duolingo.id — Implementation Plan

Phase 0 — CF wiring (5 min)

• Zone exists on CF (domain registered). Account f5dbc4f1c3e4f3fbdacb83107b8088da.
• DNS: apex A 192.0.2.1 proxied + www CNAME apex proxied (placeholders; Worker handles all).
• Worker routes after first deploy: duolingo.id/*, www.duolingo.id/*.
• Whitelist duolingo.id in web/xmoj.com/worker-cf/src/whitelist.js if using kie z-image piped through xmoj.

Phase 1 — Repo scaffold

Directory: web/duolingo.id/{worker,web,seed,scripts}. See prd.md for visual decisions.

Phase 2 — D1 schema

See schema.sql. Core tables: scripts, characters, words, units, lessons, exercises, users, user_lessons, user_state, srs_items, league_cohorts, quests, friends, events.

Binding DB → duolingo-db. Create with wrangler d1 create duolingo-db.

Phase 3 — Bindings

See wrangler.toml. Bindings: DB (D1), CACHE (KV), MEDIA (R2), AI (CF Workers AI), ASSETS (Pages-style asset binding for built SPA). Secrets: KIE_API_KEY, OPENROUTER_API_KEY, JWT_SECRET, GOOGLE_OAUTH_CLIENT_ID, GOOGLE_OAUTH_CLIENT_SECRET, WA_RELAY_TOKEN, CRON_SECRET.

Phase 4 — Worker API routes

Phase 5 — Indonesian copy mapping

Phase 6 — Content authoring workflow

1. Hardcode 23 Lontara chars in seed/02_lontara_chars.sql.
2. AI-generate words/sentences via scripts/gen_content.mjs (OpenRouter Gemini Flash Lite).
3. TTS each char + word via local pro/tts-service → R2 cache by SHA1.
4. Stroke-order SVGs scraped from Wikipedia (CC-BY where applicable).
5. Hourly cron auto-generates next lessons so users never hit a "no more content" wall.

Phase 7 — Lesson player UX

Each lesson = 6–10 exercises (~3 min). Types: translit, choose, tap-pair, listen, build-word. Each wrong = −1 nyawa. End-of-lesson: confetti + Garu cheer + XP summary.

Phase 8 — Engagement loops

• Daily: streak meter, 3 daily Misi, web push at 19:00 WIB.
• Weekly: league promotion Sun 23:59 WIB, weekly Misi.
• Social: friend streaks, share-progress link.
• Pay: Super (Rp 39k/mo, Rp 299k/yr), gems IAP.

Phase 9 — Deploy steps

cd web/duolingo.id
wrangler d1 create duolingo-db                 # → fill database_id in wrangler.toml
wrangler kv namespace create CACHE             # → fill id
wrangler r2 bucket create duolingo-media
wrangler d1 execute duolingo-db --remote --file=schema.sql
for f in seed/*.sql; do wrangler d1 execute duolingo-db --remote --file=$f; sleep 5; done
cd web && npm install && npm run build && cd ..
wrangler secret put KIE_API_KEY
wrangler secret put OPENROUTER_API_KEY
wrangler secret put JWT_SECRET
wrangler secret put CRON_SECRET
wrangler deploy

Verify:

curl https://duolingo.id/api/health   # → {ok:true}
curl -I https://duolingo.id/          # → 200, HTML

Phase 10 — Beads epic

File ucok-duo once scaffold lands. Children mirror this plan's phases.

Risks

See prd.md §10.

Step 1 done = SUCCESS criteria

• https://duolingo.id/ serves SPA shell
• /api/health returns {ok:true}
• Lontara unit 1, lesson 1 loads + grades a translit exercise
• League cohort row exists for first signup
• Lighthouse mobile Perf ≥ 85, SEO 100

⚙ HARD CONSTRAINTS (enforced for all sites)

This domain MUST operate within these constraints — no exceptions:

• 100% Cloudflare serverless — Workers + D1 + R2 + KV + Workers AI + Vectorize. NEVER PM2, NEVER VPS, NEVER Docker in production path.
• 100% AI-automated — every customer interaction, every moderation decision, every transaction reconcile = AI. No manual queue, no live human chat support, no physical fulfillment.
• 1-operator solo — one person can run the entire operation from a phone. No team meetings, no shared inbox, no shift rotation.
• WhatsApp AI bot for all support (24/7, instant response, no SLA promises that need humans).
• Mayar QRIS for all Indonesian payments (subscription auto-renew, no manual invoicing).
• Indonesian UI primary — bahasa-first, English fallback only where unavoidable.
• Privacy — opt-in only, delete-on-request honored within 24h (cron-driven).
• No physical goods, no inventory — digital products + affiliate referrals only.

If the plan above describes any flow that violates these constraints, treat the plan as ASPIRATIONAL only and rework before building. The constraint trifecta wins.