steipete/CodexBar v0.57.0

CodexBar 0.57.0

on GitHub

Highlights

• Claude cost breakdowns: inspect daily usage and top models with the new opt-in CLI cost --breakdown output (#3244).
• Faster, more accurate cost history: reuse Claude reports across launches, resume interrupted Codex scans, and recover usage missed by older parsers (#3284, #3411, #3504).
• Clearer menus and fresher dashboards: refresh spend charts when returning to the app, show money or points for balance-only providers, and honor segmented claude-swap account menus (#3107, #3494, #3498).
• Hardened updates: adopt Sparkle 2.9.6's archive-handling and package-signature protections.

Added

• CLI: add opt-in Claude cost --breakdown daily and top-model details, with consistent calendar or recorded periods and explicit partial-attribution labels (#3244). Thanks @Yuxin-Qiao!

Performance

• Claude local costs: reuse compatible cost reports across launches instead of decoding unchanged transcript caches; refresh them when history, pricing, or report semantics change (#3284). Thanks @eggyrooch-blip!
• Codex local costs: reduce repeated pricing work across daily, project, and session reports without changing token accounting or tariffs (#3476). Thanks @brzvsk!

Security

• Updates: adopt Sparkle 2.9.6 installer hardening, including archive-moving and package-signature validation fixes.

Fixed

• Usage & Spend: refresh expired charts on pane return or app activation, keep cached data visible during loading, and refresh across midnight (#3107). Thanks @Yuxin-Qiao!
• Codex local costs: resume unfinished scans after a refresh reaches its time limit, without restarting completed file work or discarding compatible history (related to #3411). Thanks @kesslerio!
• Codex local costs: accept valid JSON whitespace in usage events, retain next-day appended usage, and reparse older files once without deleting compatible stores (#3504).
• Codex local costs: include previously unpriced usage in GPT-5.6 Luna estimates, including cached reports, without discarding compatible history or scan progress (#3503, #3502). Thanks @BUKOWSKIREAL!
• Claude local usage: discard stale rows when a transcript is replaced, including across restarts, while retaining incremental parsing for genuine appends.
• Usage parsing: reject out-of-range counts in MiMo, Pi/OMP, OpenCodex, and Bedrock without crashing; preserve valid fields and refresh affected OpenCodex caches (#3486).
• Local costs: avoid overflow crashes in OpenCodex and combined reports, leaving unrepresentable totals unavailable while retaining valid neighboring token classes (#3501).
• Accounts: retain matching cached usage and widget data through transient multi-account refresh failures without refreshing measurement timestamps or reusing changed credentials.
• Claude: honor segmented account menus for claude-swap, retain unavailable-account diagnostics, and show stable slot numbers when personal information is hidden (#3498, #3382). Thanks @thatlev!
• Claude: preserve claude-swap's measurement timestamps so cached usage does not appear newly refreshed; retain the fallback for missing or malformed optional timestamps (#3485, extracted from #3452). Thanks @QuantIntellect!
• Claude: recover expired default-profile usage from fresh CLI credentials when existing consent permits, while preserving explicit-file precedence and custom-profile isolation (related to #3390).
• Menu bar: show money or points for balance-only providers in the default layout and editor preview, preserve real quota percentages, and avoid duplicate reset text (#3492, #3494). Thanks @zkforge!
• Widgets: remove redundant padding from Usage, Switcher, History, and Metric widgets so WidgetKit controls their content margins (extracted from #3137). Thanks @iamenahs!
• Overview: keep highlighted provider cards readable on macOS 15 while retaining fast GPU selection and native submenu interactions (#3173).
• Copilot: resolve Enterprise identities on the configured host, keep accounts on different hosts distinct, and avoid public GitHub budget requests for Enterprise accounts (#3341). Thanks @Fletcher-Alderton!
• Antigravity: show each CLI quota bucket once, preserve unavailable usage and reset context, and keep display filtering out of raw JSON (#3489). Thanks @urda!
• Antigravity local usage: retain valid history around bookkeeping steps without UUIDs, while continuing to reject ambiguous IDs and uncertain dates (#3462). Thanks @urda!
• Kiro: use the CLI profile's supported region for overage enrichment, reject invalid profile ARNs before sending credentials, and retain CLI fallback (#3359). Thanks @zucram!
• MiniMax: preserve cached usage and normal retries after DNS, connection, and translated offline failures.
• z.ai: preserve valid quota when optional model analytics overflow or exceed display bounds, while retaining supported Unicode labels.
• AWS Bedrock: disclose monitoring charges in both authentication modes, link Cost Explorer pricing, and clarify refresh controls and the informational budget (#3496, related to #3387). Thanks @kyen99!
• Settings: detect rapid external config replacements and edits that restore earlier app-written contents, without treating successful app writes as external changes.
• Local usage: honor the app's Low Power Mode interval during automatic Codex catch-up, while retaining manual acceleration and system thermal pauses.
• CLI login: stop cancelled Codex and Kiro logins and lingering child processes while preserving timeout output and device-flow progress.
• Subprocesses: handle very large finite timeouts without overflowing or crashing.