fix: correct usage history, account routing, and quotas

[steipete]

Correct repeated cost-history scans and projections, account credential selection, quota reset handling, and CLI cost windows. The batch also adds GitKraken API credit tracking and compact Omarchy provider logos, with provider metadata and documentation kept consistent.

The registry contains 78 providers. README, provider documentation, all 23 website locales, social HTML, website cards, and generated provider sources agree. New changelog bullets and contributor thanks are under 0.64.2 Unreleased, preserving the already-published 0.64.1 release notes.

Bifrost is deferred. This snapshot omitted its provider implementation after a source/limit aggregation concern. The provider PR has since advanced and needs a fresh review before adoption. The independently useful plugin-host support for unknown named-window usage and numeric detail progress is retained. The Bifrost provider PR remains open.

Lane verdicts

• refactor(cli): share common command option groups #3852 — dedup-2: LAND. Share Commander options, OneConsole personal-plan parsing, environment/endpoint helpers, and credit reconciliation while preserving the existing CLI/provider contracts.
• fix(cost): correct Linux Priority pricing and 30-day JSON totals #3850 — cli-cost-fixes: LAND. Adopt fix(codex): read Priority traces on Linux #3820's Linux Priority/Fast evidence support and Fix last30Days fields in extended cost JSON histories #3824's calendar-correct 30-day JSON totals. Thanks @A-l-an and @tillstriegel.
• fix(codex): confirm rolling unused weekly resets #3851 — quota-bugs: LAND Codex early weekly reset remains stale at 0–1% after #3177; 2% publishes normally #3248. Confirm rolling unused Codex weekly resets. bug(z.ai) Reset time shows 10 hrs window instead of 5 hours (again in 0.49.2) #2871 remains open for a complete payload; Claude enterprise extra usage row missing after upgrading to 0.47.0 #2681 still requires a source/privacy decision. Thanks @kcharlan.
• fix(litellm): fall back to self-scoped spend reports #3853 — litellm-spend: LAND / resolve LiteLLM: support spend reports when management routes are disabled #3834. Fall back to bounded, self-scoped spend reports when management routes are disabled, without inventing budgets or identity. Thanks @xiaoxianma.
• fix(claude): include claude-swap homes in cost scans #3849 — cost-accounts: LAND [Claude cost] Usage & Spend shows one claude-swap account because the cost scan reads a single home #2954. Include claude-swap homes in combined local cost totals with shared/copied-history deduplication. OpenRouter management keys do not surface token and model activity #3289 was already resolved on main. Thanks @Newarr.
• fix(opencodego): support API-key token accounts #3848 — accounts-dashboard: LAND OpenCode Go: support multiple API-key accounts in token accounts #3841. Route OpenCode Go per-account API keys while preserving cookie fallback. The separate Add opt-in multi-account dashboard snapshots #3832 proposal is rejected for default-redaction, adapter-failure-contract, and scope problems; none of its code is adopted.
• fix(claude): honor read-only account adapters #3846 — small-prs-c: LAND Support read-only Claude account adapters #3658. Respect read-only Claude account adapters using strict boolean parsing. Reject feat: icloud device removal #3350's insufficiently proven destructive CloudKit change and keep No way to remove a stale duplicate device record in iCloud Sync #3234 open. Thanks @benedictfrancis.
• fix(antigravity): honor explicit CLI overrides #3847 — antigravity-scoped: LAND fix(antigravity): honor CLI overrides and explain excluded reports #3689. Make an explicit CLI-path override authoritative. Reject feat(antigravity): fetch usage per saved account via scoped agy #3780's credential-staging implementation; Antigravity "Limits not available" bug. #3662 and Antigravity: menu bar ignores a signed-in agy and shows “Launch Antigravity” after the desktop app is uninstalled #3146 remain open. Thanks @hhh2210.
• fix(antigravity): retain cadence-less quota history #3856 — antigravity-history: LAND the fix(antigravity): preserve quota history and estimate local token costs #3619 quota-history subset. Retain account-scoped observations without assuming a reset cadence, including replenishment within an hour. Remaining pricing/disclosure differences are deferred. Thanks @hhh2210.
• perf(cost): reuse quota projections and unchanged Codex scans #3859 — perf-projection: LAND perf(cost): cache the quota-week projection off the main thread #3831 / resolve First provider tab switch stalls 100–200 ms: quotaWeekSummaries runs on the main thread with no cache #3827 and Codex scan load re-decodes the entire cost cache on every refresh, even when the store is unchanged #3840. Reuse immutable snapshot projections and unchanged decoded Codex scan baselines while keeping live reset inputs current. Thanks @giuseppebisemi and @CodingCanuck.
• perf(keychain): memoize concurrent code-signature checks #3857 — perf-codesign: LAND Background Chromium cookie gate re-runs a full app-bundle code-signature validation on every refresh: ~11.6% of a core at idle, 190-217% CPU bursts every 60s #3837 and the reviewed cache rewrite. Coalesce code-signature checks, invalidate after bundle changes, and preserve no-UI Keychain semantics. The original fix(keychain): keep completed signature validations fresh #3838 has advanced beyond the inspected revision and remains open for review of that delta. Thanks @jeffloo886.
• fix(claude): replay CLI frames before parsing usage #3854 — claude-pty: LAND the panel-dismissal fix. Reused sessions dismiss modal panels before subsequent commands. This train includes only panel dismissal. The diagnosis, regression fixture, and contributor credit from fix(claude): replay CLI PTY output onto a screen before parsing (#3746) #3822 are preserved in the smaller replay rewrite on still-open fix(claude): replay CLI frames before parsing usage #3854 (35bbc87f8d73); that follow-up is not included here. Fable doesn't show reliably. It's really unpredictable. No way to know if it'll show or not. #3746 and the remaining identity work in Claude CLI source: /status identity probe never runs because the reused session still has the /usage panel open (email/plan badge stay empty) #3821 await its integration and validation. Thanks @sczhui.
• feat(gitkraken): add API usage provider plugin #3855 — gitkraken: LAND Add GitKraken AI usage provider with API and CLI support #3842. Add first-party API personal weekly credits and optional organization-pool details; CLI fallback remains outside scope. Thanks @RyanTheTechMan.
• feat(omarchy): show provider logos in the compact bar #3845 — omarchy: LAND feat(omarchy): mark each provider in the bar with its own logo #3795. Add theme-tinted logos while preserving compact text, two-entry limits, stale markers, missing-logo fallback, and older backends. feat(omarchy): offer bar detail, per-model caps and provider count as preferences #3794 still needs the default-width/scope decision. Thanks @jsonMartin. The lane's earlier macOS CI failure was unclassified; this integration is being validated independently.

The already-landed Linux work on main also resolves #3799 and #3839.

Claude replay remains separate. This PR includes only panel dismissal from #3854. The accepted replay rewrite at 35bbc87f8d73, based on #3822, is outside this snapshot; both PRs remain open, along with the unresolved replay/identity portions of #3746 and #3821.

Integration and validation

Final candidate: e8f82f6ada9d0bdc252b8ef2090584138154746b, rebased onto current main at f28ddcaf3483. Its Git tree (ba0ee9131eaff7c12f7eb0635e35ac07827d143c) is byte-identical to the fully reviewed and tested integration candidate; the rebase changed only commit history.

• Complete aggregate review: independent review of every code change against current main is scoped-clean through P2.
• Focused integration tests: 792 tests across 74 suites passed, covering provider architecture, Antigravity, account routing, and fetch-pipeline behavior.
• make check: passed with zero violations in 2,570 Swift files, including generated-source, documentation, locale, installer, and packaging checks.
• Full integrated make test: all 1,334 selections in 120 groups passed on the first run, with zero retries or timeouts, on the exact reviewed tree. The preceding c426db54c313 tree also passed its complete local suite and hosted CI.
• Current-head CI: run 35763456613 is running on the rebased head. The superseded run was cancelled by the history rewrite; final landing waits for the new head.
• Freshness and credit: all 13 closing component heads match the picked commits. The later Claude replay work and the advanced original fix(keychain): keep completed signature validations fresh #3838 remain open. Human commit identities are preserved; the regression supplied by @CodingCanuck in Codex scan load re-decodes the entire cost cache on every refresh, even when the store is unchanged #3840 is explicitly credited.

The main-branch reconciliation changed only the shared changelog conflict; published 0.64.1 and earlier release notes are byte-identical. All tests use synthetic fixtures and repository test isolation with Keychain access suppressed. The two synthetic chart images are PR attachments rather than runtime assets.

An earlier integration CI failure came from a dashboard fixture expecting the old 90-day aggregate in the CLI's corrected 30-day field. The fixture now asserts the 30-day subtotal and separately preserves the full-history total; production behavior remains calendar-correct.

OpenCode Go account follow-up

Commit c426db54c313 fixes an unresolved account-settings regression: API-key account changes were switching the saved cookie source to Manual, which could disable browser imports after the final API account was removed. The provider now uses its existing credential classifier before changing cookie mode. API accounts preserve Automatic or an explicit Manual preference; Cookie accounts retain Manual behavior.

Synthetic lifecycle tests failed with 24 assertions against the previous code and passed after the repair. The broader OpenCode Go/account run passed 227 tests, make check passed with zero violations, and independent review of this follow-up was scoped-clean through P2. Full make test also passed on this head: all 1,333 selections in 120 groups succeeded on the first pass, with zero retries or timeouts. This follow-up review does not claim to review the entire aggregate.

The late Claude replay commit remains outside this snapshot, so #3854 and its original contributor PR #3822 are referenced rather than automatically closed.

Synthetic Antigravity chart proof

Before:

After:

These existing lane captures were inspected and contain synthetic chart data only. They prove the offscreen native chart, not a live-account menu workflow.

Closes #3845
Closes #3846
Closes #3847
Closes #3848
Closes #3849
Closes #3850
Closes #3851
Closes #3852
Closes #3853
Refs #3854
Closes #3855
Closes #3856
Closes #3857
Closes #3859

Fixes #3248
Fixes #3834
Fixes #3827
Closes #3840
Fixes #3837
Closes #3842
Closes #3799
Closes #3839
Closes #3795
Closes #3820
Closes #3824
Closes #3831
Refs #3838
Closes #3658
Closes #3689
Closes #3619
Fixes #2954
Fixes #3841
Closes #3289
Closes #3832
Closes #3350
Closes #3780
Refs #3822

Refs #2871
Refs #2681
Refs #3234
Refs #3662
Refs #3146
Refs #3746
Refs #3821
Refs #3794

[clawsweeper]

🦞👀
ClawSweeper picked this up.

Pull request received. I will update this pull request when review starts.

ClawSweeper review blocked

Automated review did not run, so no review verdict was produced.

Reason: The input-safety check rejected material in this revision. No detected value, path, or scanner output is reproduced here.

ClawSweeper will not retry this unchanged revision.

Next step: If this is a genuine credential, remove and rotate it. If it is an intentional test fixture, a maintainer must review and qualify it.

View the workflow run.