…ota (e.g., Fable) (#86)

* Read per-model weekly quota caps generically

* normalize the usage API's limits[] array instead of retired per-model fields
* rename showOpusWeekly to showScopedWeekly, migrating the stored value
* show usage credits and their monthly reset in the quota tooltip
* merge weekly caps that reset together into one tooltip row
* make the tooltip follow resetCountdownFormat and show a wall clock per row
* round reset times to the minute and drop implied .1f percent precision
* fill five missing pt-BR / id settings translations

* Nest per-model cap in the weekly status-bar figure

* print the shared weekly countdown once instead of per cap
* give each tooltip cap its own row and bar
* label the all-models row, localised across eight locales
* put the credit spend after its reset date

* Shorten the weekly all-models tooltip label

* read "Weekly (All)" to keep the window column narrow
* abbreviate the label in all eight locales

* Anchor quota reset tests to local wall time

* a fixed epoch put the 4.8h fixture on the next UTC day, so CI saw a
  weekday the assertions did not expect
* local-noon anchor keeps same-day and next-day true in every timezone

* Show usage credits as an amount, not a share

* replace the credit percentage with the spend, keeping the bar on the cap
* keep reporting spend after credits are toggled off
* handle an absent or unlimited cap instead of hiding the row
* pad the monthly reset date away from the amount

* Follow the API's own precision for quota shares

* print no decimal for the integer limits[] field, one for the legacy float
* correct the showScopedWeekly help text in all eight locales
* pin that an idle per-model cap leaves no trace in the status bar

* Pad the quota tooltip reset column

* gap the right-aligned share and reset cells
* drop the credits row's one-off pad

Co-authored-by: OpenAI Codex <215057067+openai-codex[bot]@users.noreply.github.com>

#90)

* Match official Claude app quota thresholds

* quota bars and status-bar background turn amber at 75%, red at 90%
* extract fill thresholds into named constants with a shared level helper
* keep the context-window indicator on its earlier 80/95 steps
* add wiring test pinning which indicator uses which pair

* Align quota tooltip credits row with bar column

* span the credits amount across the bar and share cells, left-aligned
* share column stays at percent width, so bars keep their gap to percents
* drop the credits row bar; its percent stays derived but undisplayed

* Show only spent amount in credits tooltip row

* drop the "/ $cap" suffix, since the cap is user-adjustable and may be unlimited
* reword the unreleased changelog entry to match

Add privacy-safe local Codex usage indexing, exact request token attribution, API-equivalent cost trends, weekly allowance estimates, a shared responsive dashboard, and related Claude profile and transcript-deduplication fixes.\n\nValidated by the complete Node, UI, and release-package gates.\n\nFixes #89\n\nCo-authored-by: OpenAI Codex <215057067+openai-codex[bot]@users.noreply.github.com>\nCo-authored-by: Claude <noreply@anthropic.com>