pi-zhipu-usage

A pi extension that shows your GLM Coding Plan usage and next-reset countdown in the footer status bar.

Features

• 📊 Shows both the 5-hour rolling window and weekly window usage (% + reset countdown)
• 🎨 Color-coded by usage: 🟢 green <30% · default · 🟠 orange ≥70% · 🔴 red ≥90%
• ⚙️ Zero-config — auto-detects your GLM API key from pi's configured providers
• 🌐 Works with Zhipu domestic (open.bigmodel.cn) and international (api.z.ai); also auto-detects GLM behind a local proxy (e.g. Headroom) — see How it works

Example footer line:

GLM lite  │  5h 69% · 3h3m 后刷新  │  周 28% · 3d 后刷新

Install

From npm:

pi install npm:pi-zhipu-usage

Or from GitHub:

pi install git:github.com/caimaoy/pi-glm-usage

Or add to ~/.pi/agent/settings.json:

{
  "packages": ["npm:pi-zhipu-usage"]
}

Then /reload in pi (or restart). With v0.2.0+, /reload lights up the footer on your next message — no restart needed. The status shows up once a GLM provider is configured.

Replace caimaoy with the actual GitHub owner if you fork.

How it works

Queries Zhipu's internal quota endpoint GET /api/monitor/usage/quota/limit with your Coding Plan API key (note: no Bearer prefix). The endpoint and response shape follow the production implementation in cc-switch (src-tauri/src/services/coding_plan.rs).

The 5h / weekly windows are distinguished by the unit field (3 = 5h rolling, 6 = weekly). Colors use truecolor ANSI so they render vividly regardless of theme.

Proxy / headroom setups: If your GLM provider points at a local proxy (e.g. http://127.0.0.1:8787) instead of Zhipu directly, the extension still works — it recognizes the provider by name + key shape and queries Zhipu's quota endpoint directly with your key (the request goes to open.bigmodel.cn, not your proxy). This is required because the quota endpoint isn't exposed by OpenAI-compatible proxies. Note: this means your key is sent to Zhipu's own infra for the quota lookup even when you route model traffic through a proxy.

Config

Edit extensions/glm-usage.ts top constants:

• POLL_INTERVAL_MS — refresh interval (default 3 min)
• VIVID — colors (truecolor hex)
• thresholds in colorForPct (default 30 / 70 / 90)
• DEFAULT_GLM_HOST — fallback host for proxy setups. Defaults to the domestic site https://open.bigmodel.cn; international (api.z.ai) users can override it with the GLM_QUOTA_HOST env var (e.g. GLM_QUOTA_HOST=https://api.z.ai). Direct connections are unaffected and always use the host from the provider's baseUrl.
• GLM_PROVIDER_NAME_RE / GLM_KEY_RE — heuristics that recognize a GLM provider behind a proxy; both must match for the proxy fallback to fire.

The extension is a single self-contained file with no external type imports.

Requirements

• A configured GLM provider in pi (Zhipu/GLM Coding Plan API key)

License

MIT