Skip to content

Prod Release Sprint 168 — Pix on Avenia (gated), card collateral return, QR pay gate (2026-11-16) - #4115

Draft
abalinda wants to merge 538 commits into
mainfrom
dev
Draft

abalinda wants to merge 538 commits into
mainfrom
dev

Conversation

@abalinda

@abalinda abalinda commented Nov 16, 2026 •

Copy link
Copy Markdown
Contributor

Release page: https://app.notion.com/p/419838117579817f8c2eef8e30679979

Draft until the approver's go. Opened by the captain (Aleks) on 16 Nov. The scope follows dev until the captain announces the freeze in #devs-pls; this list is refreshed whenever the head moves. main has one commit that dev lacks: the content publish #4109 (content submodule only). This PR merges cleanly over it, and the post-release back-merge (step 19) brings it to dev.

  • Migrations: none in this repo. The API release (peanut-api-ts) carries 7 and merges first.
  • Native surface changed: yes, so this is a store build only. android/app/src/meawallet/java/me/peanut/wallet/PushProvisioningPlugin.java changed (fix(card): refresh exact Google Wallet status on Android #4066). release-ota.yml stops at "Preflight OTA native compatibility" (expected). No Capacitor or Capgo package changed, so no Package.resolved check is due.
  • Deploy order: merge after the API is live. No other code merge to main until the v* tag exists.
  • Must stay closed in production: rail_gate.avenia_pix and rail_gate.avenia_pix_qr.
  • Money PRs: most have no non-author human approval yet (step 4). The release page lists them; the approver decides.

PRs in scope (55)

🤖 Generated with Claude Code

innolope-dev and others added 30 commits November 13, 2026 14:42
…g-device

Mark signup events from devices that had an account
…orks

"The verification was not accepted. Contact support." now says our Pix
partner didn't approve it, Pix isn't available for now, the rest of
Peanut is unaffected, and support is there if it's a mistake.
…main-prs-dev

ci(screens): capture and review every PR into main
A timed-out or errored status check now leaves a confirmed present state
alone instead of re-offering Add to Google Wallet for a card still on the
device. Absent and pending still clear it.
…py-dev

copy(pix): say what a Pix verification decline means and what still works
…t refused

While Bridge reviews a customer, the backend lists the corridor in neither
claimable nor unavailable, and the gate reads waiting-on-provider or pending.
The row fell to the not-offered fallback: "We cannot open COP accounts for
you. Support can check why." with a Contact support button.

A wait now opens like any other review: the row says "Under review" and the
tap opens the claim step's existing wait drawer. No new copy, no new row kind.

TASK-24909
resolveBridgeTosClose answered a yes/no. A failed confirm request and an
acceptance the account did not reflect yet both came back as "no", so the
terms step said "Something went wrong... Please try again" or "haven't been
accepted yet". Its retry minted a new Bridge page each time: one user
accepted 30 times, then deleted the account.

It now answers accepted, unconfirmed, not-accepted or failed:
- unconfirmed: the step says the acceptance was sent and the payment
  partner has not confirmed it yet, with only "Got it". No new page.
- failed: the existing confirm error and its retry.
- not-accepted: the existing "haven't been accepted yet".
The progress modal ends unconfirmed and failed on its timed-out state,
as it already did for a thrown confirm.

TASK-24909
An action_required entry the app has no step for becomes a fixable
rejection. With no message of its own, the document request fell back to
"Please upload a clearer photo of your ID to continue." Users under Bridge
review read it as a problem with their ID that they did not have.

The ask's own message still wins. The fallback now says only that more
information is needed. The es-AR overlay entry goes: es-419 has no voseo
verb in the new line.

TASK-24909
The resubmit route answers 404 when it finds no provider request to answer.
The app showed "Could not initiate document resubmission. Please try again."
A retry cannot succeed, so the 404 gets its own code and says there is no
document to upload right now. Other failures keep the retry copy.

TASK-24909
Avenia leaves dev before the next release and is parked. The legal registry
no longer lists it, so a route or history row that names it gets no provider
row, like any code no entry answers for. Its rail code, the report branch that
read its Pix rail as the local account, and its two provider sheet strings go
with it. Comments that used it as the example now name Manteca or no one.
With Avenia parked, the demo's Pix routes name Manteca, so /add-money/pix and
/withdraw/pix still open the Pix screens. The demo API answers as Manteca:
a code with no expiry, payer CPF or flat fee, a key lookup that cannot tell
the owner, and a withdrawal that opens at the submit (transfer-then-submit).
The demo Pix code drops Avenia's payee name for a generic one.

Removed handlers no demo flow or fixture reaches any more: the route terms
acceptance, the deposit rules read (no caller), the deposit-with-id funding
report and the new-key recovery.
…hots

Four states from TASK-24909: every Bridge account waiting on review, the
terms step after an acceptance Bridge has not confirmed, the same step after
a failed confirm, and a provider ask with no step and no message.

TASK-24909
The title already says the terms were sent, so the body keeps the two facts
that stop the loop: not confirmed yet, no need to accept again. 21 words
became 13, and the es-AR overlay entry goes (no voseo verb left).

Every screen that mounts BridgeTosStep can now show this status, so each grows
by 15 to 17 words in screen-wordiness. The baseline records that growth.

TASK-24909
…910)

The Pix states the Screen Library shows now run on Manteca's route, for a
resident outside Brazil: R$1 minimum, no flat fee, a code with no expiry, a
key lookup that cannot tell the owner. The states only Avenia reached go:
its terms step, its hosted identity check, the CPF it parked on, the new key
for a refused payout, the payable-code receipt, the own-key and deny-list
verdicts, the identity-first Home plan, and their page captures (p80, p82,
p83, p84, p87). p87 failed on dev for that reason: its tap opened the Pix
setup's terms, which only Avenia's route asked for.

A resident of Brazil now gets the router's real answer: every Pix route
refused for the residence, or paused with no provider while the Brazil pause
is on. The pause, residence and Brazil accounts fixtures say so.

The e2e Pix spec keeps the Manteca route and the deposit minimum.
…ail verdicts

A hand-built rail with no `resolved` verdict reads as fixable, and the demo
identity reads as unverified, so the review rows rendered "Action needed".
The rails now carry the verdict the API sends (a wait, or the terms step).

TASK-24909
…e catalogue

screen-tests checks that the trusted catalogue matches the screens the
fixture registry declares. Generated with the test's own mapping.

TASK-24909
…24910)

Tests of a screen Manteca's route still reaches now name Manteca: the code
screen, the withdrawal form and its submit, the endings, the setup's verify,
review and first-use steps, and the route-driven BRL row. The submit tests
that ran on the deposit-with-id kind and checked shared logic (the pending
guard, the token and amount contract, a moved quote) now run on
transfer-then-submit.

Tests of steps only Avenia's route reached go, while their code stays for the
follow-up PR (ruling D5): the terms step and its hook, the hosted identity
check, the CPF park, the new key for a refused payout and its receipt, the
payable-code receipt, the own-key verdicts, the identity-first Pix plan on
Home, and the deposit-with-id signing and reload states. The
never-signed reload fixture goes with them.

The Pix pages and the BRL row get the D9 case: a paused route that names no
provider still reads as the residence pause.
…dent fixture (TASK-24910)

The check asked the Avenia deposit fixture for its own routes. A resident of
Brazil now states its own routes too, refused with no provider, so the check
reads that fixture instead.
…tes left (TASK-24910)

Generated from src/dev/screens/catalogue.ts with the expression in
capture-cost-integration.test.mjs: 25 Avenia-only states go, and the deposit
minimum state opens at R$0.50, under Manteca's R$1.
The route inventory needs a repeatable scenario for every app route, and the
new-key page stays until its follow-up removal (ruling D5). An old new-key
link on a Manteca withdrawal that waits for no key shows its current state,
which needs no Avenia data.
…ur own setup

Chip on ui#4092: after a page that closed by itself, a failed confirm offered
"Try again", which fetched a new terms page. The closed page may already hold
a signature, so the retry asked for terms the user had signed: the loop this
PR closes. The retry now re-runs the confirm; only a confirmed missing
signature opens a new page.

A generic pending gate is our own setup in flight, not a provider review.
design.md reserves "Under review" for a provider or a reviewer, so that row
reads "Setting up" (processing). waiting-on-provider and requirements in
review keep "Under review".

TASK-24909
…SK-24910)

The capture answers the API outside the page, so a fixture's balance never
applies there: the screen showed the demo balance, had no shortfall, and the
wait for its field error timed out on every dev run. The QR amount now sits
five cents above the demo balance, which both the capture and a local
fixture show.
…SK-24910)

Copied from peanut-api-ts chore/TASK-24910-park-avenia at e9c2ac3b9. It loses
the Avenia webhook and health routes, the Avenia examples and avenia in the
capabilities provider unions. The stored KYC requirement enums keep AVENIA as
a retired value, so attempts that named it still read. It also carries the
API dev contract changes this copy had not synced yet.
…4910)

pnpm gen:api. The capabilities provider unions no longer name avenia, so a
fixture or test that still did would stop compiling.
…ot (TASK-24910)

The wholesale copy also carried API dev contract changes this snapshot had not
synced (update-user, email change, Rain cards, KYC questions, the saved Pix
key), and api-openapi-sync.test pins the old update-user refusal. Those belong
to their own sync. The snapshot is now the previous one with the API's
snapshot diff of e9c2ac3b9 applied by patch, which applies cleanly: the Avenia
routes, examples and provider union values go, nothing else changes.
…-leading

fix(pix): TASK-24841 the Pix logo leads as itself, and Add money by Pix always starts a new deposit
…ive-check

fix(withdraw): check the crypto network minimum as the amount changes (TASK-23878)
The residence hero swaps waving-hello for cheering on a running clock. A
single subtraction left the carried frame past the shorter loop's end;
this pins the modulo.
The 400 ms picker hold only applied after the wave, so a country picked
just before the wave ended slapped its stickers behind the closing picker
(Chip review). Any country that arrives after the mount now gets the hold;
for the geo prefill it runs out inside the wave and changes nothing.
…gs-24899

feat(setup): country stickers on the Residence step (TASK-24899)
@0xkkonrad
0xkkonrad deployed to content-publish November 16, 2026 11:12 — with GitHub Actions Active
@github-actions

Copy link
Copy Markdown
Contributor

⚠️ This PR changes CI workflows or the backdoor scan itself

The scan in tests.yml runs this PR's own workflow, so a change here could switch it off. This check stays red until the change is merged. Someone other than the author must read these files before anyone runs the branch:

  • .github/workflows/android-release.yml
  • .github/workflows/bot-approval.yml
  • .github/workflows/release-ota.yml
  • .github/workflows/staging-ota.yml
  • .github/workflows/sync-openapi.yml
  • .github/workflows/tests.yml
  • scripts/backdoor-scan.mjs

…pages

feat(TASK-24893): creator invite pages on /go with a three-page test for Nita
@0xkkonrad
0xkkonrad deployed to content-publish November 16, 2026 11:13 — with GitHub Actions Active
@github-actions

Copy link
Copy Markdown
Contributor

⚠️ This PR changes CI workflows or the backdoor scan itself

The scan in tests.yml runs this PR's own workflow, so a change here could switch it off. This check stays red until the change is merged. Someone other than the author must read these files before anyone runs the branch:

  • .github/workflows/android-release.yml
  • .github/workflows/bot-approval.yml
  • .github/workflows/release-ota.yml
  • .github/workflows/staging-ota.yml
  • .github/workflows/sync-openapi.yml
  • .github/workflows/tests.yml
  • scripts/backdoor-scan.mjs

hasTarget: /\btarget=/.test(attributes),
hasPosition: /\bdata-cta-position=/.test(attributes),
hasIcon: inner.includes('<svg'),
text: inner.replace(/<svg.*?<\/svg>/gs, '').replace(/<[^>]+>/g, ''),
@github-actions

Copy link
Copy Markdown
Contributor

⚠️ This PR changes CI workflows or the backdoor scan itself

The scan in tests.yml runs this PR's own workflow, so a change here could switch it off. This check stays red until the change is merged. Someone other than the author must read these files before anyone runs the branch:

  • .github/workflows/android-release.yml
  • .github/workflows/bot-approval.yml
  • .github/workflows/release-ota.yml
  • .github/workflows/staging-ota.yml
  • .github/workflows/sync-openapi.yml
  • .github/workflows/tests.yml
  • scripts/backdoor-scan.mjs

abalinda and others added 5 commits November 16, 2026 15:02
…the deposit screen's reason

Smoke test, release 168: a R$22 Pix paid from an account not under the
user's CPF was sent back by Avenia. The deposit screen said why ("Pay again
from an account under your CPF."); the Activity receipt read "Failed" with no
reason.

The API marks such a deposit with the existing extraData.returnReason (owner
only) and keeps the wire status FAILED. A Pix deposit that is FAILED and
carries returnReason now reads as a returned deposit, as returned bank
deposits do: "Returned" on the badge, the date row and the PDF status,
"Returned to sender" on the Activity row, and the reason line. For
third_party it is the deposit screen's own sentence (one string for both),
for other the general line. Nothing was converted, so the receipt keeps the
"≈ BRL 22" estimate and no rate or fee rows. An entry without returnReason
reads as before, and a lapsed code stays Cancelled.

The PDF status row now prints "Returned" for every returned deposit, as the
screen does; it printed "Refunded" for a returned bank deposit too. Two
Screen Library states show the returned receipt and its Activity row. The
receipt screens' word budget grows by the reused sentence (+8); no screen
shows more words than before.
…receipt too

Chip on ui#4116: the returned state came from returnReason, which the API
sends to the owner only, so the shared receipt and the anonymous PDF of the
same deposit still read "Failed" while the owner read "Returned".

The state is public now and the reason stays private: a FAILED Pix deposit
reads as returned when extraData.pixReturned is true, for every viewer.
The reason line is the deposit screen's CPF sentence only for the owner's
third_party reason, and the general line otherwise. The shared receipt
layout never prints the CPF sentence, even if a reason reached it, because
it speaks to the payer. An entry without the flag reads "Failed" as before,
whatever returnReason says.
…b6cb9e67d

GET /history/:entryId extraData gains pixReturned (true only on a Pix
deposit the provider gave back, to every viewer). Nothing else differs from
b0d06b9af, the snapshot's previous source.
…o, with its mark

Argentines add ARS from Mercado Pago, so "Bank transfer (ARS)" with a green
bank bubble named the wrong method at signup. The arsBank row now reads
"Mercado Pago (ARS)" in every locale and leads with the Mercado Pago mark;
the picker button reads "Mercado Pago" with the mark at the 24 px slot.

This mirrors the Pix (BRL) change for Brazil (TASK-24947): Mercado Pago
joins Pix as a brand mark in CONCEPT_ICONS (a logo, the whole circle, no
bubble), and the setup channel maps to it. The stored value stays arsBank:
the saved plan and the analytics are data.
…dence-manual-review

feat(profile): say when a person is reviewing a residence change (TASK-24822)
@innolope-dev
innolope-dev deployed to content-publish November 16, 2026 14:46 — with GitHub Actions Active
@github-actions

Copy link
Copy Markdown
Contributor

⚠️ This PR changes CI workflows or the backdoor scan itself

The scan in tests.yml runs this PR's own workflow, so a change here could switch it off. This check stays red until the change is merged. Someone other than the author must read these files before anyone runs the branch:

  • .github/workflows/android-release.yml
  • .github/workflows/bot-approval.yml
  • .github/workflows/release-ota.yml
  • .github/workflows/staging-ota.yml
  • .github/workflows/sync-openapi.yml
  • .github/workflows/tests.yml
  • scripts/backdoor-scan.mjs

…68-smoke

fix: release 168 smoke-test fixes (TASK-24947)
@abalinda
abalinda deployed to content-publish November 16, 2026 15:36 — with GitHub Actions Active
@github-actions

Copy link
Copy Markdown
Contributor

⚠️ This PR changes CI workflows or the backdoor scan itself

The scan in tests.yml runs this PR's own workflow, so a change here could switch it off. This check stays red until the change is merged. Someone other than the author must read these files before anyone runs the branch:

  • .github/workflows/android-release.yml
  • .github/workflows/bot-approval.yml
  • .github/workflows/release-ota.yml
  • .github/workflows/staging-ota.yml
  • .github/workflows/sync-openapi.yml
  • .github/workflows/tests.yml
  • scripts/backdoor-scan.mjs

…o-pago-funding-row

fix(TASK-25015): signup names Argentina's ARS funding row Mercado Pago, with its mark
@abalinda
abalinda deployed to content-publish November 16, 2026 15:38 — with GitHub Actions Active
@github-actions

Copy link
Copy Markdown
Contributor

⚠️ This PR changes CI workflows or the backdoor scan itself

The scan in tests.yml runs this PR's own workflow, so a change here could switch it off. This check stays red until the change is merged. Someone other than the author must read these files before anyone runs the branch:

  • .github/workflows/android-release.yml
  • .github/workflows/bot-approval.yml
  • .github/workflows/release-ota.yml
  • .github/workflows/staging-ota.yml
  • .github/workflows/sync-openapi.yml
  • .github/workflows/tests.yml
  • scripts/backdoor-scan.mjs

@github-actions

Copy link
Copy Markdown
Contributor

⚠️ This PR changes CI workflows or the backdoor scan itself

The scan in tests.yml runs this PR's own workflow, so a change here could switch it off. This check stays red until the change is merged. Someone other than the author must read these files before anyone runs the branch:

  • .github/workflows/android-release.yml
  • .github/workflows/bot-approval.yml
  • .github/workflows/release-ota.yml
  • .github/workflows/staging-ota.yml
  • .github/workflows/sync-openapi.yml
  • .github/workflows/tests.yml
  • scripts/backdoor-scan.mjs

This branch was successfully deployed

3 active (1 outdated) deployments
Preview — 78c00906 Deployed Nov 16, 2026 by vercel[bot]
content-publish — 78c00906 Deployed Nov 16, 2026 by abalinda via approve-and-merge #8201
screen-library-deploy — 664fc118 Deployed Nov 16, 2026 by innolope-dev via publish / deploy / deploy #3669
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

6 participants