Development Log — 2026-08-16 · Public Gateway v0.7

What happened

Turned the public gateway into a public knowledge graph that search

engines can discover, read, understand and select correctly — without any

fake Google claims.

Search knowledge model (knowledge/)

  • 17 system knowledge blocks with real EN / ไทย / 中文 content

(aenet-cloud-public, aenet-core, agent runtime, model router, memory,

spine, K Language, KAP, public API, snapshot layer, public time,

development queue, early access, XGRID, RGBM, UI contract, visit memory).

  • Every fact is evidence-based; systems without public evidence

(agent runtime internals, model router, memory, spine, K Language) are

marked insufficient-public-evidence — never invented.

  • Architecture relationship blocks, concepts, vocabulary.json and the

canonical relationships/system-graph.json (single source for related

systems).

Go Search Compiler (search/)

  • Build-time compiler (Go standard library only) with

audit / build / sitemap / validate / report commands.

  • Reads only allow-listed public sources (knowledge/, content/,

public-data/, ui-contract/, design/, web/locales/).

  • Generates 82 static pages under web/{en,th,zh-cn}/, per-language

sitemaps, sitemap-index.xml, robots.txt, JSON-LD, a search graph, a

search manifest and readiness reports.

  • Deterministic: identical inputs → identical HTML (verified by hash diff).
  • 9+ Go test packages (canonical, hreflang, sitemap, metadata,

graph, schema/structured data, scoring, renderer,

url_registry/config).

Search correctness

  • Locale-first URLs (/en/, /th/, /zh-cn/); each language canonicals to

itself; reciprocal hreflang sets; x-default → English.

  • Research and development entry bodies are English-only for now — localized

indexes link to the English canonical, and no mostly-English localized

URLs are generated (translation-pending policy).

  • The internal AENET Search Readiness Score (97/100) is an audit metric,

explicitly not a Google ranking score. Zero build defects, zero

warnings.

Commands

npm run search:build   # Go compiler + public build
npm run search:check   # go test + validate + artifact checks
npm run search:preview # build + readiness report
npm run build          # full pipeline (content → Go compiler → snapshots → verify)

Verification

  • go -C search test ./... — all packages pass.
  • npm run search:check — PASS (34 checks).
  • npm run check — all 34 verification groups pass.
  • npm run build — deterministic; wrangler deploy --dry-run — 223 assets,

25.38 KiB.

  • Browser-verified /en/system/kap-language, /th/..., /zh-cn/...,

/en/, /en/systems — localized content, self-canonical, reciprocal

hreflang, real language links, breadcrumbs, JSON-LD.