Add invert option for usage segments

CLAUDE.md

@@ -62,7 +62,7 @@ Key sections in `claude-code-status-line.py` (~2,200 lines):
   - **Bayesian shrinkage**: the observed burn rate is blended toward the on-track rate using a hyperbolic trust curve `f = elapsed / (halftrust + elapsed)`, so early-window noise is dampened. Config: `usage_burndown:halftrust=16` (default 16h — at 16h elapsed, 50/50 blend).
   - **Relevance filter**: burndown is suppressed when the predicted "sooner" gap is too small relative to remaining time, using a power curve `days_remaining^coeff` (in hours) divided by the trust factor `f`. The `/f` scaling makes the threshold inversely proportional to data confidence — early-window signals need proportionally larger gaps to display. Config: `usage_burndown:coeff=1.4` (default).
   - Config: `usage_burndown:verbosity=short:coeff=1.4:halftrust=16`. Short uses compact compound durations (`5d2h30m`), `~` instead of `about`, `->` instead of `then`, drops `usage`, `out ~` instead of `may run out about`; `git_status` shows working directory state using symbols: `+` staged, `!` modified, `x` deleted, `r` renamed, `?` untracked, `=` conflicted, `$` stashed, `>` ahead, `<` behind, `<>` diverged; `added_dirs` shows directories added via `/add-dir` (from `workspace.added_dirs` JSON field), sorted alphabetically, joined by separator. Config: `added_dirs:basename_only=1:separator=,` (defaults: full path with `~` home shortening, ` • ` separator). Theme color: `text_added_dirs` (dark: Nord3 muted gray `#4C566A`, light: medium gray `#7B8394`); `worktree` shows worktree info when `worktree` field is present in JSON input (from `--worktree` sessions). Displays in curly braces `{name}`. Config: `worktree:show=name` (default), supports `name`, `branch`, `path`, `origin` or comma-separated combo like `show=name,branch`. `path` and `origin` apply `~` home shortening. `origin` maps to JSON `original` field. Theme color: `text_worktree` (dark: Nord15 purple `#B48EAD`, light: muted purple `#8B6B85`). `model` segment supports `effort` option to show reasoning effort level inside the badge: `model:effort=short` (L/M/H/A) or `model:effort=full` (low/medium/high/auto). Reads from settings files (not in statusline JSON). Precedence: `CLAUDE_CODE_EFFORT_LEVEL` env var → `.claude/settings.local.json` → `.claude/settings.json` → `~/.claude/settings.json`. `max` cannot be detected (session-only).
-- **Configuration**: global settings via `SL_THEME`, `SL_USAGE_CACHE_DURATION`, `SL_UPDATE_CACHE_DURATION`, `SL_UPDATE_RETRY_DURATION`, `SL_UPDATE_CUSTOM_RETRY_DURATION`, `SL_UPDATE_VERSION_CMD`, `SL_UPDATE_VERSION_SOURCE`, `SL_SHOW_STATUSLINE_UPDATE`, `SL_THEME_FILE`, `SL_DUMP`. All per-segment config (bar width, gauge style, hide default branch, basename-only directory) via colon-separated options in `SL_SEGMENTS`.
+- **Configuration**: global settings via `SL_THEME`, `SL_USAGE_CACHE_DURATION`, `SL_UPDATE_CACHE_DURATION`, `SL_UPDATE_RETRY_DURATION`, `SL_UPDATE_CUSTOM_RETRY_DURATION`, `SL_UPDATE_VERSION_CMD`, `SL_UPDATE_VERSION_SOURCE`, `SL_SHOW_STATUSLINE_UPDATE`, `SL_THEME_FILE`, `SL_DUMP`. All per-segment config (bar width, gauge style, hide default branch, basename-only directory, invert usage display) via colon-separated options in `SL_SEGMENTS`.
 - **Dump mode**: `SL_DUMP=1` appends every stdin JSON input (with ISO timestamp) as JSONL to `/tmp/claude-statusline-dump.jsonl`. Normal rendering continues. Useful for inspecting what Claude Code sends. Auto-cleaned on reboot.
 - **Update checker**: fetches latest version from npm registry (`@anthropic-ai/claude-code`) or custom command (`SL_UPDATE_VERSION_CMD`), compares with `claude --version` output. Custom command retries 3 times with 1s delay, then falls back to npm. Returns `(version, source)` tuple where source is `npm`, `custom`, or `npm_fallback`. Cached to `~/.claude/.update_cache.json` — success cached for `UPDATE_CACHE_DURATION` (1h), custom source failures retry after `UPDATE_CUSTOM_RETRY_DURATION` (2min), total failures retry after `UPDATE_RETRY_DURATION` (10min). Falls back to stale cache when offline.
 - **Self-update**: `--self-update` flag downloads latest version from GitHub and replaces the script atomically. Status line update notifications appear on a separate line below the main output with the update command. Controlled by `SL_SHOW_STATUSLINE_UPDATE` (default on).

README.md

@@ -133,8 +133,10 @@ Each token is `segment_name` optionally followed by `:key=value` pairs. Unknown
 | `git_branch` | `hide_default` | `0`/`1` | `1` | Hide branch on `main`/`master` |
 | `usage_5hour` | `gauge` | `vertical`/`blocks`/`none` | `blocks` | Gauge style |
 | `usage_5hour` | `width` | even integer >= 2 | `4` | Gauge width (invalid values reset to 4) |
+| `usage_5hour` | `invert` | `0`/`1` | `0` | Show used % instead of remaining % |
 | `usage_weekly` | `gauge` | `vertical`/`blocks`/`none` | `blocks` | Gauge style |
 | `usage_weekly` | `width` | even integer >= 2 | `4` | Gauge width (invalid values reset to 4) |
+| `usage_weekly` | `invert` | `0`/`1` | `0` | Show used % instead of remaining % |
 | `usage_burndown` | `verbosity` | `default`/`short` | `default` | Message style (see burndown section) |
 | `usage_burndown` | `coeff` | float | `1.4` | Relevance filter power curve exponent (see burndown section) |
 | `usage_burndown` | `halftrust` | float (hours) | `16` | Bayesian shrinkage half-trust point (see burndown section) |
@@ -160,6 +162,9 @@ SL_SEGMENTS='model progress_bar percentage tokens directory:basename_only=1 git_
 # Show git branch even on main/master
 SL_SEGMENTS='model progress_bar percentage tokens directory git_branch:hide_default=0'
 
+# Show used % instead of remaining % for usage limits
+SL_SEGMENTS='model progress_bar percentage tokens directory usage_5hour:invert=1 usage_weekly:invert=1'
+
 # Multi-line layout: model and progress on first line, directory on second
 SL_SEGMENTS='model progress_bar percentage new_line directory git_branch'
 

claude-code-status-line.py

@@ -82,8 +82,8 @@ def _env_int(key, default):
     "directory": {"basename_only": "0"},
     "added_dirs": {"basename_only": "0", "separator": " • "},
     "git_branch": {"hide_default": "1"},
-    "usage_5hour": {"gauge": "blocks", "width": "4"},
-    "usage_weekly": {"gauge": "blocks", "width": "4"},
+    "usage_5hour": {"gauge": "blocks", "width": "4", "invert": "0"},
+    "usage_weekly": {"gauge": "blocks", "width": "4", "invert": "0"},
     "worktree": {"show": "name"},
     "usage_burndown": {"coeff": "1.4"},
 }
@@ -1351,6 +1351,7 @@ def format_usage_indicators(usage_data):
 
         now = datetime.now(timezone.utc)
         remaining_pct = max(0, int(100 - utilization_pct))
+        display_pct = min(100, int(utilization_pct)) if opts.get("invert") == "1" else remaining_pct
         reset_label = reset_dt.astimezone().strftime(time_fmt)
 
         # Calculate time elapsed in window
@@ -1436,7 +1437,7 @@ def format_usage_indicators(usage_data):
             gauge = get_usage_gauge(ratio)
         gauge_part = f"{gauge}\u00a0" if gauge else ""
         results[segment_name] = (
-            f"   {gauge_part}{color}{remaining_pct}\u00a0%\u00a0→\u00a0{reset_label}"
+            f"   {gauge_part}{color}{display_pct}\u00a0%\u00a0→\u00a0{reset_label}"
         )
 
     for seg in ("usage_5hour", "usage_weekly"):