@v1nvn/zai 0.45.0 → 0.46.0

package/README.md

@@ -22,7 +22,7 @@ claude plugin install zai@agentic
 In a terminal, bare:
 
 ```sh
-npx -y @v1nvn/zai@0.45.0 usage
+npx -y @v1nvn/zai@0.46.0 usage
 ```
 
 ## Usage
@@ -30,10 +30,10 @@ npx -y @v1nvn/zai@0.45.0 usage
 | Invocation | Does |
 |---|---|
 | `/zai-usage` | the account's usage report, as a pane |
-| `npx -y @v1nvn/zai@0.45.0 usage` | the same report, printed bare |
-| `npx -y @v1nvn/zai@0.45.0 usage --auth-token TOKEN` | same, key on the command line (visible in `ps`) |
-| `npx -y @v1nvn/zai@0.45.0 usage --auth-token=TOKEN` | `=` form — zsh quoting-safe |
-| `npx -y @v1nvn/zai@0.45.0 usage --base-url URL` | another GLM endpoint (default `api.z.ai`) |
+| `npx -y @v1nvn/zai@0.46.0 usage` | the same report, printed bare |
+| `npx -y @v1nvn/zai@0.46.0 usage --auth-token TOKEN` | same, key on the command line (visible in `ps`) |
+| `npx -y @v1nvn/zai@0.46.0 usage --auth-token=TOKEN` | `=` form — zsh quoting-safe |
+| `npx -y @v1nvn/zai@0.46.0 usage --base-url URL` | another GLM endpoint (default `api.z.ai`) |
 
 Each setting takes the first source that provides it:
 

package/package.json

@@ -1,10 +1,10 @@
 {
   "name": "@v1nvn/zai",
-  "version": "0.45.0",
+  "version": "0.46.0",
   "description": "Query GLM Coding Plan quota and usage — the zai-usage CLI behind the zai plugin.",
   "type": "module",
   "bin": {
-    "zai-usage": "./src/index.ts"
+    "zai-usage": "./src/index.js"
   },
   "scripts": {
     "test": "yarn run -T vitest run",
@@ -35,6 +35,6 @@
     "node": ">=23.6"
   },
   "dependencies": {
-    "@v1nvn/agentic-core": "0.45.0"
+    "@v1nvn/agentic-core": "0.46.0"
   }
 }

package/src/format.js

@@ -0,0 +1,410 @@
+/**
+ * Usage-report line model: `reportLines` lays the report out once as ink-tagged
+ * segments (the pane draws those), and `render` joins them into the CLI's
+ * monospace terminal / hook-block `reason` bytes — no markdown; alignment comes
+ * from fixed-width columns and unicode block glyphs. Input is the parsed `data`
+ * of the three ZAI/ZHIPU monitor endpoints (model-usage, tool-usage,
+ * quota/limit — the last already passed through processQuotaLimit).
+ */
+import { bar, bold, dim, doc, fmtNum, fmtTokens, line, meter, MONTHS, pad, pad2, plain, rule, RULE_WIDTH, textOf, } from '@v1nvn/agentic-core/draw';
+// Lower-half block eighths for vertical bars (1/8 .. 7/8); a full cell uses '█'.
+const VBLOCKS = '▁▂▃▄▅▆▇';
+function parseSlot(s) {
+    const m = /^(\d{4})-(\d{2})-(\d{2}) (\d{2}):(\d{2})/.exec(s ?? '');
+    if (!m) {
+        return null;
+    }
+    const [, mo, d, hh, mm] = m.slice(1);
+    return {
+        h: +hh,
+        day: `${MONTHS[+mo - 1]} ${d}`,
+        time: `${hh}:${mm}`,
+    };
+}
+/**
+ * Re-express a naive "YYYY-MM-DD HH:MM" string from one UTC offset to another.
+ * Offsets are minutes east of UTC (Beijing = 480, IST = 330). Pure — does not
+ * depend on the runtime timezone, so it is deterministic under test.
+ */
+function shiftSlot(s, fromOffsetMin, toOffsetMin) {
+    const m = /^(\d{4})-(\d{2})-(\d{2}) (\d{2}):(\d{2})/.exec(s);
+    if (!m) {
+        return null;
+    }
+    const [y, mo, d, hh, mm] = m.slice(1);
+    const utcMs = Date.UTC(+y, +mo - 1, +d, +hh, +mm) - fromOffsetMin * 60000;
+    const dt = new Date(utcMs + toOffsetMin * 60000);
+    return {
+        h: dt.getUTCHours(),
+        day: `${MONTHS[dt.getUTCMonth()]} ${pad2(dt.getUTCDate())}`,
+        time: `${pad2(dt.getUTCHours())}:${pad2(dt.getUTCMinutes())}`,
+    };
+}
+function friendlyTool(code) {
+    switch ((code ?? '').toLowerCase()) {
+        case 'search-prime':
+            return 'web search';
+        case 'web-reader':
+            return 'web read';
+        default:
+            return code || 'other';
+    }
+}
+function idleRuns(tok) {
+    const runs = [];
+    let i = 0;
+    while (i < tok.length) {
+        if ((tok[i] || 0) === 0) {
+            let j = i;
+            while (j < tok.length && (tok[j] || 0) === 0) {
+                j++;
+            }
+            runs.push({ i, j: j - 1, len: j - i });
+            i = j;
+        }
+        else {
+            i++;
+        }
+    }
+    return runs;
+}
+function idleLabel(run, x, slotFn) {
+    const a = slotFn(x[run.i] ?? '');
+    const b = slotFn(x[run.j] ?? '');
+    if (!a || !b) {
+        return `idle (${run.len}h)`;
+    }
+    const start = `${a.day} ${a.time}`;
+    const end = a.day === b.day ? b.time : `${b.day} ${b.time}`;
+    return `idle ${start} → ${end} (${run.len}h)`;
+}
+// Per the usage-revision notice: Mon–Fri 14:00–18:00 Beijing (UTC+8) are peak
+// hours, where GLM-5.2 / GLM-5-Turbo tokens are billed at a 3× coefficient
+// (GLM-4.7 stays 1× all day). A bucket labelled "HH:00" spans HH:00–HH+1, so
+// the peak buckets are HH ∈ {14,15,16,17} on a weekday.
+function isPeakBucket(s) {
+    const m = /^(\d{4})-(\d{2})-(\d{2}) (\d{2}):/.exec(s ?? '');
+    if (!m) {
+        return false;
+    }
+    const [y, mo, d, hh] = m.slice(1);
+    const wd = new Date(Date.UTC(+y, +mo - 1, +d)).getUTCDay(); // 0=Sun … 6=Sat
+    return wd >= 1 && wd <= 5 && +hh >= 14 && +hh <= 17;
+}
+/** Local-time start/end of the 14:00–18:00 Beijing peak window (e.g. 11:30 / 15:30). */
+function localPeakWindow(apiOffsetMin, toOffsetMin) {
+    const a = shiftSlot('2000-01-03 14:00', apiOffsetMin, toOffsetMin);
+    const b = shiftSlot('2000-01-03 18:00', apiOffsetMin, toOffsetMin);
+    if (!a || !b) {
+        return null;
+    }
+    return { start: a.time, end: b.time };
+}
+/**
+ * Vertical bar chart of hourly tokens. Returns an array of plain-text lines
+ * (without the section header), or null if there are fewer than 2 buckets.
+ *
+ * Bars are scaled to `maxTok` across `ROWS` rows; the topmost filled cell of a
+ * bar uses a fractional lower-block glyph (▁▂▃▄▅▆▇) for sub-row height. The
+ * y-axis labels only the peak (top) and 0 (baseline); the header window and the
+ * caller's peak annotation carry the exact values.
+ */
+function hourlyVerticalChart({ x, tok, nh, maxTok, peakIdx, slotFn, }) {
+    const ROWS = 6; // bar rows above the baseline
+    if (nh < 2 || maxTok <= 0) {
+        return null;
+    }
+    const lines = [];
+    const maxLab = fmtTokens(maxTok);
+    const L = maxLab.length;
+    const IND = 1; // leading indent
+    const AXIS = IND + L + 1; // axis spine col (1-space gap after the label)
+    const P0 = AXIS + 2; // first bar column (axis + 1-char gap)
+    const PW = RULE_WIDTH - P0; // columns available for bars + gaps
+    const slot = Math.max(1, Math.floor(PW / nh));
+    const barW = slot >= 2 ? slot - 1 : 1;
+    function barLeft(i) {
+        return P0 + i * slot;
+    }
+    function barCenter(i) {
+        return barLeft(i) + Math.floor(barW / 2);
+    }
+    const usedCols = nh * slot;
+    function blank() {
+        return Array(RULE_WIDTH).fill(' ');
+    }
+    // bar glyph for hour i at row r (0 = bottom bar-row, ROWS-1 = top)
+    function glyph(i, r) {
+        const v = tok[i] || 0;
+        if (v <= 0) {
+            return ' ';
+        }
+        const fill = (v / maxTok) * ROWS - r; // how much of row r the bar covers
+        if (fill >= 1) {
+            return '█';
+        }
+        if (fill > 0) {
+            let e = Math.round(fill * 8); // 1..7 eighths
+            if (e < 1) {
+                e = 1;
+            }
+            if (e > 7) {
+                e = 7;
+            }
+            return VBLOCKS[e - 1] ?? ' ';
+        }
+        return ' ';
+    }
+    // bar rows, top → bottom; top row carries the max label + '┤', rest are '│'
+    for (let r = ROWS - 1; r >= 0; r--) {
+        const line = blank();
+        line[AXIS] = r === ROWS - 1 ? '┤' : '│';
+        if (r === ROWS - 1) {
+            for (let k = 0; k < L; k++) {
+                line[AXIS - 1 - L + k] = maxLab[k] ?? ' ';
+            }
+        }
+        for (let i = 0; i < nh; i++) {
+            const g = glyph(i, r);
+            if (g !== ' ') {
+                for (let c = 0; c < barW; c++) {
+                    line[barLeft(i) + c] = g;
+                }
+            }
+        }
+        lines.push(line.join('').trimEnd());
+    }
+    // x-axis ticks: first, last, and every bucket landing on a 6h boundary (local)
+    function slotOf(i) {
+        return slotFn(x[i] ?? '');
+    }
+    function isTick(i) {
+        if (i === 0 || i === nh - 1) {
+            return true;
+        }
+        const s = slotOf(i);
+        return !!s && s.h % 6 === 0;
+    }
+    // baseline: '0' label + corner '└' + '─' run, with '┬' under each tick
+    const base = blank();
+    base[AXIS - 2] = '0';
+    base[AXIS] = '└';
+    for (let c = AXIS + 1; c < P0 + usedCols; c++) {
+        base[c] = '─';
+    }
+    for (let i = 0; i < nh; i++) {
+        if (isTick(i)) {
+            base[barCenter(i)] = '┬';
+        }
+    }
+    lines.push(base.join('').trimEnd());
+    // hour labels under ticks (skip if too close to the previous label)
+    const hr = blank();
+    let last = -10;
+    for (let i = 0; i < nh; i++) {
+        if (!isTick(i)) {
+            continue;
+        }
+        const col = barCenter(i);
+        if (col - last < 3) {
+            continue;
+        }
+        const s = slotOf(i);
+        const txt = s ? pad2(s.h) : '  ';
+        hr[col] = txt.charAt(0);
+        hr[col + 1] = txt.charAt(1);
+        last = col;
+    }
+    lines.push(hr.join('').trimEnd());
+    // day labels under the first tick of each new local day
+    const day = blank();
+    let lastDay = null;
+    last = -10;
+    for (let i = 0; i < nh; i++) {
+        if (!isTick(i)) {
+            continue;
+        }
+        const s = slotOf(i);
+        if (!s || s.day === lastDay) {
+            continue;
+        }
+        const col = barCenter(i);
+        if (col - last >= s.day.length && col + s.day.length <= RULE_WIDTH) {
+            for (let k = 0; k < s.day.length; k++) {
+                day[col + k] = s.day[k] ?? ' ';
+            }
+            lastDay = s.day;
+            last = col + s.day.length;
+        }
+    }
+    if (day.some(c => c !== ' ')) {
+        lines.push(day.join('').trimEnd());
+    }
+    // peak-window (↑) and peak (◂) markers under bar centers
+    const mark = blank();
+    for (let i = 0; i < nh; i++) {
+        if (isPeakBucket(x[i])) {
+            mark[barCenter(i)] = '↑';
+        }
+    }
+    if (peakIdx >= 0) {
+        mark[barCenter(peakIdx)] = '◂';
+    }
+    if (mark.some(c => c !== ' ')) {
+        lines.push(mark.join('').trimEnd());
+    }
+    return lines;
+}
+export function reportLines({ platform, model, tool, quota, apiOffsetMin = 480, localOffsetMin, now = new Date(), }) {
+    const out = [];
+    // The monitor API labels every bucket in Beijing time (UTC+8). Shift each
+    // timestamp to the caller's offset (minutes east of UTC) for display.
+    // Defaults to the JS runtime's local zone.
+    const toOffset = localOffsetMin != null ? localOffsetMin : -new Date().getTimezoneOffset();
+    function convertSlot(s) {
+        return shiftSlot(s, apiOffsetMin, toOffset) ?? parseSlot(s);
+    }
+    const x = model.x_time ?? [];
+    const tok = model.tokensUsage ?? [];
+    const calls = model.modelCallCount ?? [];
+    const nh = x.length;
+    const tu = model.totalUsage ?? {};
+    const total = tu.totalTokensUsage != null
+        ? tu.totalTokensUsage
+        : tok.reduce((a, b) => a + (b || 0), 0);
+    const totalCalls = tu.totalModelCallCount != null
+        ? tu.totalModelCallCount
+        : calls.reduce((a, b) => a + (b || 0), 0);
+    let peakIdx = -1;
+    let peakTok = 0;
+    for (let i = 0; i < tok.length; i++) {
+        const t = tok[i] || 0;
+        if (t > peakTok) {
+            peakTok = t;
+            peakIdx = i;
+        }
+    }
+    const peakCalls = peakIdx >= 0 ? calls[peakIdx] || 0 : 0;
+    const pctPeak = total > 0 ? (peakTok / total) * 100 : 0;
+    const peakSlot = peakIdx >= 0 ? convertSlot(x[peakIdx] ?? '') : null;
+    const level = quota.level
+        ? quota.level.charAt(0).toUpperCase() + quota.level.slice(1)
+        : '';
+    const left = ` GLM Coding Plan${level ? ' · ' + level : ''}`;
+    const firstSlot = convertSlot(x[0] ?? '');
+    const lastSlot = convertSlot(x[nh - 1] ?? '');
+    const win = firstSlot && lastSlot
+        ? `${firstSlot.day} ${firstSlot.time} → ${lastSlot.day} ${lastSlot.time} · ${nh}h`
+        : platform;
+    out.push(line(plain(rule())));
+    out.push(line(bold(left), dim(pad(win, RULE_WIDTH - left.length, 'left'))));
+    out.push(line(plain(rule())));
+    const head = plain(` ${fmtTokens(total)} tokens across ${fmtNum(totalCalls)} model calls`);
+    out.push(line());
+    if (peakIdx >= 0 && peakSlot) {
+        out.push(line(head, plain(' — ')));
+        out.push(line(bold(`${Math.round(pctPeak)}%`), plain(' of it in a single hour')));
+        out.push(line(plain(` (${peakSlot.day} ${peakSlot.time}, ${fmtTokens(peakTok)} tokens / ${fmtNum(peakCalls)} calls).`)));
+    }
+    else {
+        out.push(line(head, plain('.')));
+    }
+    const activeHours = tok.filter(t => t > 0).length;
+    const longest = idleRuns(tok)
+        .filter(r => r.len >= 2)
+        .sort((a, b) => b.len - a.len)
+        .at(0);
+    const tt = tool.totalUsage ?? {};
+    const searchN = tt.totalNetworkSearchCount || 0;
+    const readN = tt.totalWebReadMcpCount || 0;
+    const zreadN = tt.totalZreadMcpCount || 0;
+    const toolTotal = tt.totalSearchMcpCount != null
+        ? tt.totalSearchMcpCount
+        : searchN + readN + zreadN;
+    let peakWinActive = 0;
+    let peakWinTokens = 0;
+    for (let k = 0; k < nh; k++) {
+        if (isPeakBucket(x[k]) && (tok[k] || 0) > 0) {
+            peakWinActive++;
+            peakWinTokens += tok[k] || 0;
+        }
+    }
+    const peakWinPct = total > 0 ? (peakWinTokens / total) * 100 : 0;
+    const peakWin = localPeakWindow(apiOffsetMin, toOffset);
+    out.push(line());
+    out.push(line(plain(` Peak     ${pad(peakSlot ? peakSlot.day + ' ' + peakSlot.time : '—', 15, 'right')}${pad(fmtTokens(peakTok), 7, 'left')} tokens · ${pad(fmtNum(peakCalls), 5, 'left')} calls`)));
+    out.push(line(plain(` Active   ${pad(`${activeHours} / ${nh} hours`, 15, 'right')}${longest ? idleLabel(longest, x, convertSlot) : 'no idle gaps'}`)));
+    out.push(line(plain(` Tools    ${pad(`${toolTotal} calls`, 15, 'right')}${searchN} searches · ${readN} reads${zreadN ? ` · ${zreadN} zread` : ''}`)));
+    if (peakWin) {
+        out.push(line(plain(` Peak hrs Mon–Fri ${peakWin.start}–${peakWin.end} · GLM-5.2 3× · ${peakWinActive}h active · ${fmtTokens(peakWinTokens)} (`), bold(`${Math.round(peakWinPct)}%`), plain(')')));
+    }
+    out.push(line());
+    const chartHdr = ' Hourly tokens · ↑ peak hour ';
+    out.push(line(dim(chartHdr + '─'.repeat(Math.max(0, RULE_WIDTH - chartHdr.length)))));
+    const maxTok = peakTok || 1;
+    const chartLines = hourlyVerticalChart({
+        x,
+        tok,
+        nh,
+        maxTok,
+        peakIdx,
+        slotFn: convertSlot,
+    });
+    if (chartLines) {
+        out.push(...chartLines.map((l) => line(plain(l))));
+        if (peakIdx >= 0 && peakSlot) {
+            out.push(line(plain(`   ◂ peak  ${peakSlot.day} ${peakSlot.time}  ${fmtTokens(peakTok)} tokens · ${fmtNum(peakCalls)} calls`)));
+        }
+    }
+    else {
+        out.push(line(plain('   (not enough hourly data to chart)')));
+    }
+    out.push(line());
+    out.push(line(dim(' Model mix ' + '─'.repeat(Math.max(0, RULE_WIDTH - 11)))));
+    const mixSrc = model.modelSummaryList ?? model.modelDataList ?? [];
+    const denom = mixSrc.reduce((a, m) => a + (m.totalTokens ?? 0), 0) || 1;
+    const mixSorted = [...mixSrc].sort((a, b) => (a.sortOrder || 0) - (b.sortOrder || 0));
+    for (const m of mixSorted) {
+        const pct = ((m.totalTokens ?? 0) / denom) * 100;
+        out.push(line(plain(`   ${pad(m.modelName || '?', 11, 'right')}${pad(fmtTokens(m.totalTokens), 8, 'left')}  `), bold(pad(pct.toFixed(1) + '%', 6, 'left')), plain(`  ${bar(pct, 100, 20)}`)));
+    }
+    out.push(line());
+    out.push(line(dim(' Limits ' + '─'.repeat(Math.max(0, RULE_WIDTH - 8)))));
+    // The window does not bill on weekends, so the meter stays empty off Mon–Fri.
+    const bj = new Date(now.getTime() + apiOffsetMin * 60000);
+    const bjDay = bj.getUTCDay();
+    const inWindow = bjDay >= 1 && bjDay <= 5;
+    const elapsedMin = inWindow
+        ? Math.max(0, Math.min(240, (bj.getUTCHours() - 14) * 60 + bj.getUTCMinutes()))
+        : 0;
+    if (peakWin) {
+        out.push(line(dim(`   ${pad('Peak', 16, 'right')}`), plain(`${pad(peakWin.start, 5, 'left')}  `), ...meter(Math.round((elapsedMin / 240) * 100), 22), plain(`  ${peakWin.end}`)));
+    }
+    const limits = quota.limits ?? [];
+    const mcp = limits.find(l => /mcp/i.test(l.type ?? ''));
+    const tok5 = limits.find(l => /token/i.test(l.type ?? ''));
+    if (tok5) {
+        const reset = new Date(tok5.nextResetTime ?? '').toLocaleTimeString('en-IN');
+        out.push(line(dim(`   ${pad('Tokens · 5h', 16, 'right')}`), bold(pad(`${tok5.percentage || 0}%`, 5, 'left')), plain('  '), ...meter(tok5.percentage, 22), plain(`  ${reset}`)));
+    }
+    if (tok5 && mcp) {
+        out.push(line());
+    }
+    if (mcp) {
+        const reset = new Date(mcp.nextResetTime ?? '').toLocaleString('en-IN');
+        out.push(line(dim(`   ${pad('MCP · this month', 16, 'right')}`), bold(pad(`${mcp.percentage || 0}%`, 5, 'left')), plain('  '), ...meter(mcp.percentage, 22), plain(`  ${reset}`)));
+        const det = mcp.usageDetails ?? [];
+        const parts = det
+            .map(d => `${friendlyTool(d.modelCode)} ${fmtNum(d.usage)}`)
+            .join('  ');
+        if (parts) {
+            out.push(line(plain(`   ${fmtNum(mcp.currentUsage)}M / ${fmtNum(mcp.totol)}M · ${parts}`)));
+        }
+    }
+    out.push(line());
+    out.push(line(plain(rule())));
+    return out;
+}
+export function render(input) {
+    return textOf(doc(...reportLines(input)));
+}

package/src/index.js

@@ -0,0 +1,16 @@
+#!/usr/bin/env node
+import { runCli } from '@v1nvn/agentic-core';
+import { render, reportLines } from "./format.js";
+import { cliSpec, resolveConfig } from "./resolve.js";
+import { fetchReport } from "./usage.js";
+process.exitCode = await runCli(cliSpec(async (options) => {
+    const input = await fetchReport(resolveConfig(process.env, {
+        authToken: options.authToken || undefined,
+        baseUrl: options.baseUrl || undefined,
+        command: 'usage',
+        json: options.json === true,
+    }));
+    console.log(options.json === true
+        ? JSON.stringify({ lines: reportLines(input) })
+        : render(input));
+}), process.argv.slice(2));

package/src/resolve.js

@@ -0,0 +1,99 @@
+import { buildCli, HelpRequested, parseCli, } from '@v1nvn/agentic-core';
+const DEFAULT_BASE_URL = 'https://api.z.ai';
+const GLM_HOSTS = new Set(['api.z.ai', 'dev.bigmodel.cn', 'open.bigmodel.cn']);
+function fromEnv(value) {
+    return value ? value : undefined;
+}
+function originOf(value) {
+    try {
+        const url = new URL(value);
+        return url.protocol === 'https:' || url.protocol === 'http:'
+            ? url.origin
+            : undefined;
+    }
+    catch {
+        return undefined;
+    }
+}
+function isGlm(origin) {
+    return GLM_HOSTS.has(new URL(origin).hostname);
+}
+function resolveBaseUrl(flag, env) {
+    if (flag !== undefined) {
+        const origin = originOf(flag);
+        if (origin === undefined) {
+            throw new Error(`invalid base URL: ${flag}`);
+        }
+        return { glm: isGlm(origin), origin };
+    }
+    const inherited = fromEnv(env.ANTHROPIC_BASE_URL);
+    const origin = inherited === undefined ? undefined : originOf(inherited);
+    if (origin !== undefined && isGlm(origin)) {
+        return { glm: true, origin };
+    }
+    return { glm: false, origin: DEFAULT_BASE_URL };
+}
+export function cliSpec(run) {
+    return {
+        name: 'zai-usage',
+        description: 'GLM Coding Plan usage — one command, usage',
+        commands: [
+            {
+                name: 'usage',
+                description: 'GLM Coding Plan quota and usage report',
+                options: [
+                    {
+                        flags: '--auth-token <token>',
+                        description: 'API key',
+                        env: 'ZAI_AUTH_TOKEN',
+                    },
+                    {
+                        flags: '--base-url <url>',
+                        description: 'base URL',
+                        env: 'ZAI_BASE_URL',
+                    },
+                    {
+                        flags: '--json',
+                        description: 'print the report lines as JSON for the zai mod',
+                    },
+                ],
+                run,
+            },
+        ],
+    };
+}
+export function buildProgram() {
+    return buildCli(cliSpec());
+}
+export function parseArgs(args) {
+    const parsed = parseCli(cliSpec(), args);
+    if (parsed === undefined || parsed instanceof HelpRequested) {
+        return parsed;
+    }
+    if (parsed.command !== 'usage') {
+        return {
+            authToken: undefined,
+            baseUrl: undefined,
+            command: undefined,
+            json: false,
+        };
+    }
+    const { authToken, baseUrl, json } = parsed.options;
+    return {
+        authToken: authToken || undefined,
+        baseUrl: baseUrl || undefined,
+        command: 'usage',
+        json: json === true,
+    };
+}
+export function resolveConfig(env, parsed) {
+    const base = resolveBaseUrl(parsed.baseUrl, env);
+    const token = parsed.authToken ??
+        (base.glm ? fromEnv(env.ANTHROPIC_AUTH_TOKEN) : undefined);
+    if (token === undefined) {
+        throw new Error(env.ANTHROPIC_AUTH_TOKEN
+            ? 'ANTHROPIC_AUTH_TOKEN is set, but the resolved base URL does not name a GLM host, so it is not a GLM Coding Plan token — set ZAI_AUTH_TOKEN (or --auth-token)'
+            : 'no API key: set ZAI_AUTH_TOKEN (or --auth-token)');
+    }
+    return { token, url: base.origin };
+}

package/src/usage.js

@@ -0,0 +1,87 @@
+/**
+ * Usage query: the three GLM monitor endpoints of the resolved base URL,
+ * assembled into the report input.
+ */
+import { pad2 } from '@v1nvn/agentic-core/draw';
+// The monitor API labels every bucket in Beijing time; the query window runs
+// from yesterday at the current hour to the end of today's current hour.
+function queryParams(now = new Date()) {
+    const startDate = new Date(now.getFullYear(), now.getMonth(), now.getDate() - 1, now.getHours(), 0, 0, 0);
+    const endDate = new Date(now.getFullYear(), now.getMonth(), now.getDate(), now.getHours(), 59, 59, 999);
+    function formatDateTime(date) {
+        return `${date.getFullYear()}-${pad2(date.getMonth() + 1)}-${pad2(date.getDate())} ${pad2(date.getHours())}:${pad2(date.getMinutes())}:${pad2(date.getSeconds())}`;
+    }
+    return `?startTime=${encodeURIComponent(formatDateTime(startDate))}&endTime=${encodeURIComponent(formatDateTime(endDate))}`;
+}
+// Rename the API's limit kinds to their display names; `totol` is the field the
+// renderer reads (sic — the upstream spelling).
+function processQuotaLimit(data) {
+    if (!data.limits) {
+        return data;
+    }
+    data.limits = data.limits.map(item => {
+        if (item.type === 'TOKENS_LIMIT') {
+            return {
+                type: 'Token usage(5 Hour)',
+                percentage: item.percentage,
+                nextResetTime: item.nextResetTime,
+            };
+        }
+        if (item.type === 'TIME_LIMIT') {
+            return {
+                type: 'MCP usage(1 Month)',
+                percentage: item.percentage,
+                currentUsage: item.currentValue,
+                totol: item.usage,
+                usageDetails: item.usageDetails,
+                nextResetTime: item.nextResetTime,
+            };
+        }
+        return item;
+    });
+    return data;
+}
+async function fetchJson(apiUrl, label, authToken, query = '') {
+    const res = await fetch(apiUrl + query, {
+        headers: {
+            Authorization: authToken,
+            'Accept-Language': 'en-US,en',
+            'Content-Type': 'application/json',
+        },
+    });
+    const body = await res.text();
+    if (res.status !== 200) {
+        throw new Error(`[${label}] HTTP ${res.status}\n${body}`);
+    }
+    let json;
+    try {
+        json = JSON.parse(body);
+    }
+    catch (e) {
+        throw new Error(`[${label}] could not parse response: ${e.message}`, { cause: e });
+    }
+    if (json.success === false) {
+        throw new Error(`[${label}] ${json.msg ?? 'request failed'} (code ${json.code ?? '?'})`);
+    }
+    return json.data ?? json;
+}
+export async function fetchReport(config) {
+    const platform = config.url.includes('bigmodel') ? 'ZHIPU' : 'ZAI';
+    const query = queryParams();
+    try {
+        const [model, tool, quotaRaw] = await Promise.all([
+            fetchJson(`${config.url}/api/monitor/usage/model-usage`, 'Model usage', config.token, query),
+            fetchJson(`${config.url}/api/monitor/usage/tool-usage`, 'Tool usage', config.token, query),
+            fetchJson(`${config.url}/api/monitor/usage/quota/limit`, 'Quota limit', config.token),
+        ]);
+        return {
+            platform,
+            model: model,
+            tool: tool,
+            quota: processQuotaLimit(quotaRaw),
+        };
+    }
+    catch (e) {
+        throw new Error(`Request failed: ${e.message}`, { cause: e });
+    }
+}