[tickernelz]

Summary

README audited against main @ 1055d37f42f9deee3e0377f7be23b05b9747fa9e (v2.40.1 + #486).

Stale, now fixed

• Core Features said "multi-provider AI support (OpenAI, Anthropic)". src/config.ts has 10 manual memoryProvider values plus any OpenCode provider via opencodeProvider. Rewritten as a list covering features that exist in code (Markdown vault, update-on-capture, dreaming, credential redaction, encryption, export/import/migrate, Basic Auth).
• Dreaming ended with "Deeper related-threshold dreaming (update/supersede/flag, cross-project) lands in the follow-up PR for Feature Request: keep memories current — update-on-capture and deeper dreaming (follow-up to #460) #464". That shipped in feat(dreaming): related-threshold consolidation with supersede and flag #466 (dreaming-service.ts actions merge|update|supersede|keep|flag, dreaming.crossProject). Removed.
• Manual memory modes table lacked help (tool enum in src/index.ts / src/v2/adapter.ts).
• Web UI section listed only timeline/profile; the UI has 5 routes (web/src/lib/routes.ts). Added the views, webServerPort, webServerEnabled.
• Encryption said AES-256-GCM only; databaseEncryptionCipher accepts 7 ciphers (default aes256gcm).
• Dev commands now bun install / build / test / check (matches package.json scripts and AGENTS.md), plus how the site/docs are generated and the site/docs-map.json rule.

Added

• Website and Docs badges at the top (dropped from docs by the existing dropLines), Website/Docs/npm in License & Links.
• "Commonly tuned top-level options" table; defaults taken from DEFAULTS in src/config.ts, meanings from opencode-mem.example.jsonc. autoCleanup wording checked in cleanup-service.ts (once a day, updated_at cutoff, pinned skipped).
• No headings added or renamed, so site/docs-map.json is unchanged.

Translations (.github/i18n/README.{de,zh,ar,tr,nl}.md): updated for the changes above and for older gaps a token check found: recall/recallScope/retrieval_score rows and paragraphs, embeddingIdleTimeoutMs, the missing-config toast, autoCaptureProviderOrder, Session.NotFoundError and V2 event bullets. After the update, every backticked token of every English section exists in the matching translated section (35/35 sections, 5 languages). "Memory credential redaction" stays untranslated as listed in docs-map.json.

package.json: homepage and bugs.url only. release.yml runs on push: tags: v*, so this does not publish. files, version, scripts and dependencies are unchanged.

Site metadata: the live landing page already had canonical, OG, a Twitter card, JSON-LD and lang; docs had canonical, hreflang, OG and Twitter via Starlight; robots.txt and the sitemap were already in place. This PR adds:

• Landing: og:site_name, og:locale, og:image:width/height/alt, twitter:image:alt, rel=sitemap. JSON-LD gains description, image, sameAs (GitHub, npm), author (copyright holder in LICENSE) and isAccessibleForFree.
• Docs: og:image:width/height/alt.
  No ratings or counts were added.

Size and Lighthouse (same local static server, mobile, median of 3)

Page	HTML raw before → after	gzip before → after	Perf	A11y	BP	SEO	LCP
/	66189 → 67227 B	15645 → 15822 B	98 → 98	100 → 100	100 → 100	100 → 100	1951 → 1952 ms
/docs/	37442 → 37697 B	8838 → 8924 B	97 → 97	100 → 100	100 → 100	100 → 100	2177 → 2176 ms

Test plan

• cd site && bun run build && bun run check && node scripts/check-dist.mjs: build OK (158 pages; docs-map and anchor checks pass), astro check 0 errors, check-dist OK.
• prettier --check . clean; pre-commit tsc --noEmit + lint-staged passed.
• npm pack --dry-run: only LICENSE, README.md, package.json and dist/.
• Built dist/index.html and dist/docs/index.html heads inspected; JSON-LD parses.

Checklist

• Conventional Commits
• README.md and the 5 translations updated together
• No new claims that the code at 1055d37 does not support