[Zorato]

Summary

Add pace indicators for the rolling, weekly, and monthly usage windows.

• Show +Npp when usage is below the elapsed share.
• Show -Npp when usage is above the elapsed share.
• Round the absolute deviation before selecting the status. A rounded deviation of five points or less shows on pace.
• Use exact five-hour and seven-day windows for rolling and weekly usage.
• Derive the monthly start one UTC calendar month before resetsAt, with the day clamped to the previous month when necessary.
• Label the monthly row 1mo instead of 30d.
• Count all calendar time, including weekends.
• Omit the pace label when the reset timestamp is missing or invalid.

Align labels, bars, percentages, pace indicators, and reset countdowns with fixed columns. Use a ten-cell bar and preserve its track color through an inline span.

This branch is rebased onto main after #5. It preserves setup-owned state, polling, and cleanup. Renderer tests invoke plugin cleanup before disposing views and detect leaked intervals.

This PR adds no configuration options and does not change credential resolution.

Verification

• npm test: 43 tests pass across five files, including all refresh and lifecycle tests from fix(tui): keep usage state and polling in plugin setup #5.
• npm run typecheck: passes for production sources.
• Pace tests cover 28-, 29-, 30-, and 31-day months, leap years, year boundaries, end-of-month clamping, and UTC time preservation.
• Rounding tests cover symmetric deviations of 5.4 and 5.5 points.
• All 18 pace tests also pass with TZ=America/New_York.
• OpenTUI renderer tests cover column alignment, track color, live pace updates, and cleanup.
• Mutation checks detect fixed-thirty-day monthly calculations, inconsistent status rounding, and missing renderer cleanup.

Tests use synthetic responses, not live credentials or API requests.

[Zorato]

This is how it looks like on my OpenCode setup:

For myself I have excluded weekends (Saturday & Sunday). Let me know if you want to add config for that or exclude them by default.

[PentSec]

Thanks, the pace math and the renderer tests are solid, and the alignment reads well in the screenshot.

Two things before this can merge:

1. tui.tsx conflicts with fix(tui): keep usage state and polling in plugin setup #5. fix(tui): keep usage state and polling in plugin setup #5 moves the state and polling into setup; this branch rewrites the same file from main, so resolving the conflict by taking this version brings back the remount reset that fix(tui): keep usage state and polling in plugin setup #5 fixes. This needs a rebase on main once fix(tui): keep usage state and polling in plugin setup #5 lands, plus cleanup?.() in the test teardown, otherwise each test leaves a refresher running.

2. The monthly window is not 30 days. Real payload: monthly.resetsAt is 2026-11-15T18:12:24Z, the subscription anniversary, so the window is a calendar month of 28 to 31 days. Using 30 days shifts the expected share by up to about 3 points, which is enough to cross your 5 point band and flip on pace. Subtracting one calendar month from resetsAt for the monthly row would be exact. Weekly (Monday 00:00 UTC) and rolling (5h) are exact as written.

Minor: a delta of 5.4 renders as -5pp, which per your own table means "on pace". Pick the bucket with the rounded value, or ceil the magnitude.

On weekends: calendar time as the default is fine with me. If we add an option later, it goes behind config, not in this PR.

[Zorato]

Thanks for the review. Addressed in b5fb1d7:

• Rebased onto main after fix(tui): keep usage state and polling in plugin setup #5 and preserved setup-owned state and polling.
• Added plugin cleanup to renderer test teardown, plus a check for leaked intervals.
• Changed monthly pace to subtract one UTC calendar month from resetsAt, with end-of-month clamping. Renamed the row to 1mo.
• Rounded the absolute deviation before selecting the status: ±5.4 stays on pace, while ±5.5 displays ±6pp.

Calendar time remains the default, with no weekend option in this PR.

All 43 tests pass, including the lifecycle tests from #5. Production typecheck passes. The 18 pace tests also pass with TZ=America/New_York.