Scan one business by name, or a category and city, from Google Maps (Apify compass~crawler-google-places) | /owners, /dashboard, pnpm scan | Live, 2026-10-26 (12 Austin restaurants; by-name scans). Fixture pipeline tests |
| Diagnosis where every finding cites a real review, listing fact or website quote; a code check drops any quote that is not found word for word | /businesses/[id]/diagnosis | Live with xAI grok-4.7, 2026-10-26. Citation tests; eval "Diagnosis cites real reviews" |
| Your peers: anonymized Austin tier list and leaderboard, with the tier formula shown | diagnosis page | src/data/peers.json: 111 anonymized Austin businesses in 6 industries, scanned 2026-10-26. Eval "Peers stay anonymized" |
| Game plan: scorecard, the one binding constraint, tiered steps labeled "Ringy does this", "Ringy does this after your OK" or "You do this" | diagnosis page | Tests; eval "Overlap rule" |
| Adjustable ROI calculator: every input shows where it came from (listing, vendor pricing page, or labeled assumption); edits stay on screen and are never saved | diagnosis page, business page, 1-pager | src/data/calculator.json; tests calculator.test.ts, figures-traceable.test.ts |
| 30-day game plan: up to 3 focus fixes, weeks 1 to 4 split into "Ringy does this" and "You do this", each with a starting number and its source | diagnosis page (#thirty-day), 1-pager page 2 | Tests thirtyday.test.ts on the real Aire Serv snapshot |
| Published studies per recommendation: 6 sourced case studies, each with its URL and the exact quote; vendor sources labeled "Vendor-reported"; problems with no source show none | diagnosis page, game plan items | src/data/case-studies.json (checked 2026-10-27); scripts/verify-case-studies.mjs re-fetches each page and finds the quote; tests case-studies.test.ts |
| Three ways to get each fix done: do it yourself (time cost, labeled estimate), Ringy does it (after your OK, price not set), or hire it out (a category; vendors and prices named only with a public pricing page, no affiliation) | game plan items | src/data/fix-paths.json + solutions.json; tests fixpaths.test.ts, figures-traceable.test.ts |
| Talk to Ringy: the owner picks a channel and ticks a consent box; the exact consent text and time are stored (append-only, withdraw any time, contacts masked on read). Browser voice is live; email, phone, SMS and WhatsApp are a waitlist and nothing is sent | diagnosis page (#talk), POST /api/businesses/[id]/channels | Migration 20261030000000_channel_optins.sql (RLS on, no client access); tests channels.test.ts against the local store |
| Printable 1-pager (no price, no money-back promise) | /businesses/[id]/one-pager | Tests; screenshots |
Owner interview: Ringy asks the questions from src/data/interview.json (and the game plan's open questions), the owner answers by voice in the browser or by typing, and answers are saved to the business profile | /businesses/[id]/interview | Tested end to end in a real browser (headless Chrome) against a real ElevenLabs Conversational AI session with a scripted microphone (recorded answers played into the mic input), 2026-10-26: all 5 questions asked in order, all 5 answers transcribed, mapped, saved and shown after reload (scripts/e2e/interview-voice.mjs). Not yet tried by a person on a real microphone. 18 unit and route tests |
| Weekly check-in: light rating and review-count check, new reviews only when the count went up, a deeper refresh monthly, under a monthly cap (default $1.00) | Vercel Cron /api/cron/checkin, Mondays 15:00 UTC | Tests with mocked Apify (no spend). Needs CRON_SECRET in the deploy env; not set here, so not run live |
| Phone-agent config per business: AI disclosure first, every fact numbered and cited, rules for unknown answers, bookings, after hours, upset callers | business page | Tests; simulated caller evals 10/10 (fixture), 8/8 live LLM on 2026-10-26 before the 2 new cases were added |
| Demo-safe fallback: if Supabase, Apify or xAI fail or hang (8 s timeout), the Aire Serv flow renders from a committed real report labeled "Saved snapshot from Oct 26, 2026", with editing, voice and saving off | every Aire Serv page | Tests simulate each failure; screenshots with Supabase pointed at a dead port |
| Eval panel: rule checks run on real saved reports, plus simulated callers | /evals, pnpm eval | 8/8 rule checks and 10/10 callers at the last commit |
| Customer mode: nearby businesses from OpenStreetMap on a map, and chat with a business's AI built only from cited facts | /find | Tests; bundled Austin OSM snapshot when public Overpass mirrors are down |
