pi-light-statusline

A lightweight colored statusline for pi — official layout, your colors, live tok/s & cache-hit rate, and optional AI working vibes.

中文说明

What Is This

pi's built-in footer is informative but monochrome, and it can't show generation speed. pi-powerline-footer is powerful but heavy — 20+ segments, presets, themes, welcome overlays.

pi-light-statusline is the middle path: the built-in footer's clean layout (left stats, right-aligned model + thinking level), plus:

• colors & Nerd Font icons for every segment (ASCII terminals degrade to plain text)
• tok/s — live output speed of the current assistant message, frozen at its final value
• cache-hit rate — prompt cache efficiency of the latest turn
• AI working vibes (optional, off by default) — themed loading messages generated by any model you configure

 gists.lanlance.cn   main   53k/1.0M (5.3%)
 47 tok/s  78% Cache                               glm-5.3  󰠚 xhigh

Other extensions' footer statuses (e.g. subagent tooling) keep working: they render on their own line below the statusline, exactly like the built-in footer.

Install

pi install npm:pi-light-statusline

or via git:

pi install git:https://github.com/L2ncE/pi-light-statusline

or try without installing:

pi -e /path/to/pi-light-statusline

No configuration needed — sensible defaults apply. Restart pi (or /reload).

Segments

| id | default side | icon | shows | |---|---|---|---| | model | right | | active model id, lowercased | | thinking | right | 󰠚 | thinking level; high/xhigh/max get a rainbow color | | path | line 1 | | current directory basename | | git | line 1 | | current branch | | context | line 1 | | context usage 53k/1.0M (5.3%); warning > 70%, error > 90% | | tps | line 2 | | 47 tok/s live during streaming, frozen after; hidden the first second | | cache_rate | line 2 | | cache hit rate of the latest turn, 78% |

Empty segments (no git branch, no tokens yet) hide themselves.

Configuration

Everything lives in one lightStatusline block of ~/.pi/agent/settings.json. Arrays set segment order; omit them to keep defaults.

{
  "lightStatusline": {
    "line1": ["path", "git", "context"],
    "line2": ["tps", "cache_rate"],
    "right": ["model", "thinking"],
    "colors": {
      "model": "#0ABAB5",
      "path": "#e06c75",
      "git": "#1a7f37",
      "context": "#98c379"
    },
    "icons": { "tps": "" },
    "vibes": { "enabled": false }
  }
}

• line1 / line2 / right — ordered segment ids per row: line 1 (path/git/context), line 2 stats (tps/cache_rate), right-aligned (model/thinking). A legacy single left array is split automatically. Unknown ids are ignored.
• colors — per-segment color: any pi theme color name (accent, warning, dim, …) or #RRGGBB. Defaults are muted theme colors; context always escalates to warning/error at 70%/90%.
• icons — per-segment icon override. Detection follows your terminal (set LIGHT_STATUSLINE_NERD_FONTS=1 to force Nerd Font icons, =0 to disable).

Vibes

While pi works, replace the boring default loader with a themed, task-aware message generated by a model of your choice — for example:

⠦ 凝神聚念，洞彻微观玄机...

{
  "lightStatusline": {
    "vibes": {
      "enabled": true,
      "theme": "神龙尊者",                    // your persona
      "model": "mccodex/LongCat-Flash-Chat", // any configured pi model
      "fallback": "龙威震世",                 // shown on failure/timeout
      "prompt": "…",                          // optional custom template
      "maxLength": 18,
      "timeoutMs": 3000,
      "color": "rainbow"                      // optional: color name, #hex, or "rainbow"
    }
  }
}

One generation per agent run (no per-tool-call refresh spam). The prompt template supports {theme}, {task} (the current prompt), {exclude} (recent vibes, to avoid repetition) and {maxLength}. On timeout or error the fallback message stays — vibes never break your session.

License

MIT © L2ncE