Acceptance run v0.1.0
Version: 3 · 2026-10-02 · every row of the acceptance contract checked against the live product for the v0.1.0 tag
This is the record of the v0.1.0 acceptance run against the acceptance contract version 2. Every row was checked on production: the web app at https://app.bankroll40.com and the API at https://api.bankroll40.com, both deployed by the pipeline from main. The API checks ran against f3018b63 and the last browser check against 221f3af8. The run used fresh SES simulator accounts (success+accept-ui@, success+accept-api@ and success+accept-api2@simulator.amazonses.com), and the purge script removed them from D1 and R2 afterwards.
How each row was checked
- API checks: a script signs in through a real magic link, then drives the API the way the app does and reads production D1 to confirm what was stored. It ran 22 checks; all passed on the final run.
- Browser walks: Playwright with Chromium against the live site. Every walk starts with a signed-out cold load of
/and/session/liveat 390 px in both themes, then signs in and screenshots each screen at 390 and 1280 px in light and dark. Page errors and horizontal overflow fail the walk. Seven walks ran; none saw a page error. - Unit and API tests: the full suite on
mainat 221f3af8, 208 tests, all passing. Where the contract names a test as evidence, the row cites it.
Results: pass means the evidence holds on production. Fixed means the row failed, the fix shipped through its own merge request, and the row passed on a re-check. Owner means the check needs something only the owner has. Deferred v0.2.0 means the release page puts the row in v0.2.0.
Money and currencies
| # | Result | What was checked live | Evidence |
|---|---|---|---|
| M1 | fixed | Every *_minor column in production D1 is INTEGER, and every table with one carries a currency column on the same row. The first run found session_buyin and tournament_entry without one. |
app !19 (issue #16) added the column, set it on every insert and backfilled from the parent row. Re-check: 14 money columns across 22 tables, all INTEGER, none without a code; no blank codes in 31 buy-ins and 36 bullets. Schema test "every money column is an integer in minor units and sits beside a currency column" has no exemptions now. |
| M2 | pass | A past HKD 25/50 session at Wynn Palace logged against a USD pot. | Stored currency HKD, fx_to_pot and fx_to_home 0.12743 (USD home), home_currency USD. API test "a HKD game from a USD pot stores the rate to the pot and to home". |
| M3 | pass | Home currency switched USD to GBP and back. | Total went from 20,191.15 USD to 15,257.44 GBP; every session row was byte-identical before and after. API test "changing the home currency changes the total and rewrites no row". |
| M4 | pass | A 500 USD transfer from Main to a HKD pot. | Two rows sharing one link_id, each naming the other pot in counter_pot_id, both with fx_applied 7.8472; 500.00 USD out, 3,923.60 HKD in. |
| M5 | pass | fx_rate in production, and MOP, which ECB does not publish. |
ECB rows for HKD, GBP and PHP on 2026-10-01 (USD base), written by the fill path on first use. MOP answered 404 until a rate of 8.05 was typed, then 8.05 one way and 0.12422 the other. The first cron run, 30 16 * * * on 2026-10-02, captured with wrangler tail: outcome ok, log {"event":"fx.fill","day":"2026-10-02","stored":["GBP","HKD","PHP"],"missing":[]}; fx_rate then held ECB rows for 2026-10-01 and 2026-10-02. API test "the daily fill stores USD, HKD, GBP, PHP and every pot's currency". |
The discipline engine
| # | Result | What was checked live | Evidence |
|---|---|---|---|
| D1 | pass | The rule set of a pot created by sign-up. | 30, 40, 50 of the next stake plus 100 h, stop-loss 3, cool-off 12 h, session cap 8 h, re-entry cap 2, budget 5 percent, as in DEFAULT_RULES. |
| D2 | pass | The start screen of a pot holding 87.5 buy-ins at 1/2. | 1/2 and 2/5 open; 5/10, 10/20 and 25/50 greyed with "$12,510 short", "$42,510 short", "$132,510 short". Screenshot 51-start-stakes. API test "the stake list greys out what the roll does not keep open, with the shortfall". |
| D3 | pass | $6,150 withdrawn through Pots from a $20,000 roll at 2/5. | Home: "2/5 ↓ 1/2, 27.7 buy-ins · $20k brings 2/5 back". Screenshot 12-home-red-arrow. |
| D4 | pass | A live session to 3 buy-ins, in the app and through the API. | App: the nudge one buy-in before, then "Stop-loss reached: rebuy paused"; the fourth buy-in asked for a reason. API: the fourth buy-in without a reason answered 409 reasonRequired; with one, 201 and a stop-loss rule break. Screenshots 17-live-warn, 18-live-paused. |
| D5 | pass | The next start after the stop-loss session. | API: 409 with a cool-off hold until 12 h after the end. Home: "Cool-off: next session in 11 h 58 min". Screenshot 50-home-cool-off-next. |
| D6 | pass | The Metro Poker Manila event, 15 bullets. | Bullets 1 to 3 passed; bullet 4, the third re-entry, was held for a reason; 12 reentry rule breaks for the 12 bullets past the cap. App walk: the fourth bullet of DeepStack Main held on screen. Screenshot 32-fire-held. |
| D7 | pass | The same event against the month's budget, 5 percent of the roll. | Bullet 4 took the month's spend over the cap and was held when fired, alongside the re-entry hold; each of the 12 bullets past the cap became a tournament-budget break. API test "is checked the moment a bullet is fired, against 5 percent of the cash pot". |
| D8 | pass | A softer stop-loss (4) during a live session, then after it ended. | 409 while live; 200 cold, with a rule-change rule break carrying the note. |
| D9 | deferred v0.2.0 | The weekly report is v0.2.0. The numbers it will show already come from /api/stats. |
/api/stats live: clean 3 sessions, +$739.15; broken 1 session, −$2,000, 1 break; tournaments 24 breaks. API test "D9 in numbers". |
| D10 | pass | Home's next-stake line. | "2/5 opens at $25k · $7,510 to go · 5 of 100 h logged" before, "open now · one tap on the card above" after 100 hours. Screenshots 50-home-cool-off-next, 52-home-green-arrow. |
| D11 | pass | Tournaments screen and /api/tournaments/budget. |
"Max buy-in today: $136.90 · $13,690 in the pot spread over 100 buy-ins"; the API names limitedBy: roll. Screenshot 30-tournaments. |
| D12 | pass | A fresh account answered the questions on the live site: 20,000 USD at 2/5, 12 h a week. | Every card on screen equals suggestSetup for the same answers. The stored setup holds hoursPerWeek 12, and its safe roll, stake, reserve and next stop equal suggestSetup. A roll held as 12,000 GBP at a typed 1.35 gave the API and suggestSetup the same 16,200 safe roll, 2/5 and 34 months. Hours a week shipped in app !17. Screenshots 03 to 05. |
| D13 | pass | Red arrow: Stay with a note, then Play 1/2. Green arrow after 104 h at 1/2 and a deposit to $25,001: Play 2/5. | Each tap changed the verdict; the Stay note is in rule_break as stake "reading the table, one more night at 2/5". Screenshots 12, 13, 14, 52, 53. |
Sessions, tournaments, online
| # | Result | What was checked live | Evidence |
|---|---|---|---|
| S1 | pass | A live session with the browser offline: reload, timer, end with cash-out, then back online. | The timer kept running after an offline reload; all four buy-ins, the stop-loss note and the end reached the API in order on reconnect; result −$500. Screenshots 19-live-offline, 20-home-offline. A phone in airplane mode is the owner's V2. |
| S2 | pass | An online session at GGPoker, 1,200 hands, 4 tables, through the API and the Log screen. | Row has site, hands, tables and no venue; 2 bb/100 through the API, 10 bb/100 in the app walk's session. Screenshot 22-log-online. |
| S3 | pass | The Metro Poker Manila case: 15 bullets over flights 1A to 1D, bullet 7 bag paid $1,500, bullet 15 bagged on Day 1D, Day 2, closed. | Open after Day 1 and after Day 2 was logged; closed with 15 bullets and 5 days. The app walk closed a 4-bullet event the same way. Screenshots 34-event, 36-event-closed. |
| S4 | pass | The closed Metro event and an open online event. | ROI 65.2 percent over all 15 bullets; closed summary counts 1 event, the open one shows as 1 event in play. |
Media, profile, journal
| # | Result | What was checked live | Evidence |
|---|---|---|---|
| P1 | pass | A receipt on a deposit through the API; a receipt on a bullet and a photo on a cash-out through the app. | Signed PUT 200, complete 200, row ready under u/<user id>/, same bytes downloaded. Screenshots 33-capture, 38-session-media. |
| P2 | pass | A JPEG carrying Manila GPS coordinates (14°31′N, 121°3′E) attached through the web capture screen. | The object stored in R2 has no EXIF and no GPS block. The app lane had already verified the same on a real upload. The native-device version of this check comes with the v0.3.0 builds. |
| P3 | pass | Delete the receipt. | 200; the row is gone and a still-valid signed URL then answers 404. |
| P4 | deferred v0.2.0 | The journal timeline is v0.2.0. | Not in this release. |
| P5 | pass | Media without a session, and a signed URL past its expiry. | Private by default holds: 401 without the cookie; a 1 s URL answered 403 after 3 s. The shared case is not built; per-item sharing arrives with the v0.2.0 share card. |
Access
| # | Result | What was checked live | Evidence |
|---|---|---|---|
| A1 | owner | Magic link and passkey on the web. Google and Apple are hidden until their secrets exist. | Passkey registered under Me and used to sign in alone, with Chrome's virtual platform authenticator. Screenshots 41, 42. The owner needs to: add a passkey on a real phone; supply the Google OAuth client id and secret; supply the Apple Services ID, signed client secret and developer account. /api/auth/providers answers google false, apple false. |
| A2 | pass | The audit rows of the API account after every kind of write. | 62 audit rows over user, setup, pot, rule_set, transaction, fx_manual, session, tournament and media, all actor_kind app, none without one. |
| A3 | pass | /api/export.csv per table against D1. |
Equal counts for pots 2, transactions 4, sessions 4, buy-ins 6, events 2, bullets 16, days 5, rule breaks 26, rule sets 2, venues 4, manual rates 1, audit 62. |
Deployment and verification
| # | Result | What was checked live | Evidence |
|---|---|---|---|
| V1 | pass | Cloudflare's deployment history of both Workers against the main pipelines. |
Each of the last 10 deployments of bankroll40-api and of bankroll40-app, 20 in all from 05:33 to 14:28 UTC, falls inside a deploy-api or deploy-app job of a main pipeline. Worker secrets were set once, at bootstrap. |
| V2 | owner | The owner's own session at a Manila venue with the tunnel off. | Needs the owner at a venue with the phone. |
| V3 | pass | The signed v0.1.0 tag and the outcome note. |
Tag v0.1.0 (object 8e81ceed) on main at 221f3af8, signed, git tag -v reports a good signature; the outcome note is on app issue #15. |
Fixed during the run
| Merge request | What it fixed |
|---|---|
| app !17 | Sign-up asks hours a week instead of sending 20; GET / on the API redirects to the app |
| app !18 | The purge script for simulator accounts (cleanup before and after this run) |
| app !19 | M1: buy-in and bullet rows carry their own currency code |
| app !20 | Stats read "1 sessions" for a venue with one untimed session |
| app !21 | After the tag: Stats and Home read /api/stats in the player's timezone (the phone used UTC months); Pots show money in play. Tagged v0.1.1 |
Screenshots
The full set is attached to docs issue #1: acceptance-run-v0.1.0-shots.zip, 134 PNGs. Each name below exists at 390 and 1280 px in light and dark, except where noted.
| Shot | Screen |
|---|---|
| 00-cold, 00-cold-session-live | signed-out cold load of / and a deep link, 390 px, both themes |
| 03-questions-empty, 04-questions-filled | the eight sign-up questions, hours a week in the first |
| 05-suggested-setup, 06-home-after-setup | the suggested setup and Home right after Accept |
| 10-home, 11-pots | Home at $20,000 on 2/5; Pots after a withdrawal |
| 12-home-red-arrow, 13-home-stay-note (390), 14-home-moved | the red arrow, the note for Stay, after Play 1/2 |
| 15-start, 16-live, 17-live-warn, 18-live-paused | start, live, one buy-in from the stop-loss, rebuy paused |
| 19-live-offline, 20-home-offline (390) | live session and Home with no network |
| 21-log, 22-log-online, 23-stats, 24-me | past live and online sessions, Stats, Me |
| 25-stats-untimed-venue | Stats after the fix in app !20: "1 session" for an untimed venue |
| 30-tournaments, 31-event-new (390), 32-fire-held, 33-capture (390) | budget and max buy-in, a new event, a held bullet, the capture screen |
| 34-event, 35-day-close, 36-event-closed | Day 2 in play, the close form, the closed event |
| 37-session, 38-session-media | a session and its gallery |
| 40-profile-pictures | avatar and cover |
| 41-passkey-saved, 42-passkey-signed-in (390) | passkey added under Me, then used to sign in |
| 50-home-cool-off-next, 51-start-stakes | the cool-off countdown and next-stake line, the start screen's greyed stakes |
| 52-home-green-arrow, 53-home-moved-up | the green arrow and the one-tap move up |
Cleanup
Production held only simulator accounts before the run. apps/api/scripts/purge-simulator-users.sh removed them on 2026-10-02 at 14:11 UTC: 209 rows, 3 of them users, and 5 R2 objects, with every table losing exactly its simulator rows (app issue #14). The three accounts this run created were removed the same way at 14:30 UTC: 397 rows, 3 of them users, and 4 R2 objects. Production then held no users and an empty media bucket.
After the v0.1.0 tag
Box 3's report raised two wiring gaps that landed after v0.1.0, in app !21, tagged v0.1.1 (0c264863). First, the phone computed Stats and Home's this-month line itself, with UTC months. Second, Pots ignored inPlay. A live check on a fresh account set to Asia/Manila passed: Home counted a session that ended at 01:30 on 1 October Manila time ("+$300 · 4 h · 15 bb/h"), Pots read "$230 in play in 1 open tournament", and Stats matched /api/stats. Shots 60 to 62 are in the zip on docs issue #1. The account was purged afterwards: 18 rows. The Me screen already read its headline from /api/profile.