Port upstream 0.64.0: Usage & Spend provider grouping, consolidated chart selector, clear selected day

[Finesssee]

Found by the 0.60.4-0.69.0 port gap audit (gap G9, audit-0.64.0).

Summary

Settings -> Usage & Spend now groups model history by provider, folds the daily and hourly charts behind one compact selector, and adds a control that clears the selected day.

• The flat "Models" list is replaced by collapsible provider groups. Each group header shows the provider name and its spend/tokens (with ~ when some models are unpriced). Each group shows six models, with Show all / Show less, and a "Partial model breakdown" note when local history is incomplete.
• The standalone hourly heatmap is now one panel with a Day / Hour selector. The Day view is a new daily spend bar chart built from the existing contract.daily points. The Hour view is the existing heatmap. With only one view available the selector is hidden.
• Clicking a day bar selects it. The chip with the date has a clear (x) button and clicking the selected bar again also clears it. Changing the 7d/30d/all-time range clears the selection.

Upstream reference

• Release note (v0.64.0, Improved): "Usage & Spend: group sources and models by provider, consolidate detail and daily/hourly chart selectors, and add a clear selected-day control while preserving incomplete-history warnings, local-history labels, calendar boundaries, privacy, and exports (feat(spend): group usage details by provider steipete/CodexBar#3353)."
• Commit 212c8d63a (feat(spend): group usage details by provider steipete/CodexBar#3353). Tag-pinned v0.64.0 files read: Sources/CodexBar/SpendDashboardProviderBreakdown.swift (spendDashboardProviderBreakdowns ordering, spendProviderModelDisplayLimit = 6, per-provider model rows, partial model warning), SpendDashboardSummary.swift (selected-day chip and clear button), PreferencesSpendDashboardPane.swift (SpendDashboardTrendSection daily/hourly picker), SpendDashboardModel.swift (incompleteModelProviders).

Ported / Deferred

Ported: provider grouping and ordering rule, six-model collapse, partial-model-history label, single Day/Hour selector, selected-day chip and clear control.

Deferred, because the Windows data model does not carry it and inventing fields is out of bounds:

• Per-provider model rows for providers other than the contract provider. The shell builds one model-level contract (contract.providerId, Codex, merged with OpenCodex imports), so only that provider gets a group. Other providers stay in the existing provider table. The grouping helper takes the contract, so more groups appear when the backend supplies them.
• Upstream's "Sources / Accounts" children per provider group: there is no per-native-source cost row on Windows.
• Selected day only scopes the Spend card (from contract.daily). Models, projects and hourly activity stay on the full range because the contract has no per-day breakdown of them. The panel says so when a day is selected.
• Upstream's Sessions/Projects/Providers detail picker, daily ledger, and incompleteModelProviders per-provider set (Windows uses historyCoverageEstablished for the single contract).

Unchanged: incomplete-history labels, calendar/reporting day, privacy redaction, share card and JSON export, provider table.

Validation

• pnpm --dir apps/desktop-tauri exec vitest run (full): 69 files, 413 tests passed, including 11 new tests (usageSpendBreakdown.test.ts, UsageSpendBreakdownPanels.test.tsx).
• pnpm --dir apps/desktop-tauri run lint: no findings in the touched files (pre-existing warnings elsewhere).
• pnpm --dir apps/desktop-tauri run build: passed.
• cargo +1.98.0 fmt --all: clean.
• cargo +1.98.0 test -p codexbar locale: 16 passed (new locale keys are in the enum, en-US.ftl, and keys.ts).
• cargo +1.98.0 clippy --workspace --all-targets -- -D warnings: passed (workspace, both crates)

Affected areas

• Settings UI (Usage & Spend tab)
• Locale catalog (6 new en-US keys; other locales fall back)
• Rust provider/backend logic (none)

UI proof

Pending: coordinator will capture CUA proof on a fresh build (CODEXBAR_PROOF_MODE=settings:usageSpend with a seeded spend file).

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[Finesssee]

Adversarial validation (lane-B review)

Head validated: c25f083a ("Port upstream 0.64.0: group Usage & Spend models by provider, consolidate chart selector, clear selected day"). Spec: audit-0.64.0.md G9.

Verdict: no blocking defects. The grouping/selector/selected-day behavior matches the spec:

• usageSpendBreakdown.ts implements upstream ordering (highest known cost, then tokens, unknown last, then name with numeric/base compare) and the 6-model display limit (spendProviderModelDisplayLimit).
• Partial-cost markers (~ via costIsPartial when priceCoverage.unpriced > 0) and partial-history markers (hasPartialModelHistory from historyCoverageEstablished) match the partial-cost/partial-history semantics.
• UsageSpendTab.tsx consolidates the daily/hourly selectors into UsageSpendTrendPanel (155 lines, own component) with a clear-selected-day control; the selected day flows into SpendContractOverview.
• Locale keys added for the new controls (en-US, 6 keys; check-locale 885 OK).
• Provider grouping note in the spec ("one group per provider") is correctly reduced: the desktop shell builds model-level history for one provider contract, so only that provider renders a group; the rest stay in the provider table — consistent with the data model, documented in buildSpendProviderGroups's doc comment.

Validation re-run at head (pinned 1.98.0, E-cores): cargo +1.98.0 fmt --all -- --check clean; cargo +1.98.0 clippy --workspace --all-targets -- -D warnings clean; cargo +1.98.0 test -p codexbar --lib 2160 passed / 0 failed (1 ignored); frontend: pnpm test 68 files / 413 tests passed (11 in the two new suites); pnpm run check-locale OK (885 keys); pnpm run build OK.

UI proof: waived per user 2026-10-02 directive (fast-track); component tests + build cover the new surfaces.

Integrator: merge with #610/#609 coordination noted in the spec (usage_spend.rs is #610's file — no conflict at this head since the change is frontend-only).

[Finesssee]

UI proof (browser-use)

Combined build of integrate/v0.70.0-ports at ddd85594 (all 40 port PRs and the follow-ups). Debug desktop build in proof mode, isolated config/data dirs, driven by browser-use over the WebView2 DevTools protocol with DOM events only (no mouse or keyboard input, no focus changes). All data is synthetic: local mock servers for L1, a seeded usage snapshot plus synthetic local logs for L2.

Scenario	Check	Result
L2 seeded snapshot + synthetic local logs	#705 Usage & Spend: Day/Hour trend toggle present with labelled day bars	PASS
L2 seeded snapshot + synthetic local logs	#705 Hour view renders the hourly activity chart	PASS
L2 seeded snapshot + synthetic local logs	#705 provider groups use the release period caption ('Local model history for: ...'), no 'days of local history'	PASS
L2 seeded snapshot + synthetic local logs	#705 section order: trend before provider groups before projects	PASS

Every surface also passed the privacy check (no email-like text, account e-mail nodes or profile paths in the DOM) and theme auto rendered dark on the tray flyout, float bar and settings windows.

Validation at ddd85594: cargo fmt --all --check, cargo clippy --workspace --all-targets -- -D warnings, cargo test -p codexbar (3567 passed, 0 failed) and cargo test -p codexbar-desktop-tauri (606 passed, 0 failed). Screenshots were captured for each scenario and kept with the local proof kit.

[Finesssee]

Shipped in v0.70.0: this PR's head is included in main via #735 (merge commit 9d0a37a). Closing as integrated.