feat: ladder v1 plan mechanics (flagged off) by Git-on-my-level · Pull Request #21488 · BasedHardware/omi

Git-on-my-level · 2026-10-14T12:06:01Z

Implements and integrates ladder-v1 plan mechanics across backend, mobile, web and macOS behind default-off flags. Legacy prices and installed-client entitlements stay unchanged. Keep draft; do not merge or mark ready.

Integrated status

Six Stripe lookup-key bindings (ladder_v1_{plus,pro,architect}_{month,year}), config-bound price IDs, collision/legacy guards, read-only price attestation, renewal/reconciliation recognition and interval-safe reactivation. No Stripe writes.
Shared monthly budget admission, 80% nudge and 100% hold/choice. Mobile consumes backend at_cap strings, reads/writes the capability-gated monthly choice before enabling on-device processing, and opens upgrades with trigger=usage_budget. Device audio remains in durable custody; a choice never grants cloud entitlement.
Merged lane/sub-ladder-clients at d70df24 and lane/ondevice-overflow-ux / feat(app): on-device overflow engines, processing and summaries (flagged off) #21466 through 55ce061 with merge commits. Includes native engines, scheduled runner, local summaries, provenance and phone-only WAL policy. Merged main through a42840d (including shipped feat: subscription clarity for everyone (pre-ladder) #21474), then overflow 55ce061. feat(app): on-device overflow engines, processing and summaries (flagged off) #21466 is fully contained; the lead will close it. Main’s shipped freemium fallback remains authoritative; Parakeet, runner, scheduling and provenance remain behind mobile-on-device-overflow. Phone-only overflow takes precedence over a still-valid cloud batch authorization (real packet-owner regression); no rebase or force push.
Fixed mobile budget_80 trigger to contract usage_budget and omitted absent triggers instead of sending an invalid empty literal; added backend pro_budget80 golden and byte-identical mobile fixture. Pro sees Architect only at 100% monthly usage.
Backend-owned monthly→yearly confirmation lines and CTA used before billing on mobile/web/macOS. Mobile avoids the duplicate confirmation. Generated OpenAPI, Swift, Dart and TypeScript projections regenerated.
Versioned recording rate card v2: live and batch both remain 130 units/worn hour today, with a 98% background / 2% live advertised blend. On-device transcription is zero units; cloud text summaries still consume LLM units. Capacity-derived copy bands have tests at provisional/measured rates and the 720/1,440-hour rounding boundaries; headline and meter stay consistent, and percent rounding never offers Architect below the real cap; v1 event pricing/digest remain unchanged.
Pro proactive/screen entry points and observed daily MCP-cap entries on existing mobile/macOS feature surfaces. Both contextual flags and an explicit backend gate/cap are required; legacy/default-off paths remain inert. Daily caps remain 100/500/5,000/5,000 with atomic reservations and reset metadata.
Configurable approved decoder floors and separate default-off pro alias retirement. Existing client behavior is preserved until the audited retirement step.
On-device copy uses the one shared usage budget. Settings show the storefront usage meter (percent, meter_line, reset_label) and "Get more usage" in the same card. The settings row and the queue banner's Get more usage button are shown only when the contextual plans sheet can open, and both call openContextualPlans with trigger=usage_budget. openOnDeviceUpgrade has no legacy Usage page fallback. The storefront catalog now has every app locale. Flutter, web, and macOS send the UI locale as locale. The backend string is the meter, the get-more label, and the at-cap title. hours_left is still set only from 80% to below 100% when the estimate is positive. meter_line still includes the hours estimate only in that band. meter_line still includes the hours estimate only from 80% to below 100%; below that, and at the cap, it is percent and reset (meter.line_short). An unresolved engine stays loading. A failed or timed-out lookup is unavailable with retry and never opens Parakeet; Parakeet setup opens only after a positive non-OS-managed engine. SpeechAnalyzer turns on with no setup screen; a speech-asset install is a settings row ("Getting speech ready · 40%"), and the speech-model row stays hidden while the engine is OS-managed. Parakeet settings show the real 443 MB size. The save-usage toggle is "Always transcribe on this phone" / "Saves your usage. No speaker labels." Inspected renders: /Volumes/Ephemeral/scratch/ondevice-ux-renders-2026-10-15/ (17-ios26-first-run-no-setup.png removed; it duplicated the settings screen).

Flags (all default OFF)

Server: STOREFRONT_LADDER_V1_ENABLED, LADDER_V1_LANE_METERING_ENABLED, LADDER_V1_BUDGET_ENFORCEMENT_ENABLED, LADDER_V1_PRO_GATES_ENABLED, LADDER_V1_MCP_DAILY_CAPS_ENABLED, LADDER_V1_WIRE_IDENTITY_ENABLED, LADDER_V1_RETIRE_PRO_ALIAS_ENABLED, existing HELD_CONTENT_ENABLED (plus client capability).

Clients: web-plan-storefront, web-contextual-upsells, mobile-plan-storefront, mobile-contextual-upsells, desktop-plan-storefront, desktop-contextual-upsells, mobile-on-device-overflow (debug compile opt-in only for local verification).

Remaining cutover checklist

David approves Stripe writes → create six prices on existing products → approved read-only lookup-key/product/amount receipt → bind six LADDER_V1_*_PRICE_ID env IDs. Preserve all legacy prices and env vars.
Do not flip the storefront flag until the batch lane is live and measured fully loaded blended economics support the bands: Pro 720 worn hours needs about ≤ $0.052/hour; Plus 240 hours needs about ≤ $0.058/hour. Batch STT vendor estimates alone are insufficient. Today's rate deliberately does not claim these bands.
Accept ledger producer coverage, rate-card calibration, monthly reset and two-week measurement baseline. The ledger remains a partial lower bound; admission is not a concurrent prepaid reservation system.
Validate integrated device holding, retry, on-device fallback, reset and actual product behavior. Builds, hermetic tests and offscreen renders are not production/provider/device acceptance.
Approve decoder floors, audit installed clients and stored rows; retire the old pro alias last.
Ramp shadow first, then identity; storefront web → mobile → desktop. Server flags are global. Enable hold/budget/Pro/daily enforcement only after acceptance, and review conversion/retention/held/reset outcomes at each step.
David ruling (October 14 evening): chat pauses at 100% until reset or upgrade on every ladder tier, including Free. No cheap model or capped fallback chat. Shared backend chat_at_cap copy supplies the pause, reset date and upgrade path (Pro → Architect). Mobile, macOS and both web composers consume it and open plans with usage_budget. Capture, held content and on-device transcription/summaries continue. Legacy prices and flag-off admission retain their prior behavior. uses_shared_budget also prevents macOS from applying the older question counter below the budget cap; legacy quota responses omit both new fields.

Rollback

Disable storefront/client flags to stop new sales. Independently disable budget/Pro/MCP/identity/overflow flags to restore prior admission. Keep price ID bindings for renewal/webhook recognition; restore alias acceptance if needed. Never rewrite subscriptions, usage history or captured content. Reset recovery retains its separate flag/runbook.

Final integration validation and conflicts

Previous integrated head 13e43149735b8d5987734af8c71065e71ca52f58 had all 14 full CI workflows green. Head 43bfb8ebe48bbf7f2609073b48156cd68512bdb3 keeps Hangul meter segments together. Previous head 600a3574978280b721f4eabf8804243a0bbae6ea formats the at-cap date from the reset timestamp instead of parsing the localized reset label, translates the plans-sheet chrome, and wraps CJK meter segments at the middle dot. It localizes the storefront catalog for every app locale and keeps the shared usage meter: plans-sheet-only get-more, unknown-engine loading, one usage card, and the speech-asset progress row. Speaker-label router, contract test, marker keys and review telemetry match current main; the app-client OpenAPI diff against main is the storefront fields, including more_label, with no extra speaker constraints. Main had replaced the home processing card with the pending row; the on-device queue stays beside it. Draft/open; not merged. No rebase or force push. feat(app): on-device overflow engines, processing and summaries (flagged off) #21466 is fully contained; the lead will close it.
Apple/freemium conflict: configured-language readiness uses one selected-engine path; SFSpeech is probed as SFSpeech. Main's shipped kill-switch/readiness/retry behavior is authoritative when overflow is off. Enhanced engines remain behind the overflow flag.
WAL conflicts and automatic merge cleanup: deduplicated custody fields and constructor arguments; custody is latched per recording, not derived from the current preference at save time. Historical imports remain processable; explicit local-result custody is retained. The local-conversation regression remains in the test suite.
Whisper/polling integration retains normalized, bounded plugin calls, transient-failure retry and consumed-audio clocks, including silence. Dropped polling audio advances its source clock.
Localization conflicts had identical values: retained the integrated ARBs and regenerated Dart. Independent flag rows were combined and the registry regenerated; ladder-only rows sit at the end of the active list. Pipeline docs retain both ladder admission and current main guidance.
Home-page conflict keeps the on-device queue banner beside main's pending row. Phone-only WALs stay out of that row. The processing-card test left with the card main deleted; the pending-row test covers on-device-only recordings.
Later main capture-forwarding conflict retains onboarding audio taps/notifications and the phone-only cloud-send guard. The as-captured audio main merge was clean; later archive-test conflict takes main’s equivalent is_file() assertion. Current main messaging/Firestore provisioning changes are retained; OpenAPI and Dart/Swift/TypeScript artifacts were regenerated and match the tree.
Backend focused suites pass via backend/test.sh: quota/budget/catalog/copy/metering, sync-v2 (182 tests), storage ownership, as-captured ingest and archive contracts. Legacy prices, flag-off behavior and monthly reset remain covered.
Dart focused WAL/polling/provider/runner/capture/chat suites pass, including two overflow-enabled configured-language/selected-engine readiness tests. Combined onboarding integration: 218 tests pass; latest archive-schema refresh: 179 tests pass. Analyzer ratchet passes without baseline changes; 49 locales and 4030 keys are consistent.
Web bun run check: typecheck, 6 Bun smoke tests and 721 Vitest tests pass. macOS swift build and 63 focused quota/storefront/decoder/render tests pass. iOS simulator and Android Dev debug builds pass with overflow enabled.
make preflight and normal bounded push gates pass; the expanded integration/workflow scope runs 88 repository checks. actionlint passes. No failed, pending or cancelled full workflow remains at verification. Analyzer and both overflow-enabled mobile debug builds pass.
CI harness corrections preserve every check and limit: iOS shares the existing pinned Swift-tool binary cache (flow coverage dropped from 983s to ~2s); the native revision-retirement fixture explicitly holds queued work until the policy transition (forced fast-writer negative control plus 30 successful runs); sync cyclic-mock cleanup covers every case, retaining the 0.30s CPU guard; archive acceptance counts payload files, not .bin-named prefix directories, while retaining the exact byte comparison.
Inherited engine recovery renders were inspected; integrated storefront/chat evidence remains applicable. CodeQL scratch containment, linear email redaction and private-value/import guards remain covered without suppressions.
Rate-card v1/event digests and legacy Stripe rows remain unchanged. Current lane rates preserve 130 units/hour; measured blended capacity remains a cutover gate. No Stripe writes, flag flips or deployment.

Runbook: backend/docs/ladder-v1-cutover.md. Refreshed and inspected renders: /Volumes/Ephemeral/scratch/sub-ladder-renders-2026-10-14/integrated/ (real component/widget render harnesses, not live authenticated product acceptance).

UI evidence:

Chat ruling renders (actual production pause components with backend goldens; all four tiers inspected):

INV-DATA-1: canonical account/data-plane authority is preserved; the headless engine reuses existing Firebase/profile initialization. Local debug fixtures are not shipped routing changes.
INV-CAP-1: captured audio remains in durable phone/WAL custody while cloud admission is held; capture retention and sync-fence regressions pass.
INV-UI-2: composer pause and contextual plans use existing Omi controls and shared confirmation; modal dismissal consumes Escape.
INV-CHAT-1: composer presentation uses the existing shared quota/provider; no new timeline, journal writer or continuity handler.
INV-UI-3: on-device provenance never invents speaker separation or named/count labels.
Failure-Class: FC-stt-allowance-not-enforced-at-dispatch
Earlier CI exposed eager Firestore imports and unnecessary executor work while ladder flags were off. Database resolution now occurs only after the relevant guard; test_flag_off_cold_import_never_resolves_database and the existing async-boundary suites pin that contract.

Line-Count-Exception: app/lib/pages/conversation_detail/conversation_detail_provider.dart | 1518 -> 1527 | Narrow flagged native processing and provenance integration; engine implementation remains in dedicated modules
Line-Count-Exception: app/lib/pages/settings/transcription_settings_page.dart | 1572 -> 1575 | Narrow entry to the isolated flagged on-device settings
Line-Count-Exception: app/lib/providers/conversation_provider.dart | 2121 -> 2133 | Narrow local conversation provenance and account-owned WAL integration
Line-Count-Exception: app/lib/services/capture/capture_controller.dart | 4475 -> 4495 | Narrow retained-capture overflow hooks; native runner stays separate
Line-Count-Exception: app/lib/services/wals/local_wal_sync.dart | 2690 -> 2714 | Flagged phone-only WAL exclusion from automatic and manual cloud drains
Line-Count-Exception: backend/routers/developer.py | 2752 -> 2753 | Optional on-device STT provenance reaches the existing transcript ingest seam
Line-Count-Exception: backend/routers/payment.py | 1746 -> 1752 | Narrow flagged ladder price checkout binding
Line-Count-Exception: backend/routers/sync.py | 2714 -> 2726 | Shared flagged budget admission before file consumption
Line-Count-Exception: backend/utils/subscription.py | 1950 -> 2015 | Shared flagged ladder policy at existing admission/reporting boundaries

Line-Count-Exception: backend/utils/sync/pipeline.py | 3212 -> 3215 | Validate owned scratch before the existing recursive cleanup seam; helper and regression tests remain separate

Line-Count-Exception: backend/routers/users.py | 2572 -> 2579 | Additive flagged chat pause projection at the existing quota endpoint; copy and admission live in dedicated modules

Cap enforcement follow-up (#6 / #7)

Merged main at 1239634f1bd (#21866) with a merge commit. Ladder live budgets refresh on each 60-second usage flush, including with legacy dispatch dark; funded dispatch uses the ladder budget. Sync and GPU workers reauthorize at execution/microbatch boundaries and preserve denied audio through the existing hold/defer recovery. Ladder-only activation bypasses legacy monthly-seconds dispatch.

Enforced ladder totals include residual transcription seconds at the provisional unknown-provider speech rate (no silence discount), plus residual/unpriced gateway attempts at the standard-chat estimate. Coverage histograms expose residual seconds/attempts per evaluation; no historical writes. Normal live detection overshoot is about 60 seconds of speech per session plus persistence/read latency; concurrent in-flight jobs are not reserved. Flags remain OFF; PR remains draft.

Validation: BACKEND_PYTEST_WORKERS=1 BACKEND_UNIT_TEST_FILE_LIST=<list> nice -n 10 bash backend/test.sh: targeted 205 passed; touched 578 passed; recovery 116 passed. Flutter analyzer ratchet passed (94 existing diagnostics); 49 focused storefront/checkout Flutter tests and 8 web billing tests passed. Full repository PR preflight: 89 selected checks passed. Final refresh regressions: 43 passed. CI caught duplicate coverage histogram registration during isolated auth imports; collectors now reuse the process registry (the existing observability pattern). The failing API-key suite plus ladder/STT regressions pass locally; exact-head CI is recorded in the PR checks tab. No prod/GCP or Stripe access.

🤖 Generated with Claude Code

When live hours run out (or by choice), Omi keeps recording and can transcribe on the phone. This adds the mobile UI for that, behind the default-off PostHog flag mobile-on-device-overflow: - at-cap choice sheet (transcribe on this phone / upgrade), offered from the existing freemium prompt only when the flag is on and live hours are used up - model setup (step 1) and schedule picker (step 2: while charging, every few hours, as you talk) - conversations-list banner for audio waiting on the phone, waiting chips on recording rows, finish-setup and unsupported-phone variants - provenance chips on conversation rows and a provenance section on the summary (why, Redo with Omi Cloud, keep everything on this phone) - On-Device Transcription settings page and its entry row - settings model persisted via SharedPreferencesUtil, a pure scheduling policy, an OnDeviceModel abstraction (Android Whisper, iOS unavailable until Parakeet lands), and placeholder benchmark numbers in one file With the flag off every new surface renders nothing. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Settings > Plan renders GET /v2/payments/plans behind web-plan-storefront (default off): backend header, current line, legacy note, usage meter, interval toggle, tier cards, blocked lines and CTAs, verbatim. Purchases use the payload's price ids through the existing checkout-session and upgrade-subscription rails, gated by the same checkout decision as mobile. web-contextual-upsells adds the budget meter, the 80% nudge and the exhausted-budget entry (trigger=usage_budget, Architect for Pro at 100%). Unknown fields are ignored; unknown actions, intervals, usage kinds and plan ids render inert. The registry scanner now sees TypeScript PostHog flag constants, and the four web/desktop flags are registered. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

At 100% the payload has no nudge_line, so the banner drew a bare button. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

… payload Settings > Plan & Usage renders GET /v2/payments/plans behind desktop-plan-storefront (default off) in place of the local plan card and grid: backend header, current line, legacy note, usage meter, interval chips, tier cards, blocked lines and CTAs, verbatim. Purchases use the payload's price ids through the existing startCheckout rail (upgrade for paid subscribers, checkout session for Free and resubscribe), gated by the same checkout decision as mobile and web. desktop-contextual-upsells adds the 80% nudge and the exhausted-budget sheet (trigger=usage_budget). Tests run every backend golden through the decoder and checkout boundary, and render the review states offscreen with ImageRenderer. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Add gated native engines, phone-only WAL custody, charging schedulers, and hash-bound summary provenance. Co-Authored-By: Codex <noreply@openai.com>

Preserve both main localization updates and flagged on-device copy through the l10n tool. Co-Authored-By: Codex <noreply@openai.com>

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Regenerate the complete merged locale set and record all 22 audited UI states with SF Pro fonts. Co-Authored-By: Codex <noreply@openai.com>

Use SafeDefaults for the background charging policy and guard the dynamically loaded native bridge. Co-Authored-By: Codex <noreply@openai.com>

Co-Authored-By: Codex <noreply@openai.com>

mintlify · 2026-10-14T12:06:05Z

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated
omi	🟢 Ready	View Preview	Oct 16, 2026, 11:06 PM

💡 Tip: Enable Automations to automatically generate PRs for you.

…ne/sub-ladder-cutover

…lane/sub-ladder-cutover # Conflicts: # backend/docs/listen_pusher_pipeline.mdx # config/feature-flags.yaml

# Conflicts: # backend/docs/listen_pusher_pipeline.mdx # config/feature-flags.yaml # scripts/test_check_feature_flag_registry.py

Clients can open the plans sheet from the usage meter without inventing unit copy, and the at-cap title matches on every surface. Co-authored-by: Cursor <cursoragent@cursor.com>

Settings follow the backend meter, OS-managed speech skips setup unless assets are installing, and the save-usage toggle no longer upsells Pro. Co-authored-by: Cursor <cursoragent@cursor.com>

The usage-meter push updated the branch without starting the pull_request workflows. Co-authored-by: Cursor <cursoragent@cursor.com>

…row. Main replaced the processing card with the pending row. Phone-only recordings stay on the on-device banner and out of the cloud row. Co-authored-by: Cursor <cursoragent@cursor.com>

That file's diarization assertions currently fail on main, and a local addition made pre-push rerun them. Co-authored-by: Cursor <cursoragent@cursor.com>

The exporter and the checked-in document drifted in the merge, so the contract check rejected the branch. Co-authored-by: Cursor <cursoragent@cursor.com>

Main already rejects unsafe speaker and person ids, but publishing those path constraints narrows the released contract. The handlers still reject them before any document lookup. Co-authored-by: Cursor <cursoragent@cursor.com>

The conversations page always mounts the queue banner. Loading the storefront while the flag is off constructs the API client and reads Env, which the outage page test never initializes. Co-authored-by: Cursor <cursoragent@cursor.com>

…row. Main restored the backend unit shards and moved speaker-id checks out of the released app-client contract. The conversations page still mounts the on-device banner only as a no-op while the flag is off. Co-authored-by: Cursor <cursoragent@cursor.com>

The get-more row only opens the plans sheet, an unknown engine stays loading until detection finishes, and non-English devices use the app strings for the new usage labels. Co-authored-by: Cursor <cursoragent@cursor.com>

The released app-client contract stays free of extra speaker constraints, and the marker keys and review logs follow the bounds already on main. Co-authored-by: Cursor <cursoragent@cursor.com>

Daily Recaps stay above Today. The on-device queue sits with the pending row and stays empty while the overflow flag is off. Co-authored-by: Cursor <cursoragent@cursor.com>

Below 80%, and at 100%, the line is percent and reset, so it never says about 0 hours. Co-authored-by: Cursor <cursoragent@cursor.com>

The queue banner opens the plans sheet only when it is available, a failed engine lookup stays unavailable, and non-English devices localize the meter from hours_left. Co-authored-by: Cursor <cursoragent@cursor.com>

The plans sheet, reset date, billing-period words, and usage meter now follow the locale the client sends, with English kept as the per-key fallback. Co-authored-by: Cursor <cursoragent@cursor.com>

The route inventory and contract export import the app, and the catalog now formats dates with Babel. Co-authored-by: Cursor <cursoragent@cursor.com>

…sed. Co-authored-by: Cursor <cursoragent@cursor.com>

Co-authored-by: Cursor <cursoragent@cursor.com>

# Conflicts: # app/lib/l10n/app_ar.arb # app/lib/l10n/app_be.arb # app/lib/l10n/app_bg.arb # app/lib/l10n/app_bn.arb # app/lib/l10n/app_bs.arb # app/lib/l10n/app_ca.arb # app/lib/l10n/app_cs.arb # app/lib/l10n/app_da.arb # app/lib/l10n/app_de.arb # app/lib/l10n/app_el.arb # app/lib/l10n/app_en.arb # app/lib/l10n/app_es.arb # app/lib/l10n/app_et.arb # app/lib/l10n/app_fa.arb # app/lib/l10n/app_fi.arb # app/lib/l10n/app_fr.arb # app/lib/l10n/app_he.arb # app/lib/l10n/app_hi.arb # app/lib/l10n/app_hr.arb # app/lib/l10n/app_hu.arb # app/lib/l10n/app_id.arb # app/lib/l10n/app_it.arb # app/lib/l10n/app_ja.arb # app/lib/l10n/app_kn.arb # app/lib/l10n/app_ko.arb # app/lib/l10n/app_localizations.dart # app/lib/l10n/app_localizations_ar.dart # app/lib/l10n/app_localizations_be.dart # app/lib/l10n/app_localizations_bg.dart # app/lib/l10n/app_localizations_bn.dart # app/lib/l10n/app_localizations_bs.dart # app/lib/l10n/app_localizations_ca.dart # app/lib/l10n/app_localizations_cs.dart # app/lib/l10n/app_localizations_da.dart # app/lib/l10n/app_localizations_de.dart # app/lib/l10n/app_localizations_el.dart # app/lib/l10n/app_localizations_en.dart # app/lib/l10n/app_localizations_es.dart # app/lib/l10n/app_localizations_et.dart # app/lib/l10n/app_localizations_fa.dart # app/lib/l10n/app_localizations_fi.dart # app/lib/l10n/app_localizations_fr.dart # app/lib/l10n/app_localizations_he.dart # app/lib/l10n/app_localizations_hi.dart # app/lib/l10n/app_localizations_hr.dart # app/lib/l10n/app_localizations_hu.dart # app/lib/l10n/app_localizations_id.dart # app/lib/l10n/app_localizations_it.dart # app/lib/l10n/app_localizations_ja.dart # app/lib/l10n/app_localizations_kn.dart # app/lib/l10n/app_localizations_ko.dart # app/lib/l10n/app_localizations_lt.dart # app/lib/l10n/app_localizations_lv.dart # app/lib/l10n/app_localizations_mk.dart # app/lib/l10n/app_localizations_mr.dart # app/lib/l10n/app_localizations_ms.dart # app/lib/l10n/app_localizations_nl.dart # app/lib/l10n/app_localizations_no.dart # app/lib/l10n/app_localizations_pl.dart # app/lib/l10n/app_localizations_pt.dart # app/lib/l10n/app_localizations_ro.dart # app/lib/l10n/app_localizations_ru.dart # app/lib/l10n/app_localizations_sk.dart # app/lib/l10n/app_localizations_sl.dart # app/lib/l10n/app_localizations_sr.dart # app/lib/l10n/app_localizations_sv.dart # app/lib/l10n/app_localizations_ta.dart # app/lib/l10n/app_localizations_te.dart # app/lib/l10n/app_localizations_th.dart # app/lib/l10n/app_localizations_tl.dart # app/lib/l10n/app_localizations_tr.dart # app/lib/l10n/app_localizations_uk.dart # app/lib/l10n/app_localizations_ur.dart # app/lib/l10n/app_localizations_vi.dart # app/lib/l10n/app_localizations_zh.dart # app/lib/l10n/app_lt.arb # app/lib/l10n/app_lv.arb # app/lib/l10n/app_mk.arb # app/lib/l10n/app_mr.arb # app/lib/l10n/app_ms.arb # app/lib/l10n/app_nl.arb # app/lib/l10n/app_no.arb # app/lib/l10n/app_pl.arb # app/lib/l10n/app_pt.arb # app/lib/l10n/app_ro.arb # app/lib/l10n/app_ru.arb # app/lib/l10n/app_sk.arb # app/lib/l10n/app_sl.arb # app/lib/l10n/app_sr.arb # app/lib/l10n/app_sv.arb # app/lib/l10n/app_ta.arb # app/lib/l10n/app_te.arb # app/lib/l10n/app_th.arb # app/lib/l10n/app_tl.arb # app/lib/l10n/app_tr.arb # app/lib/l10n/app_uk.arb # app/lib/l10n/app_ur.arb # app/lib/l10n/app_vi.arb # app/lib/l10n/app_zh.arb # app/lib/pages/settings/widgets/plans_sheet.dart # backend/docs/feature-flag-registry.md # backend/routers/payment.py # config/feature-flags.yaml # desktop/macos/Desktop/Sources/MainWindow/Pages/Settings/Components/SettingsContentView+BillingHelpers.swift # desktop/macos/Desktop/Sources/MainWindow/Pages/Settings/Sections/SettingsContentView+AccountBilling.swift # web/app/src/components/settings/__tests__/SettingsSubscriptionClarity.test.tsx

Git-on-my-level and others added 18 commits October 14, 2026 13:21

docs(app): UI evidence for on-device transcription overflow screens

df16fb3

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

chore(app): changelog fragment for flagged-off on-device overflow UI

3880dec

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

fix(web): give the exhausted-budget banner the backend's header line

a05d482

At 100% the payload has no nudge_line, so the banner drew a bare button. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

docs(evidence): add ladder-v1 plans renders for web and macOS

bb9caae

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

docs(desktop): add changelog fragment for the plans section

bf030ad

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

test(desktop): cover the storefront plan files from the plan-usage flow

278f979

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

feat(app): run on-device overflow transcription and summaries

3987baa

Add gated native engines, phone-only WAL custody, charging schedulers, and hash-bound summary provenance. Co-Authored-By: Codex <noreply@openai.com>

chore: merge origin/main into on-device overflow lane

ca58395

Preserve both main localization updates and flagged on-device copy through the l10n tool. Co-Authored-By: Codex <noreply@openai.com>

test(desktop): unwrap the dark appearance without a force unwrap

d70df24

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

feat(app): refresh overflow audit renders and localization

38e1e29

Regenerate the complete merged locale set and record all 22 audited UI states with SF Pro fonts. Co-Authored-By: Codex <noreply@openai.com>

refactor(ios): validate overflow scheduler defaults and bridge access

ea3685e

Use SafeDefaults for the background charging policy and guard the dynamically loaded native bridge. Co-Authored-By: Codex <noreply@openai.com>

chore(api): regenerate on-device transcription provenance contract

19cb374

Co-Authored-By: Codex <noreply@openai.com>

feat(app): align on-device upgrade copy with product ruling

3c0b17a

Co-Authored-By: Codex <noreply@openai.com>

chore(app): satisfy overflow analyzer and API contracts

daf2bef

Co-Authored-By: Codex <noreply@openai.com>

feat: ladder v1 plan mechanics (flagged off)

8edf942

mintlify Bot deployed to staging - docs October 14, 2026 12:06

View deployment

Git-on-my-level added 9 commits October 14, 2026 19:22

Merge remote-tracking branch 'origin/lane/sub-ladder-clients' into la…

00d7871

…ne/sub-ladder-cutover

Merge remote-tracking branch 'origin/lane/ondevice-overflow-ux' into …

0bc75a7

…lane/sub-ladder-cutover # Conflicts: # backend/docs/listen_pusher_pipeline.mdx # config/feature-flags.yaml

feat: integrate ladder clients and efficiency contracts

55d2f0b

Merge remote-tracking branch 'origin/main' into lane/sub-ladder-cutover

b8f5408

# Conflicts: # backend/docs/listen_pusher_pipeline.mdx # config/feature-flags.yaml # scripts/test_check_feature_flag_registry.py

test: verify integrated ladder contracts and renders

06cd90a

test: capture ladder review evidence and isolate runtime setup

5a084d0

feat: use shared shell for ladder annual confirmation

8e5acea

test: render native ladder confirmation through AppKit hosting

20f9978

fix(app): omit absent storefront trigger from requests

212c314

mintlify Bot deployed to staging - docs October 14, 2026 13:26

View deployment

Merge home processing card while retaining separate on-device queue

13e4314

mintlify Bot deployed to staging - docs October 15, 2026 03:39

View deployment

Git-on-my-level mentioned this pull request Oct 15, 2026

feat(app): on-device overflow engines, processing and summaries (flagged off) #21466

Closed

Git-on-my-level and others added 2 commits October 16, 2026 00:13

Expose a backend-owned usage CTA and one shared monthly-limit headline.

de80c8a

Clients can open the plans sheet from the usage meter without inventing unit copy, and the at-cap title matches on every surface. Co-authored-by: Cursor <cursoragent@cursor.com>

Show the shared usage meter on device and retire live-hours copy.

15089f6

Settings follow the backend meter, OS-managed speech skips setup unless assets are installing, and the save-usage toggle no longer upsells Pro. Co-authored-by: Cursor <cursoragent@cursor.com>

mintlify Bot deployed to staging - docs October 15, 2026 18:02

View deployment

Git-on-my-level and others added 5 commits October 16, 2026 01:16

ci: retrigger pull request workflows

6c950b5

The usage-meter push updated the branch without starting the pull_request workflows. Co-authored-by: Cursor <cursoragent@cursor.com>

Merge origin/main and keep the on-device queue beside Home's pending …

ebe8a7c

…row. Main replaced the processing card with the pending row. Phone-only recordings stay on the on-device banner and out of the cloud row. Co-authored-by: Cursor <cursoragent@cursor.com>

Keep the ladder speech shadow test out of the listen runtime file.

912a6fa

That file's diarization assertions currently fail on main, and a local addition made pre-push rerun them. Co-authored-by: Cursor <cursoragent@cursor.com>

Refresh the app-client OpenAPI document after merging main.

1ce5e35

The exporter and the checked-in document drifted in the merge, so the contract check rejected the branch. Co-authored-by: Cursor <cursoragent@cursor.com>

mintlify Bot deployed to staging - docs October 15, 2026 19:30

View deployment

Git-on-my-level and others added 5 commits October 16, 2026 02:57

Match main's speaker-label storage and telemetry.

e71622f

The released app-client contract stays free of extra speaker constraints, and the marker keys and review logs follow the bounds already on main. Co-authored-by: Cursor <cursoragent@cursor.com>

Merge origin/main into the ladder cutover branch.

f2fd19f

Daily Recaps stay above Today. The on-device queue sits with the pending row and stays empty while the overflow flag is off. Co-authored-by: Cursor <cursoragent@cursor.com>

mintlify Bot deployed to staging - docs October 16, 2026 05:05

View deployment

Git-on-my-level and others added 2 commits October 16, 2026 12:58

Show remaining hours on the usage meter only from 80% until the cap.

735e400

Below 80%, and at 100%, the line is percent and reset, so it never says about 0 hours. Co-authored-by: Cursor <cursoragent@cursor.com>

Keep on-device upgrade and engine detection from guessing a fallback.

3141eab

The queue banner opens the plans sheet only when it is available, a failed engine lookup stays unavailable, and non-English devices localize the meter from hours_left. Co-authored-by: Cursor <cursoragent@cursor.com>

mintlify Bot deployed to staging - docs October 16, 2026 07:35

View deployment

Git-on-my-level and others added 4 commits October 16, 2026 15:46

Localize the storefront catalog for every app locale.

cfc789e

The plans sheet, reset date, billing-period words, and usage meter now follow the locale the client sends, with English kept as the per-key fallback. Co-authored-by: Cursor <cursoragent@cursor.com>

Put Babel on the OpenAPI runner so storefront import succeeds.

b6393c3

The route inventory and contract export import the app, and the catalog now formats dates with Babel. Co-authored-by: Cursor <cursoragent@cursor.com>

Format the at-cap date directly so localized reset labels are not par…

600a357

…sed. Co-authored-by: Cursor <cursoragent@cursor.com>

Keep Korean meter segments from wrapping inside a word.

43bfb8e

Co-authored-by: Cursor <cursoragent@cursor.com>

Git-on-my-level mentioned this pull request Oct 16, 2026

Enforce the transcription cap where cloud STT is dispatched (flag off) #21866

Merged

Git-on-my-level added 2 commits October 17, 2026 05:40

fix: bound ladder budget refresh and conservatively charge coverage gaps

7f261e8

mintlify Bot deployed to staging - docs October 16, 2026 23:06

View deployment

fix: reuse ladder coverage collectors across isolated imports

bd37341

Navigation Menu

Repository navigation

feat: ladder v1 plan mechanics (flagged off) - #21488

feat: ladder v1 plan mechanics (flagged off)#21488
Git-on-my-level wants to merge 84 commits into
mainfrom
lane/sub-ladder-cutover

Git-on-my-level commented Oct 14, 2026 •

edited

Loading

Uh oh!

mintlify Bot commented Oct 14, 2026 •

edited

Loading

Uh oh!

This branch was successfully deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

Git-on-my-level commented Oct 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Integrated status

Flags (all default OFF)

Remaining cutover checklist

Rollback

Final integration validation and conflicts

Cap enforcement follow-up (#6 / #7)

Uh oh!

mintlify Bot commented Oct 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This branch was successfully deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Git-on-my-level commented Oct 14, 2026 •

edited

Loading

mintlify Bot commented Oct 14, 2026 •

edited

Loading