[Programcode]

Brings the iOS app up to date with main and restyles it as DebateOwl. Base is feat/native-ios. Don't merge yet. The open questions for Adam are at the end.

Catching up with main

origin/main (222 commits, through #251) is merged into the branch with a normal merge (baec2758). These 22 files conflicted.

• Billing (_shared/entitlements.ts and its test, use-subscription.ts, settings-store.ts, SettingsPage.tsx). Main's resolveBillingState stays the one place that decides a plan. Main added payment grace, extended access and subscription_ended; the branch's effective_entitlement() SQL function predates them. Rather than let the SQL win, resolveBillingState now also takes App Store rows. App Store Pro ranks with Stripe Pro, and the later end shows. App Store Basic (20 a day) comes after legacy free and the payment grace, before a trial. Revoked, upgraded and expired rows grant nothing, and Apple's billing grace extends a row. The AI functions, the website and the app all read user_settings, subscriptions and app_store_subscriptions and resolve them the same way. A database without the App Store table answers PGRST205, which counts as no App Store rows, so web and functions keep working before that migration ships. I ported the pgTAP scenarios for the App Store rules to billing-state.test.ts.
• client.ts, streaming.ts, sparring-store.ts, SparringPage.tsx take main's versions, since main split the AI client into per-feature files. The branch's extractions come back on top in later commits (see "Moved to core").
• ai/index.ts keeps main's re-indented block plus the branch's sparring_scored push. notifyScored became optional so main's new judge tests run unchanged.
• Edge tests take main's versions. The branch's tests that mocked /rpc/effective_entitlement now mock the table reads like main's; the testing/entitlement.ts helper is gone.
• Locales went through a key-level three-way merge. No key changed on both sides.
• package.json test:edge takes main's flags plus the branch's App Store and Expo env names.
• package-lock.json is main's lockfile with the mobile workspace's resolved versions laid over it, so the Expo SDK 57 set is unchanged. I restored two packages the Expo CLI got from Tailwind 3's hoisted tree (arg, micromatch).
• Docs (architecture.md, operations.md, supabase/README.md) keep main's text plus the branch's install-command, App Store and push sections, reworded for the TypeScript plan resolver.

Other fixes the merge needed:

• 40 keys main deleted would have shown as raw keys in the app. They now point at main's current keys, such as the one progress line per tool and the empty-state lines, or live under mobile.* in en/es/hu where the website no longer has them.
• temperature is gone from the app's tool settings, as on the website.
• The app's web preview couldn't read the plan because supabase.web.ts lacked the new helper. 3ee7cf9e fixes it.

Moved to core

• packages/core/src/supabase/billing.ts is the plan reader web and app share. It sits on contract/billing-state.ts, a re-export of the Deno resolver in the existing contract/ pattern.
• ai/feature-configs.ts holds the tool defaults, model normalisation and the canonical check, taken from the settings store.
• A Codex child moved the AI parsers, merged in f79ace8a. They are the Generate, Brainstorm and Factoids transforms, discovered-source and assessment parsing, retention and the autosave row, SSE framing, source-url.ts (URL matching), Sparring round moves and timers, and the quota session key. The web imports them with no behaviour change. Tests moved with the code (packages/core/src/**/*.test.ts now runs in vitest).

Case structure v2 is not in this PR. #244 is still a draft, conflicts with main and has failing CI, so there is no v2 on main to move into core or render. Its schema and case-view.ts already live in packages/core, so the app can pick them up once #244 lands. The case screenshots show the current (v1) format.

App catch-up with main

• The You tab's plan card shows main's billing states, dates and wording (SubscriptionStatus.tsx). Usage is the server's counter (ai_usage_today).
• Crossfire sends the round id as quotaSession, so the server charges a round once (Charge Crossfire and Sparring per round, refund runs that produce nothing (batch T5) #201). A test covers it. Sparring already was, through its sessionId.
• Research autosave runs one batch at a time and refreshes a page the debate already has, matched by normalised URL.
• A Sparring turn's clock starts when the student starts typing or presses play, as on the website.
• Ask falls back to the whole-answer request when a failed stream sent only whitespace (Fall back to the structured Ask when a failed stream sent only whitespace #187).

DebateOwl restyle

• Name DebateOwl, slug debateowl, links to https://debateowl.com.
• Colours come from the website's tokens in light and dark (plume, ink, gold, mint, coral). Old Tailwind values map to the website's owl remap of the same family and step.
• Fredoka for titles, the motion and button labels, Nunito for text (expo-font via @expo-google-fonts).
• Buttons are the website's tactile ones, a blue face on a ledge that sinks when pressed. Sentence-case labels come from main's catalogs, and a long label wraps to a second line instead of being cut off. Pinned action bars (case, attack, scorecard) sit on a strip of the page colour with a fade above it. Cards and fields have 2pt edges, and fields turn blue when focused. Switches use the brand fill.
• Success is mint with a check mark (sign-in "code sent"). The You tab's plan badges are the website's outline badges: Pro gold, trial and Basic blue, extended access and legacy free mint with a check mark, ended plans coral.
• Ollie appears in the app icon (the website's icon-1024.png), the splash (pointing), the sign-in mark, empty states (reading, surprised, sleepy, point), tool progress, the Ask coach avatar and the Sparring opponent. The PNGs come from the website's SVGs via scripts/render-owls.ts.
• On iPad the Library list takes the full width until a source opens, as the website's Library does, and Ask's option chips wrap. On iPhone the sideways chip rows fade at the edge.
• Titles and choices that Hungarian or Spanish cut off (picker values, collapsed items, recent work, flowchart edge labels) wrap to two lines.
• No streak exists in the app or on the website, so the feather streak icon appears nowhere.
• No feature was added or removed.

Review polish (second round)

From the review of this PR:

1. A pinned button label no longer gets an ellipsis. Labels wrap to a second centred line, and on the case screen "Attack this case" gets the wider button. It fits in en, hu and es at 390pt.
2. Pinned action bars (case, attack, Brainstorm, Factoids, scorecard) sit on a strip of the page colour with a fade above it. The page and any notice clear the bar's measured height.
3. Plan badges match the website's outline badges (15% fill, 30% edge, darker text). Pro is gold, as on the website; the earlier "mint plan badges" in this description was wrong. The website's Basic badge now uses the trial blue instead of an unmapped teal.
4. On iPad the Library list takes the full width (in an 880pt column) until a source opens. Then it keeps 40% of the width beside the source, and titles and dates wrap.
5. On iPad Ask's option chips wrap. On iPhone the chip rows (options and citations) fade at the edge only while they overflow, into the surface they sit on.

The full screenshot pass found and fixed more of the same kind: picker values, collapsed document items, recent work, source rows and flowchart edge labels that Hungarian or Spanish cut off now take two lines, and switches use the brand fill with a white thumb.

Tests

All results are local runs on Adam's VPS (Node 22.23.2), not CI.

• npm run verify at the final head 1ef3b2e3 passed. That covers docs, lint (eslint, knip, repo rules), typecheck (web, agentic, core and the app), vitest (121 files, 790 tests, including apps/mobile and packages/core), collaboration (30), Deno edge tests (543), build and site checks, the service worker check, and the Playwright desktop smoke set (47). An earlier npm ci plus verify at 6fb2664e also passed.
• The database gate (./.superset/db-gate.sh --db-only) on the merged migration order passed 33 pgTAP files and 655 tests, including app_store_entitlements_test.sql and push_tokens_test.sql, plus the quota, Opus-default, admin-drift and workspace concurrency scripts.
• The two Codex review fixes after 6fb2664e added use-subscription tests (11 pass), and the Crossfire quota test fails when the round id is dropped.
• The app's npm run -w @debatelab/mobile typecheck and export:web pass, and the screenshot script exits 0 at every size (it fails on any page error or unexpected console error).
• Codex review round 1 at 6fb2664e raised two findings, both fixed. A failed App Store read on the website now leaves the plan unknown instead of "Expired" (3bb421db, with tests), and the owl loader stays still under Reduce Motion (a9f5a0ad). Round 2 at a9f5a0ad raised one finding, which I dismissed. When the plan query fails, Settings falls back to an "Expired" card. Main already does this for every failed read, such as a failed settings read. The harm the indeterminate-read rule records (fix: finish UI cleanup and OpenRouter model catalog #47), a persisted downgrade of the model choice, can't happen: the settings store decides Pro through readProModelAccess, which returns unknown for a partial read. Showing an "unknown" plan in Settings needs a new card state and strings, so it's listed under gaps as a follow-up.
• GitHub Actions did not run. Every job on this PR ended within 3 seconds with no steps, which the repo guide describes as the account being out of Actions minutes, not a code failure. CI is not green and has not been tried.
• Not run were WebKit, the staging browser suite, live AI (verify-debatelab --live) and anything on a real iOS device or simulator.

Screenshots

These come from the web export through apps/mobile/scripts/screenshots.ts (staging, the seeded Pro account, nothing written), at 390, 820 and 1180 points.

• before: ~/.t3/scratch/ios-reskin-shots/before/, light, English.
• after: ~/.t3/scratch/ios-reskin-shots/after/, light and dark, English, Hungarian and Spanish. 648 shots from 1ef3b2e3, all checked by eye. clipped.json there lists every text a shot cuts off; only two remain, both by design: the motion as Ask's one-line header subtitle on iPhone, and the opponent's speech in Sparring, collapsed to four lines until tapped.

The script gained an 11-inch iPad Air size (820×1180 and 1180×820), Spanish, SHOTS_DIR, current sentence-case labels, a filter for the expected 404 from a project without the App Store table, and the clipped-text report: each shot lists the texts React Native Web draws with an overflowing ellipsis or line clamp.

Known gaps and flags

• Bundle id still com.adam.debatelab, and the URL scheme is still debatelab. A new id means a new App Store Connect app, so that's Adam's call.
• Slug is now debateowl. On eas init day this decides the Expo project name (@adamgaborjani/debateowl).
• Migrations. The branch's four 20261014* migrations are older than main's 2026-10-16 to 2026-10-24 ones, so a project that already has main's needs supabase db push --include-all. The database gate passes on the merged order. effective_entitlement() is defined but no longer called. A follow-up migration can drop it.
• Web Settings and the App Store. The website doesn't know an App Store subscriber is billed by Apple. An App Store Pro account would see "Set up billing" and Basic would see "Upgrade to Pro". No real account can hit this yet, because in-app purchase isn't live and the products are placeholders. hasAppStoreSubscription is plumbed for when it is.
• Unknown plan in web Settings. When the plan read fails, useSubscription falls back to an expired snapshot, so Settings says "Expired" instead of "couldn't load". This predates the branch and covers every failed read; a follow-up can add an unknown state to the Settings card.
• Live streaming of tool answers (Stream every AI answer live, token by token #149) is not in the app. Generate, Attack, Brainstorm and Factoids still wait for the whole answer, which the server still supports. Drain live AI streams until the body closes #204 and T18: keep AI streams alive, recover dropped results, fix Auto Pro charge, refunds, Sparring judge errors #224's run recovery apply to those live streams only.
• Not proven on a device. Native tabs, large titles, sheets and the button press were not seen on iOS. EAS Simulator needs an EAS project (eas init), which is a build-day step, so it wasn't used.
• npx expo install is blocked by the repo's allowScripts policy, so the two font packages were installed from the root with npm install -w.
• The Android icons were regenerated with the owl, though this is an iOS app.

Open questions for Adam

• Keep com.adam.debatelab for the store, or move to a DebateOwl bundle id before the first TestFlight build?
• When should eas init run (it creates @adamgaborjani/debateowl)? It also unlocks EAS Simulator for a native check of this restyle.

🤖 Generated with Claude Code

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

Run configuration

• Source mode: Auto
• Repository: Programcode/DebateLab
• Branch: native-ios/catch-up-debateowl
• File: .coderabbit.yaml
• Commit: 1ef3b2e
• Layers: Repository YAML (base), Organization UI (inherited)
• Review profile: CHILL
• Plan: Free
• Run ID: 2c1e4057-ac80-4986-b3fb-5b7fc21b64ce

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Note

🎁 Summarized by CodeRabbit Free

Your organization is on the Free plan. CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please upgrade your subscription to CodeRabbit Essentials by visiting https://app.coderabbit.ai/settings/billing.

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

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota.