@v1nvn/zai 0.35.1 → 0.36.1

package/README.md

@@ -1,14 +1,18 @@
 # @v1nvn/zai
 
-The CLI behind the zai plugin: it reports GLM Coding Plan quota and usage for
-the current account, straight to the terminal — no model tokens, no browser.
+The home of the zai plugin — and its plugin root: the `zai-usage` CLI in `src/`,
+the mod's hooks module at `hooks/register.tsx` (`.ts`/`.tsx` the engine loads
+directly, no build), engine tests at `tests/`. It reports GLM Coding Plan quota
+and usage for the current account — models, tools, limits — straight to the
+terminal; no browser, nothing written.
 
 User-facing docs: [root README](../../README.md).
 
 ## Quickstart
 
-In Claude Code, the plugin is the way in — `/zai:usage` runs this CLI through
-a `UserPromptExpansion` hook with zero model tokens:
+In Claude Code, the plugin is the way in — a mod serves the one surface the
+model never reads: `/zai-usage` execs the shipped CLI and opens its report as
+a pane (nothing execs until the command runs).
 
 ```sh
 claude plugin marketplace add v1nvn/agentic
@@ -18,17 +22,18 @@ claude plugin install zai@agentic
 In a terminal, bare:
 
 ```sh
-npx -y @v1nvn/zai@0.35.1
+npx -y @v1nvn/zai@0.36.1
 ```
 
 ## Usage
 
 | Invocation | Does |
 |---|---|
-| `npx -y @v1nvn/zai@0.35.1` | usage report — models, quota window, remaining balance |
-| `npx -y @v1nvn/zai@0.35.1 --auth-token TOKEN` | same, key on the command line (visible in `ps`) |
-| `npx -y @v1nvn/zai@0.35.1 --auth-token=TOKEN` | `=` form — zsh quoting-safe |
-| `npx -y @v1nvn/zai@0.35.1 --base-url URL` | another GLM endpoint (default `api.z.ai`) |
+| `/zai-usage` | the account's usage report, as a pane |
+| `npx -y @v1nvn/zai@0.36.1` | the same report, printed bare |
+| `npx -y @v1nvn/zai@0.36.1 --auth-token TOKEN` | same, key on the command line (visible in `ps`) |
+| `npx -y @v1nvn/zai@0.36.1 --auth-token=TOKEN` | `=` form — zsh quoting-safe |
+| `npx -y @v1nvn/zai@0.36.1 --base-url URL` | another GLM endpoint (default `api.z.ai`) |
 
 Each setting takes the first source that provides it:
 
@@ -42,8 +47,8 @@ Each setting takes the first source that provides it:
 to a GLM Coding Plan. Claude Code routed elsewhere (plain Anthropic, another
 proxy) is not a zai configuration; set `ZAI_AUTH_TOKEN`. Bigmodel accounts
 point the base URL at their host; the monitor paths are identical
-(`ZAI_BASE_URL=https://open.bigmodel.cn`). The hook reads the same env from
-the Claude Code process; flags are a CLI affordance — `hooks.json` is static.
+(`ZAI_BASE_URL=https://open.bigmodel.cn`). The mod's exec runs inside the
+session's shell, so it inherits the same environment `npx zai-usage` reads.
 
 The API labels every bucket in Beijing time (UTC+8); the report shifts each
 timestamp to your local zone for display.
@@ -51,22 +56,34 @@ timestamp to your local zone for display.
 ## Develop
 
 ```sh
-yarn workspace @v1nvn/zai build
-yarn workspace @v1nvn/zai test
-yarn lint && yarn typecheck       # from the repo root
+yarn workspace @v1nvn/zai test       # the CLI side (vitest)
+claude plugin validate packages/zai  # the mod, as the engine reads it
+claude plugin test packages/zai      # the mod, through the engine
+yarn lint && yarn typecheck         # from the repo root (typecheck
+                                    #   covers tsconfig.mods.json too)
 ```
 
+The mod API is early access and moves between Claude Code releases — a build
+that refuses the module loads nothing, so validate after every engine update
+and re-vendor `types/claude-code.d.ts` (repo root) from the engine-laid
+`.claude-plugin/types/` when it changes. Read `references/mods.md` first.
+
 ## Modules
 
 | File | Role |
 |---|---|
 | `src/index.ts` | bin entry (`zai-usage`) — dispatch, exit codes |
-| `src/usage.ts` | the GLM API call and the report it builds |
+| `src/usage.ts` | the GLM API call and the report input it assembles |
 | `src/resolve.ts` | auth-token resolution chain |
-| `src/format.ts` | the printed table |
+| `src/format.ts` | the report's lines — printed by the CLI, drawn by the pane |
+| `hooks/register.tsx` | the mod: registers `/zai-usage`, execs the CLI, opens its report as a pane |
+| `bin/usage.mjs` | the standalone build the mod execs — committed, synced by `yarn build` |
 
 ## Contracts
 
 - One job: print the report, exit 0. Nothing is written anywhere.
-- The hook path and the terminal path run the same bin — the hook only
-  feeds it the session's transcript context.
+- `/zai-usage` is one `node <plugin root>/bin/usage.mjs --json` exec through
+  the session's Bash tool — the same query `npx zai-usage` runs, awaited
+  as-is. The one allow rule it needs is
+  `Bash(node <plugin root>/bin/usage.mjs --json)`; the first run asks for it
+  once, then it is remembered. Its report is a pane the model never reads.

package/dist/index.js

@@ -1,42 +1,6 @@
 #!/usr/bin/env node
 import { Command, Option } from "commander";
 //#region ../core/dist/index.js
-function readAll(stream) {
-	return new Promise((resolve, reject) => {
-		let data = "";
-		stream.setEncoding("utf8");
-		stream.on("data", (chunk) => {
-			data += chunk;
-		});
-		stream.on("end", () => {
-			resolve(data);
-		});
-		stream.on("error", reject);
-	});
-}
-/**
-* Read the Claude Code hook event JSON from stdin; empty or unparseable input
-* yields an empty event, never a thrown error — a hook must always answer.
-*/
-function readHookEvent(stream = process.stdin) {
-	return readAll(stream).then((data) => {
-		try {
-			return JSON.parse(data);
-		} catch {
-			return {};
-		}
-	}, () => ({}));
-}
-/**
-* Print the UserPromptExpansion decision for a zero-token command: `block`
-* keeps the command from reaching the model, with the output as the `reason`.
-*/
-function emitHookBlock(reason) {
-	process.stdout.write(`${JSON.stringify({
-		decision: "block",
-		reason
-	})}\n`);
-}
 function parseQuietly(program, args, recover) {
 	try {
 		program.allowExcessArguments(false).exitOverride().configureOutput({
@@ -52,90 +16,8 @@ function printUsageAndExit(program) {
 	console.error(program.helpInformation());
 	process.exit(1);
 }
-async function hookOrPrint(hook, failure, run, stdin = process.stdin) {
-	if (hook) {
-		let reason;
-		try {
-			reason = await run(await readHookEvent(stdin));
-		} catch (e) {
-			reason = `${failure}: ${e.message}`;
-		}
-		emitHookBlock(reason);
-		return;
-	}
-	try {
-		console.log(await run(void 0));
-	} catch (e) {
-		console.error(e.message);
-		process.exit(1);
-	}
-}
 Number.MAX_SAFE_INTEGER;
 //#endregion
-//#region src/resolve.ts
-var DEFAULT_BASE_URL = "https://api.z.ai";
-var GLM_HOSTS = /* @__PURE__ */ new Set([
-	"api.z.ai",
-	"dev.bigmodel.cn",
-	"open.bigmodel.cn"
-]);
-function fromEnv(value) {
-	return value ? value : void 0;
-}
-function originOf(value) {
-	try {
-		const url = new URL(value);
-		return url.protocol === "https:" || url.protocol === "http:" ? url.origin : void 0;
-	} catch {
-		return;
-	}
-}
-function isGlm(origin) {
-	return GLM_HOSTS.has(new URL(origin).hostname);
-}
-function resolveBaseUrl(flag, env) {
-	if (flag !== void 0) {
-		const origin = originOf(flag);
-		if (origin === void 0) throw new Error(`invalid base URL: ${flag}`);
-		return {
-			glm: isGlm(origin),
-			origin
-		};
-	}
-	const inherited = fromEnv(env.ANTHROPIC_BASE_URL);
-	const origin = inherited === void 0 ? void 0 : originOf(inherited);
-	if (origin !== void 0 && isGlm(origin)) return {
-		glm: true,
-		origin
-	};
-	return {
-		glm: false,
-		origin: DEFAULT_BASE_URL
-	};
-}
-function buildProgram() {
-	return new Command().name("zai-usage").addOption(new Option("--auth-token <token>", "API key").env("ZAI_AUTH_TOKEN")).addOption(new Option("--base-url <url>", "base URL").env("ZAI_BASE_URL")).option("--hook", "emit a UserPromptExpansion block instead of printing");
-}
-function parseArgs(args) {
-	const program = parseQuietly(buildProgram(), args);
-	if (program === void 0) return;
-	const { authToken, baseUrl, hook } = program.opts();
-	return {
-		authToken: authToken || void 0,
-		baseUrl: baseUrl || void 0,
-		hook: hook ?? false
-	};
-}
-function resolveConfig(env, parsed) {
-	const base = resolveBaseUrl(parsed.baseUrl, env);
-	const token = parsed.authToken ?? (base.glm ? fromEnv(env.ANTHROPIC_AUTH_TOKEN) : void 0);
-	if (token === void 0) 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
-	};
-}
-//#endregion
 //#region ../tokens/dist/text.js
 /**
 * Plain-text rendering primitives for the fixed-width reports. Output targets
@@ -213,13 +95,12 @@ function meter(pct, width) {
 //#endregion
 //#region src/format.ts
 /**
-* Plain-text usage report renderer.
-*
-* Rendered for a monospace terminal / hook-block `reason`, so it must NOT rely on
-* markdown (bold, tables, fenced code). 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).
+* 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).
 */
 var VBLOCKS = "▁▂▃▄▅▆▇";
 function parseSlot(s) {
@@ -392,7 +273,22 @@ function hourlyVerticalChart({ x, tok, nh, maxTok, peakIdx, slotFn }) {
 	if (mark.some((c) => c !== " ")) lines.push(mark.join("").trimEnd());
 	return lines;
 }
-function render({ platform, model, tool, quota, apiOffsetMin = 480, localOffsetMin, now = /* @__PURE__ */ new Date() }) {
+function plain(text) {
+	return { text };
+}
+function dim(text) {
+	return {
+		text,
+		ink: "dim"
+	};
+}
+function bold(text) {
+	return {
+		text,
+		ink: "bold"
+	};
+}
+function reportLines({ platform, model, tool, quota, apiOffsetMin = 480, localOffsetMin, now = /* @__PURE__ */ new Date() }) {
 	const out = [];
 	const toOffset = localOffsetMin != null ? localOffsetMin : -(/* @__PURE__ */ new Date()).getTimezoneOffset();
 	function convertSlot(s) {
@@ -419,14 +315,16 @@ function render({ platform, model, tool, quota, apiOffsetMin = 480, localOffsetM
 	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(rule());
-	out.push(left + padL(win, 68 - left.length));
-	out.push(rule());
-	let lead = ` ${fmtTokens(total)} tokens across ${fmtNum(totalCalls)} model calls`;
-	if (peakIdx >= 0 && peakSlot) lead += ` — ${Math.round(pctPeak)}% of it in a single hour (${peakSlot.day} ${peakSlot.time}, ${fmtTokens(peakTok)} tokens / ${fmtNum(peakCalls)} calls)`;
-	lead += ".";
-	out.push("");
-	out.push(lead);
+	out.push([plain(rule())]);
+	out.push([bold(left), dim(padL(win, 68 - left.length))]);
+	out.push([plain(rule())]);
+	const head = plain(` ${fmtTokens(total)} tokens across ${fmtNum(totalCalls)} model calls`);
+	out.push([]);
+	if (peakIdx >= 0 && peakSlot) {
+		out.push([head, plain(" — ")]);
+		out.push([bold(`${Math.round(pctPeak)}%`), plain(" of it in a single hour")]);
+		out.push([plain(` (${peakSlot.day} ${peakSlot.time}, ${fmtTokens(peakTok)} tokens / ${fmtNum(peakCalls)} calls).`)]);
+	} else out.push([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 ?? {};
@@ -442,13 +340,17 @@ function render({ platform, model, tool, quota, apiOffsetMin = 480, localOffsetM
 	}
 	const peakWinPct = total > 0 ? peakWinTokens / total * 100 : 0;
 	const peakWin = localPeakWindow(apiOffsetMin, toOffset);
-	out.push("");
-	out.push(` Peak     ${padR(peakSlot ? peakSlot.day + " " + peakSlot.time : "—", 15)}${padL(fmtTokens(peakTok), 7)} tokens · ${padL(fmtNum(peakCalls), 5)} calls`);
-	out.push(` Active   ${padR(`${activeHours} / ${nh} hours`, 15)}${longest ? idleLabel(longest, x, convertSlot) : "no idle gaps"}`);
-	out.push(` Tools    ${padR(`${toolTotal} calls`, 15)}${searchN} searches · ${readN} reads${zreadN ? ` · ${zreadN} zread` : ""}`);
-	if (peakWin) out.push(` Peak hrs Mon–Fri ${peakWin.start}–${peakWin.end} · GLM-5.2 3× · ${peakWinActive}h active · ${fmtTokens(peakWinTokens)} (${Math.round(peakWinPct)}%)`);
-	out.push("");
-	out.push(" Hourly tokens · ↑ peak hour " + "─".repeat(Math.max(0, 39)));
+	out.push([]);
+	out.push([plain(` Peak     ${padR(peakSlot ? peakSlot.day + " " + peakSlot.time : "—", 15)}${padL(fmtTokens(peakTok), 7)} tokens · ${padL(fmtNum(peakCalls), 5)} calls`)]);
+	out.push([plain(` Active   ${padR(`${activeHours} / ${nh} hours`, 15)}${longest ? idleLabel(longest, x, convertSlot) : "no idle gaps"}`)]);
+	out.push([plain(` Tools    ${padR(`${toolTotal} calls`, 15)}${searchN} searches · ${readN} reads${zreadN ? ` · ${zreadN} zread` : ""}`)]);
+	if (peakWin) out.push([
+		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([]);
+	out.push([dim(" Hourly tokens · ↑ peak hour " + "─".repeat(Math.max(0, 39)))]);
 	const chartLines = hourlyVerticalChart({
 		x,
 		tok,
@@ -458,48 +360,126 @@ function render({ platform, model, tool, quota, apiOffsetMin = 480, localOffsetM
 		slotFn: convertSlot
 	});
 	if (chartLines) {
-		out.push(...chartLines);
-		if (peakIdx >= 0 && peakSlot) out.push(`   ◂ peak  ${peakSlot.day} ${peakSlot.time}  ${fmtTokens(peakTok)} tokens · ${fmtNum(peakCalls)} calls`);
-	} else out.push("   (not enough hourly data to chart)");
-	out.push("");
-	out.push(" Model mix " + "─".repeat(Math.max(0, 57)));
+		out.push(...chartLines.map((l) => [plain(l)]));
+		if (peakIdx >= 0 && peakSlot) out.push([plain(`   ◂ peak  ${peakSlot.day} ${peakSlot.time}  ${fmtTokens(peakTok)} tokens · ${fmtNum(peakCalls)} calls`)]);
+	} else out.push([plain("   (not enough hourly data to chart)")]);
+	out.push([]);
+	out.push([dim(" Model mix " + "─".repeat(Math.max(0, 57)))]);
 	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(`   ${padR(m.modelName || "?", 11)}${padL(fmtTokens(m.totalTokens), 8)}  ${padL(pct.toFixed(1) + "%", 6)}  ${barField(pct, 100, 20)}`);
+		out.push([
+			plain(`   ${padR(m.modelName || "?", 11)}${padL(fmtTokens(m.totalTokens), 8)}  `),
+			bold(padL(pct.toFixed(1) + "%", 6)),
+			plain(`  ${barField(pct, 100, 20)}`)
+		]);
 	}
-	out.push("");
-	out.push(" Limits " + "─".repeat(Math.max(0, 60)));
+	out.push([]);
+	out.push([dim(" Limits " + "─".repeat(Math.max(0, 60)))]);
 	const bj = new Date(now.getTime() + apiOffsetMin * 6e4);
 	const bjDay = bj.getUTCDay();
 	const elapsedMin = bjDay >= 1 && bjDay <= 5 ? Math.max(0, Math.min(240, (bj.getUTCHours() - 14) * 60 + bj.getUTCMinutes())) : 0;
-	if (peakWin) out.push(`   ${padR("Peak", 16)}${padL(peakWin.start, 5)}  ${meter(Math.round(elapsedMin / 240 * 100), 22)}  ${peakWin.end}`);
+	if (peakWin) out.push([dim(`   ${padR("Peak", 16)}`), plain(`${padL(peakWin.start, 5)}  ${meter(Math.round(elapsedMin / 240 * 100), 22)}  ${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(`   ${padR("Tokens · 5h", 16)}${padL(`${tok5.percentage || 0}%`, 5)}  ${meter(tok5.percentage, 22)}  ${reset}`);
+		out.push([
+			dim(`   ${padR("Tokens · 5h", 16)}`),
+			bold(padL(`${tok5.percentage || 0}%`, 5)),
+			plain(`  ${meter(tok5.percentage, 22)}  ${reset}`)
+		]);
 	}
-	if (tok5 && mcp) out.push("");
+	if (tok5 && mcp) out.push([]);
 	if (mcp) {
 		const reset = new Date(mcp.nextResetTime ?? "").toLocaleString("en-IN");
-		out.push(`   ${padR("MCP · this month", 16)}${padL(`${mcp.percentage || 0}%`, 5)}  ${meter(mcp.percentage, 22)}  ${reset}`);
+		out.push([
+			dim(`   ${padR("MCP · this month", 16)}`),
+			bold(padL(`${mcp.percentage || 0}%`, 5)),
+			plain(`  ${meter(mcp.percentage, 22)}  ${reset}`)
+		]);
 		const parts = (mcp.usageDetails ?? []).map((d) => `${friendlyTool(d.modelCode)} ${fmtNum(d.usage)}`).join("  ");
-		if (parts) out.push(`   ${fmtNum(mcp.currentUsage)}M / ${fmtNum(mcp.totol)}M · ${parts}`);
+		if (parts) out.push([plain(`   ${fmtNum(mcp.currentUsage)}M / ${fmtNum(mcp.totol)}M · ${parts}`)]);
+	}
+	out.push([]);
+	out.push([plain(rule())]);
+	return out;
+}
+function render(input) {
+	return reportLines(input).map((line) => line.map((seg) => seg.text).join("")).join("\n");
+}
+//#endregion
+//#region src/resolve.ts
+var DEFAULT_BASE_URL = "https://api.z.ai";
+var GLM_HOSTS = /* @__PURE__ */ new Set([
+	"api.z.ai",
+	"dev.bigmodel.cn",
+	"open.bigmodel.cn"
+]);
+function fromEnv(value) {
+	return value ? value : void 0;
+}
+function originOf(value) {
+	try {
+		const url = new URL(value);
+		return url.protocol === "https:" || url.protocol === "http:" ? url.origin : void 0;
+	} catch {
+		return;
+	}
+}
+function isGlm(origin) {
+	return GLM_HOSTS.has(new URL(origin).hostname);
+}
+function resolveBaseUrl(flag, env) {
+	if (flag !== void 0) {
+		const origin = originOf(flag);
+		if (origin === void 0) throw new Error(`invalid base URL: ${flag}`);
+		return {
+			glm: isGlm(origin),
+			origin
+		};
 	}
-	out.push("");
-	out.push(rule());
-	return out.join("\n");
+	const inherited = fromEnv(env.ANTHROPIC_BASE_URL);
+	const origin = inherited === void 0 ? void 0 : originOf(inherited);
+	if (origin !== void 0 && isGlm(origin)) return {
+		glm: true,
+		origin
+	};
+	return {
+		glm: false,
+		origin: DEFAULT_BASE_URL
+	};
+}
+function buildProgram() {
+	return new Command().name("zai-usage").addOption(new Option("--auth-token <token>", "API key").env("ZAI_AUTH_TOKEN")).addOption(new Option("--base-url <url>", "base URL").env("ZAI_BASE_URL")).addOption(new Option("--json", "print the report lines as JSON for the zai mod"));
+}
+function parseArgs(args) {
+	const program = parseQuietly(buildProgram(), args);
+	if (program === void 0) return;
+	const { authToken, baseUrl, json } = program.opts();
+	return {
+		authToken: authToken || void 0,
+		baseUrl: baseUrl || void 0,
+		json: json === true
+	};
+}
+function resolveConfig(env, parsed) {
+	const base = resolveBaseUrl(parsed.baseUrl, env);
+	const token = parsed.authToken ?? (base.glm ? fromEnv(env.ANTHROPIC_AUTH_TOKEN) : void 0);
+	if (token === void 0) 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
+	};
 }
 //#endregion
 //#region src/usage.ts
 /**
-* Usage query. Fetches model usage, tool usage, and quota limits from the GLM
-* Coding Plan monitor API of the resolved base URL (paths are fixed) and
-* renders the plain-text report.
+* Usage query: the three GLM monitor endpoints of the resolved base URL,
+* assembled into the report input.
 */
 function queryParams(now = /* @__PURE__ */ new Date()) {
 	const startDate = new Date(now.getFullYear(), now.getMonth(), now.getDate() - 1, now.getHours(), 0, 0, 0);
@@ -555,12 +535,12 @@ async function fetchReport(config) {
 			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 render({
+		return {
 			platform,
 			model,
 			tool,
 			quota: processQuotaLimit(quotaRaw)
-		});
+		};
 	} catch (e) {
 		throw new Error(`Request failed: ${e.message}`, { cause: e });
 	}
@@ -568,7 +548,13 @@ async function fetchReport(config) {
 //#endregion
 //#region src/index.ts
 var parsed = parseArgs(process.argv.slice(2)) ?? printUsageAndExit(buildProgram());
-await hookOrPrint(parsed.hook, "query failed", () => fetchReport(resolveConfig(process.env, parsed)));
+try {
+	const input = await fetchReport(resolveConfig(process.env, parsed));
+	console.log(parsed.json ? JSON.stringify({ lines: reportLines(input) }) : render(input));
+} catch (e) {
+	console.error(e.message);
+	process.exit(1);
+}
 //#endregion
 export {};