose behavior, state, and traceable sources, not secrets or internal permissions.

AI-use disclosure

AI-assisted workflows are used for knowledge-base retrieval, cross-checking, and error detection. Governance documents are benchmarked against OpenAI, Anthropic, and Google Gemini public frameworks. Production model usage is disclosed only when code/config evidence exists.

Quality and safety gates

Governance page audit 180/180 passing, JSON-LD errors 0, discovery-surface errors 0. Status pages report critical / 5 critical, 0 warnings as-is.

Data traceability

Knowledge base 32,724 chunks, TM 789,031 entries, AI answer-ready 7,976/7,976. Public metrics trace to data/state-machine/*, data/*audit*.json, and transparency reports.

Governance areaOpenAIAnthropicGoogle GeminiStarnum implementation evidence
Model/system-card disclosureOpenAI models + safety docsClaude model docs + system/model cardsGemini model docs + safety settingssystem-card, model-card, methodology, benchmark, transparency-log
Safety evaluation and use boundariesSafety best practices / deployment checklistResponsible Scaling / safety policyGemini safety controls / policyAI safety, acceptable-use, ethics, risk-boundary copy, crawler policy audit
Data governanceData controls / privacy controlsprivacy and data handling docsGemini API data governance referencesprivacy, ai-data-governance, KB/TM source tracking, SHA-256 hashes
Monitoring and releaseproduction checklist / eval disciplinesystem-card transparency disciplinemodel/version documentation disciplinedeploy.js, status.html, SLA report, trust-pages-machine-audit, sitemap/hreflang audits

The V2.0 goal is not more claims; it separates implemented controls from planned controls. Production usage, benchmark alignment, status exceptions, GPG signing, and SLA breaches are disclosed from source data.

Release Integrity And GPG

GPG signing configured locally. signingkey=0934DFA0EDA6363A. GitHub verification pending until the public key upload and Verified badge are confirmed.

OpenAI / Anthropic / Google Gemini Alignment

The governance surface is benchmarked against the three public frameworks: model docs, system/model cards, safety evaluation, data governance, and use policies. This is benchmark alignment, not a claim that every provider is active in production inference. Official docs checked: 2026-07-30

ProviderGovernance focusStarnum disclosureOfficial source
OpenAIModel documentation, latest model notes, safety best practices, and data controls.No verifiable production model setting was found in the production code scan; providers are listed as governance benchmarks.https://platform.openai.com/docs/models
AnthropicClaude model documentation, system/model cards, Responsible Scaling, and safety policy.No verifiable production model setting was found in the production code scan; providers are listed as governance benchmarks.https://docs.anthropic.com/en/docs/about-claude/models
Google GeminiGemini API model documentation, safety settings, data governance, and platform policy.No verifiable production model setting was found in the production code scan; providers are listed as governance benchmarks.https://ai.google.dev/gemini-api/docs/models