#38

Demo script (RAI-264), architecture diagram, judge-demo hosting steps

Drafttobi-techywants to merge intofrom

Docs and one check script only. No code path changes, N0, nothing deployed.

What is in it

  • docs/DEMO-SCRIPT.md (RAI-264): six scenes, 2:45 total, on the seeded shops shop-040 and shop-008 (receipts), shop-035 (fake f08 then genuine g02), shop-014 (one tap + voice correction mid-task), shop-011 (credit with SAM 3.1 shelf count) and the owner. For each: the exact events, messages and fixtures the sim sends, Mimi's recorded replies, the console screen, the voiceover and the timing. Plus a 10 s opening line (before vs this week) and a shot list (desktop console, phone-size "simulated shop phone", "synthetic data" visible).
  • scripts/demo_script_check.py (make demo-check): runs the e2e replay through the real API, store and workers (recorded Muse Spark, SAM 3.1 and persona answers, N0) and checks every line the script quotes, plus the owner question's route on the demo-mode API with the offline model. 16/16 hold.
  • docs/architecture.mmd and docs/architecture.png, Mermaid in the README: shop phone or simulator to /v1/turns, front door, router, helpers (Order, Collector, Credit, Stock, Proof), agent loop with budgets, claims gate, durable store and queue, console; Muse Spark, SAM 3.1, Muse voice, Spitch. Built this week vs before the event vs not live yet.
  • docs/DEPLOY-DEMO.md: the API in MIMI_MODE=demo from the Dockerfile on port 3000 (GET /), SQLite in /tmp, approved caps (N100 per address a day, N2,000 a day, 1 h sessions, 20 per address), kill switch; Atlasflow and Railway steps with prices read 30 Oct; the console build to demo.userail.money on Cloudflare; checks after deploy.
  • JUDGES.md: a placeholder for the hosted demo link.

Found while checking (documented, not fixed here)

  1. Scene 1's recorded reply is typed Yoruba, not a Pidgin voice note; a voice take has no recorded verdict, so it cannot replay for N0.
  2. Scene 5's answer needs a live Muse Spark call; only its route is checked offline.
  3. Demo mode refuses approving a credit review (403, by design), so scene 4's approve click must be filmed on a live-mode run.
  4. MIMI_ENV=production refuses SQLite, so the approved SQLite-in-/tmp hosting runs with MIMI_ENV unset; the salt and pseudonym key are set by hand.
  5. /tmp resets the day's cap counts on every restart; Spitch TTS spend is not inside the naira caps.

Merge needs Tobi's yes.

Activity

2
coderabbitai[bot]1d ago

IMPORTANT

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

  • Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

This pull request is a draft

Merging is unavailable until it is marked ready for review.