dsh-livebench-panel 0.1.11 → 0.1.13

package/lib/client.js

@@ -92,11 +92,12 @@ window.__ModuleLoader__.load({
 		function LiveBenchView() {
 			const [config, setConfig] = useState(null);
 			const [configError, setConfigError] = useState(null);
-			const [sel, setSel] = useState({ provider: "", model: "", reasoning: "default", cats: [], tasks: [], release: "2024-11-25", begin: "", end: "", maxTokens: "32000" });
+			const [sel, setSel] = useState({ models: [], reasoning: "default", cats: [], tasks: [], release: "2024-11-25", begin: "", end: "", maxTokens: "32000" });
+			const [modelDdOpen, setModelDdOpen] = useState(false);
+			const modelDdRef = useRef(null);
 			const [busy, setBusy] = useState(false);
 			const [running, setRunning] = useState(false);
-			const [log, setLog] = useState("");
-			const [exitCode, setExitCode] = useState(null);
+			const [runsList, setRunsList] = useState([]);
 			const [startError, setStartError] = useState(null);
 			const [homeInput, setHomeInput] = useState("");
 			const [homeBusy, setHomeBusy] = useState(false);
@@ -108,6 +109,16 @@ window.__ModuleLoader__.load({
 			const dragKey = useRef(null);
 			const logRef = useRef(null);
 
+			// 点击面板外部时关闭模型下拉
+			useEffect(() => {
+				if (!modelDdOpen) return;
+				const onDocDown = (event) => {
+					if (modelDdRef.current && !modelDdRef.current.contains(event.target)) setModelDdOpen(false);
+				};
+				document.addEventListener("mousedown", onDocDown);
+				return () => document.removeEventListener("mousedown", onDocDown);
+			}, [modelDdOpen]);
+
 			const loadConfig = useCallback(async () => {
 				const payload = await api("/config");
 				if (payload.ok) {
@@ -115,14 +126,12 @@ window.__ModuleLoader__.load({
 					setConfigError(null);
 					setSel((prev) => {
 						const next = { ...prev, release: payload.releases.includes(prev.release) ? prev.release : "2024-11-25" };
-						const provider = payload.providers.find((p) => p.id === prev.provider) ?? payload.providers.find((p) => p.models.length > 0) ?? null;
-						if (provider) {
-							next.provider = provider.id;
-							if (!provider.models.some((m) => m.id === prev.model)) {
-								next.model = provider.models[0]?.id ?? "";
-							}
-						}
-						if (!payload.releases.includes(next.release)) next.release = payload.releases[0] ?? "2024-11-25";
+						// 保留仍存在的选择，剔除失效项
+						next.models = prev.models.filter((v) => {
+							const [pid] = v.split("::");
+							const p = payload.providers.find((x) => x.id === pid);
+							return p ? true : false;
+						});
 						return next;
 					});
 				} else {
@@ -138,11 +147,8 @@ window.__ModuleLoader__.load({
 			const refreshStatus = useCallback(async () => {
 				const payload = await api("/status");
 				if (payload.ok) {
+					setRunsList(payload.runs ?? []);
 					setRunning(payload.running === true);
-					if (payload.hasRun) {
-						setLog(payload.log ?? "");
-						setExitCode(payload.exitCode);
-					}
 					return payload.running === true;
 				}
 				return false;
@@ -165,10 +171,6 @@ window.__ModuleLoader__.load({
 				return () => clearInterval(timer);
 			}, [refreshStatus, loadResults]);
 
-			useEffect(() => {
-				if (logRef.current) logRef.current.scrollTop = logRef.current.scrollHeight;
-			}, [log]);
-
 			const categories = useMemo(() => (config ? Object.keys(config.tasks) : []), [config]);
 			// 有效题数：LiveBench 会丢弃「发布晚于所选 release」和「在所选 release
 			// 前已退役」的题目，这里按题目桶 (发布日, 移除日) 精确复算。
@@ -197,21 +199,31 @@ window.__ModuleLoader__.load({
 				const next = list.includes(value) ? list.filter((v) => v !== value) : [...list, value];
 				return { ...prev, [key]: next };
 			});
-			const selectedProvider = useMemo(
-				() => (config ? config.providers.find((p) => p.id === sel.provider) ?? null : null),
-				[config, sel.provider],
-			);
-			const selectedModel = useMemo(
-				() => selectedProvider?.models.find((m) => m.id === sel.model) ?? null,
-				[selectedProvider, sel.model],
-			);
+			// 已选模型（provider::model 值列表）的解析与 effort 并集
+			const selectedModelEntries = useMemo(() => sel.models.map((value) => {
+				const [pid, ...rest] = value.split("::");
+				const mid = rest.join("::");
+				const provider = config?.providers.find((p) => p.id === pid) ?? null;
+				const model = provider?.models.find((m) => m.id === mid) ?? null;
+				return { providerId: pid, modelId: mid, efforts: model?.efforts ?? [] };
+			}).filter((entry) => entry.modelId !== ""), [sel.models, config]);
 			const effortOptions = useMemo(() => {
-				const efforts = selectedModel?.efforts ?? [];
-				return efforts.map((e) => ({ value: e, label: e === "off" ? "off（关闭思考）" : e }));
-			}, [selectedModel]);
+				const seen = new Set();
+				const union = [];
+				for (const entry of selectedModelEntries) {
+					for (const e of entry.efforts) {
+						if (!seen.has(e)) { seen.add(e); union.push({ value: e, label: e === "off" ? "off（关闭思考）" : e }); }
+					}
+				}
+				return union;
+			}, [selectedModelEntries]);
 
 			const onStart = async () => {
 				setStartError(null);
+				if (selectedModelEntries.length === 0) {
+					setStartError("请先在下拉框中至少选择一个模型。");
+					return;
+				}
 				const benchNames = [];
 				if (sel.cats.length === 0) {
 					benchNames.push("live_bench");
@@ -227,22 +239,28 @@ window.__ModuleLoader__.load({
 				}
 				setBusy(true);
 				try {
-					const payload = await api("/start", {
-						method: "POST",
-						headers: { "content-type": "application/json" },
-						body: JSON.stringify({
-							provider: sel.provider,
-							model: sel.model,
-							reasoningEffort: sel.reasoning,
-							benchNames,
-							release: sel.release,
-							begin: sel.begin,
-							end: sel.end,
-							maxTokens: sel.maxTokens,
-						}),
+					// 每个 selected 模型并发启动一个评测（服务端有并发上限保护）
+					const failures = [];
+					const launches = selectedModelEntries.map(async (entry) => {
+						const payload = await api("/start", {
+							method: "POST",
+							headers: { "content-type": "application/json" },
+							body: JSON.stringify({
+								provider: entry.providerId,
+								model: entry.modelId,
+								reasoningEffort: sel.reasoning,
+								benchNames,
+								release: sel.release,
+								begin: sel.begin,
+								end: sel.end,
+								maxTokens: sel.maxTokens,
+							}),
+						});
+						if (!payload.ok) failures.push(`${entry.modelId}: ${payload.error ?? "启动失败"}`);
 					});
-					if (!payload.ok) setStartError(payload.error ?? "启动失败");
-					else await refreshStatus();
+					await Promise.all(launches);
+					if (failures.length > 0) setStartError(failures.join("；"));
+					await refreshStatus();
 				} catch (error) {
 					setStartError(String(error.message ?? error));
 				} finally {
@@ -275,14 +293,7 @@ window.__ModuleLoader__.load({
 
 			const setField = (key) => (event) => {
 				const value = event.target.value;
-				setSel((prev) => {
-					const next = { ...prev, [key]: value };
-					if (key === "provider") {
-						const provider = config?.providers.find((p) => p.id === value) ?? null;
-						next.model = provider?.models[0]?.id ?? "";
-					}
-					return next;
-				});
+				setSel((prev) => ({ ...prev, [key]: value }));
 			};
 
 			// ---- 成绩矩阵：model 为行标识、category/task 为列标识 ----
@@ -385,18 +396,48 @@ window.__ModuleLoader__.load({
 				configError !== null && h("p", { className: c("error") }, `加载配置失败：${configError}`),
 				h("div", { className: c("card") },
 					h("div", { className: c("grid") },
-						h("div", { className: c("field") },
-							h("label", { className: c("label") }, "模型（harness 全部模型）"),
-							h("select", { className: c("select"), value: sel.provider + "::" + sel.model, onChange: (event) => {
-								const [providerId, ...rest] = event.target.value.split("::");
-								const modelId = rest.join("::");
-								setSel((prev) => ({ ...prev, provider: providerId, model: modelId, reasoning: "default" }));
-							} }, modelOptions(config?.providers ?? [])),
+						h("div", { className: c("field"), ref: modelDdRef, style: { position: "relative" } },
+							h("label", { className: c("label") }, "模型（harness 全部模型，可多选）"),
+							h("button", {
+								type: "button", className: c("select"), style: { textAlign: "left", width: "100%", overflow: "hidden", textOverflow: "ellipsis", whiteSpace: "nowrap" },
+								onClick: () => setModelDdOpen((v) => !v),
+							}, sel.models.length === 0
+								? "点击选择模型…"
+								: sel.models.length === 1
+									? sel.models[0].split("::")[1]
+									: `已选 ${sel.models.length} 个模型`),
+							modelDdOpen && h("div", {
+								style: {
+									position: "absolute", zIndex: 30, top: "calc(100% + 4px)", left: 0, right: 0,
+									maxHeight: "320px", overflowY: "auto", border: "1px solid var(--dsw-alias-border-l2)",
+									borderRadius: "8px", background: "var(--dsw-alias-bg-layer-1)", padding: "6px",
+									boxShadow: "var(--dsw-shadow-lv1, 0 4px 16px rgba(0,0,0,.25))",
+								},
+							},
+								(config?.providers ?? []).map((p) => h("div", { key: p.id, style: { marginBottom: "4px" } },
+									h("div", { style: { fontSize: "11px", color: "var(--dsw-alias-label-tertiary)", padding: "3px 4px" } },
+										p.name + (p.routable ? "" : " ·未接LiveBench")),
+									p.models.map((m) => {
+										const value = p.id + "::" + m.id;
+										const checked = sel.models.includes(value);
+										return h("label", {
+											key: value,
+											style: { display: "flex", alignItems: "center", gap: "6px", padding: "3px 4px", cursor: "pointer", borderRadius: "6px" },
+										},
+											h("input", { type: "checkbox", checked, onChange: () => {
+												setSel((prev) => {
+													const set = new Set(prev.models);
+													if (set.has(value)) set.delete(value); else set.add(value);
+													return { ...prev, models: [...set] };
+												});
+											} }),
+											h("span", { style: { fontSize: "12.5px" } }, m.name));
+									}))),
 						),
 						h("div", { className: c("field") },
 							h("label", { className: c("label") }, "推理强度"),
 							h("select", { className: c("select"), value: sel.reasoning, onChange: setField("reasoning"), disabled: effortOptions.length === 0 },
-								h("option", { value: "default" }, effortOptions.length === 0 ? "（该模型不支持）" : "（模型默认）"),
+								h("option", { value: "default" }, effortOptions.length === 0 ? "（按模型默认）" : "（模型默认）"),
 								effortOptions.map((e) => h("option", { key: e.value, value: e.value }, e.label))),
 						),
 						h("div", { className: c("field") },
@@ -443,22 +484,25 @@ window.__ModuleLoader__.load({
 						),
 					),
 					h("div", { className: c("row") },
-						h("button", { className: c("btn"), onClick: onStart, disabled: busy || running || !config?.available || sel.model === "" },
-							running ? "评测运行中…" : busy ? "启动中…" : "开始评测"),
-						running && h("button", { className: c("btn") + " " + c("btnDanger"), onClick: onStop }, "停止"),
+						h("button", { className: c("btn"), onClick: onStart, disabled: busy || running || !config?.available || sel.models.length === 0 },
+							running ? `评测运行中（可另起）…` : busy ? "启动中…" : `开始评测（${sel.models.length} 个模型并发）`),
+						running && h("button", { className: c("btn") + " " + c("btnDanger"), onClick: onStop }, "停止全部"),
 						h("button", { className: c("btnGhost") + " " + c("btn"), onClick: () => { loadConfig(); loadResults(); } }, "刷新"),
-						selectedProvider && !selectedProvider.routable && h("span", { className: c("hint") },
-							"该 provider 的协议或端点未知，无法自动路由：LiveBench 将按模型名原生尝试，未注册的模型名会失败。"),
 					),
 					startError !== null && h("p", { className: c("error") }, startError),
 				),
-				(log.length > 0 || running) && h("div", { className: c("card") },
+				runsList.length > 0 && h("div", { className: c("card") },
 					h("div", { className: c("row") },
 						h("span", { className: c("badge"), "data-ok": running ? "1" : "0" },
-							running ? "运行中" : exitCode === 0 ? "已完成" : exitCode === null ? "待运行" : `已退出 (${exitCode})`),
+							running ? "有评测运行中" : "无运行中的评测"),
 					),
-					h("pre", { className: c("log"), ref: logRef }, log || "（暂无输出）"),
-				),
+					runsList.map((r) => h("details", { key: r.runId, open: r.running },
+						h("summary", { style: { cursor: "pointer", fontSize: "12px", color: "var(--dsw-alias-label-secondary)" } },
+							h("span", { className: c("badge"), "data-ok": r.running ? "1" : "0" }, r.running ? "运行中" : r.exitCode === 0 ? "完成" : `退出(${r.exitCode})`),
+							" ",
+							r.displayName),
+						h("pre", { className: c("log"), ref: r.running ? logRef : null }, r.log || "（暂无输出）"),
+					)))),
 				sortedModels.length > 0 && h("div", { className: c("card") },
 					h("div", { className: c("row") },
 						h("span", { className: c("label") }, "评测成绩（model 为行；分数 = 平均分 ×100；行首可拖动排序，顺序保存在本机）"),
@@ -498,8 +542,22 @@ window.__ModuleLoader__.load({
 									h("td", null, model),
 									matrix.tasks.map((taskKeyCol) => {
 										const row = matrix.cells.get(`${model}\u0000${taskKeyCol}`);
-										return h("td", { key: taskKeyCol, className: c("score") },
-											row ? `${row.score.toFixed(1)} (${row.judged}/${row.total})` : "—");
+										// ×= 该模型未做过此任务；E = 做了但全部 API 失败
+										let cellText = "×";
+										let title = "该模型未执行此任务";
+										if (row) {
+											if (row.answered > 0 && row.errors >= row.answered) {
+												cellText = `E (${row.errors})`;
+												title = "该任务全部回答均 API 失败（$ERROR$），计 0 分";
+											} else if (row.judged === 0) {
+												cellText = "未判分";
+												title = "有答案但尚未判分";
+											} else {
+												cellText = `${row.score.toFixed(1)} (${row.judged}/${row.total})`;
+												if (row.errors > 0) title += ` · 其中 ${row.errors} 题 API 失败计 0 分`;
+											}
+										}
+										return h("td", { key: taskKeyCol, className: c("score"), title }, cellText);
 									}),
 									h("td", null, h("button", {
 										className: c("btnGhost") + " " + c("btn"), title: "删除该模型的所有评测成绩",

package/lib/index.js

@@ -102,8 +102,8 @@ const RELEASES = [
 ];
 /** Sane cap so a runaway run cannot eat memory with its log. */
 const LOG_MAX_LINES = 600;
-/** One evaluation at a time — LiveBench grading is not cheap. */
-const MAX_CONCURRENT_RUNS = 1;
+/** One evaluation at a time per model, but several models may run concurrently. */
+const MAX_CONCURRENT_RUNS = 6;
 
 /** Resolve the profile directory from the config-tree anchor (plugin-market pattern). */
 function resolveProfileDir(ctx) {
@@ -434,6 +434,7 @@ function readJsonl(path) {
 function computeResults(dataDir) {
 	const judged = new Map(); // `${model}\u0000${task}` -> {model, category, task, sum, n, time}
 	const totals = new Map(); // task -> question count
+	const answers = new Map(); // `${model}\u0000${task}` -> {answered, errors}
 	if (!existsSync(dataDir)) return { rows: [], taskTotals: {} };
 	for (const category of readDirSafe(dataDir)) {
 		const catDir = join(dataDir, category);
@@ -441,6 +442,23 @@ function computeResults(dataDir) {
 			const taskDir = join(catDir, task);
 			const questions = readJsonl(join(taskDir, "question.jsonl"));
 			totals.set(task, questions.length);
+			// answer side: how many answers each model produced, and how many of
+			// them were $ERROR$ (API call failures) — distinguishes "didn't run"
+			// from "ran and failed" from "ran and scored".
+			const answerDir = join(taskDir, "model_answer");
+			for (const file of readDirSafe(answerDir)) {
+				if (!file.endsWith(".jsonl")) continue;
+				for (const line of readJsonl(join(answerDir, file))) {
+					const model = typeof line.model_id === "string" ? line.model_id : file.slice(0, -".jsonl".length);
+					const turns = line.choices?.[0]?.turns;
+					const text = Array.isArray(turns) ? String(turns[0] ?? "") : "";
+					const key = `${model}\u0000${task}`;
+					const stat = answers.get(key) ?? { answered: 0, errors: 0 };
+					stat.answered += 1;
+					if (text.startsWith("$ERROR$")) stat.errors += 1;
+					answers.set(key, stat);
+				}
+			}
 			const judgments = readJsonl(join(taskDir, "model_judgment", "ground_truth_judgment.jsonl"));
 			for (const row of judgments) {
 				const model = typeof row.model === "string" ? row.model : null;
@@ -456,15 +474,29 @@ function computeResults(dataDir) {
 			}
 		}
 	}
-	const rows = [...judged.values()].map((entry) => ({
-		model: entry.model,
-		category: entry.category,
-		task: entry.task,
-		score: entry.n > 0 ? (entry.sum / entry.n) * 100 : 0,
-		judged: entry.n,
-		total: totals.get(entry.task) ?? 0,
-		time: entry.time > 0 ? entry.time : null,
-	}));
+	const keys = new Set([...judged.keys(), ...answers.keys()]);
+	const rows = [...keys].map((key) => {
+		const entry = judged.get(key);
+		const stat = answers.get(key) ?? { answered: 0, errors: 0 };
+		const [model, task] = key.split("\u0000");
+		let category = entry?.category;
+		if (!category) {
+			for (const cat of readDirSafe(dataDir)) {
+				if (existsSync(join(dataDir, cat, task))) { category = cat; break; }
+			}
+		}
+		return {
+			model,
+			category: category ?? "",
+			task,
+			score: entry && entry.n > 0 ? (entry.sum / entry.n) * 100 : null,
+			judged: entry?.n ?? 0,
+			total: totals.get(task) ?? 0,
+			time: entry && entry.time > 0 ? entry.time : null,
+			answered: stat.answered,
+			errors: stat.errors,
+		};
+	});
 	const taskTotals = Object.fromEntries(totals);
 	return { rows, taskTotals };
 }
@@ -519,16 +551,17 @@ function validateBenchName(bn) {
 }
 
 function apply(ctx) {
-	/** The single run slot. */
-	let run = null;
+	/** All runs: runId -> record. Several evaluations may run concurrently. */
+	const runs = new Map();
 
 	const startRun = async (body) => {
 		const layout = livebenchLayout();
 		if (!layout.available) {
 			return { status: 409, payload: { ok: false, error: `LiveBench venv not found under ${layout.root}` } };
 		}
-		if (run && run.exitCode === null) {
-			return { status: 409, payload: { ok: false, error: "another evaluation is already running" } };
+		const runningCount = [...runs.values()].filter((r) => r.exitCode === null).length;
+		if (runningCount >= MAX_CONCURRENT_RUNS) {
+			return { status: 409, payload: { ok: false, error: `已有 ${runningCount} 个评测在并发运行（上限 ${MAX_CONCURRENT_RUNS}），请等待或停止部分评测` } };
 		}
 		const providerId = typeof body.provider === "string" ? body.provider : "";
 		const modelId = typeof body.model === "string" ? body.model.trim() : "";
@@ -567,11 +600,16 @@ function apply(ctx) {
 			}
 		}
 
-		const hasEffortSuffix = reasoningEffort !== null && reasoningEffort !== "off";
-		const displayName = displayModelName(providerId || "direct", modelId) + (hasEffortSuffix ? "@" + reasoningEffort : "");
+		// A global effort only applies to models whose own effort ladder
+		// includes it; others silently fall back to their default behaviour.
+		const modelMeta = provider?.models.find((m) => m.id === modelId) ?? null;
+		const effort = reasoningEffort !== null && modelMeta?.efforts?.length > 0 && !modelMeta.efforts.includes(reasoningEffort)
+			? null
+			: reasoningEffort;
+		const hasEffortSuffix = effort !== null && effort !== "off";
+		const displayName = displayModelName(providerId || "direct", modelId) + (hasEffortSuffix ? "@" + effort : "");
 		let cliModel = modelId;
-		let writeError = null;
-		// Anthropic-protocol proxies cannot go through --api-base (that path
+		let writeError = null;		// Anthropic-protocol proxies cannot go through --api-base (that path
 		// speaks OpenAI Chat Completions). Instead the generated model config
 		// selects LiveBench's native anthropic client and the spawn env points
 		// the SDK at the proxy (ANTHROPIC_BASE_URL / ANTHROPIC_API_KEY).
@@ -587,7 +625,7 @@ function apply(ctx) {
 			writeError = writeGeneratedModelConfig(layout, loadYaml(profileDir), {
 				displayName,
 				modelId,
-				reasoningEffort: isAnthropicRoute ? null : reasoningEffort,
+				reasoningEffort: isAnthropicRoute ? null : effort,
 				protocol: isAnthropicRoute ? "anthropic" : isOpenAIResponsesRoute ? "openai_responses" : "openai",
 			});
 			if (writeError) return { status: 500, payload: { ok: false, error: writeError } };
@@ -657,14 +695,21 @@ function apply(ctx) {
 		});
 
 		const record = {
-			id: `${Date.now()}`,
+			id: `${Date.now()}-${Math.random().toString(36).slice(2, 6)}`,
 			proc: child,
+			displayName,
 			exitCode: null,
 			startedAt: new Date().toISOString(),
 			command: [layout.pythonExe, ...args].join(" "),
 			log: [`$ ${[...args].join(" ")}`],
 		};
-		run = record;
+		runs.set(record.id, record);
+		// keep history bounded: drop oldest finished runs beyond 20 entries
+		if (runs.size > 20) {
+			for (const [id, r] of runs) {
+				if (r.exitCode !== null && runs.size > 20) runs.delete(id);
+			}
+		}
 
 		child.stdout.on("data", (chunk) => appendLog(record, chunk));
 		child.stderr.on("data", (chunk) => appendLog(record, chunk));
@@ -747,8 +792,9 @@ function apply(ctx) {
 				sendJson(res, 400, { ok: false, error: "invalid request body" });
 				return;
 			}
-			if (run !== null && run.exitCode === null && runsAlive() >= MAX_CONCURRENT_RUNS) {
-				sendJson(res, 409, { ok: false, error: "another evaluation is already running" });
+			const runningCount = [...runs.values()].filter((r) => r.exitCode === null).length;
+			if (runningCount >= MAX_CONCURRENT_RUNS) {
+				sendJson(res, 409, { ok: false, error: `已有 ${runningCount} 个评测在并发运行（上限 ${MAX_CONCURRENT_RUNS}）` });
 				return;
 			}
 			const result = await startRun(body);
@@ -764,20 +810,15 @@ function apply(ctx) {
 				sendJson(res, 405, { ok: false, error: "method not allowed" });
 				return;
 			}
-			if (run === null) {
-				sendJson(res, 200, { ok: true, running: false, hasRun: false });
-				return;
-			}
-			sendJson(res, 200, {
-				ok: true,
-				running: run.exitCode === null,
-				hasRun: true,
-				runId: run.id,
-				exitCode: run.exitCode,
-				startedAt: run.startedAt,
-				command: run.command,
-				log: run.log.slice(-120).join("\n"),
-			});
+			const list = [...runs.values()].reverse().map((r) => ({
+				runId: r.id,
+				displayName: r.displayName,
+				running: r.exitCode === null,
+				exitCode: r.exitCode,
+				startedAt: r.startedAt,
+				log: r.log.slice(-80).join("\n"),
+			}));
+			sendJson(res, 200, { ok: true, running: list.some((r) => r.running), runs: list });
 		},
 	}), `${name}: status route`);
 
@@ -793,24 +834,27 @@ function apply(ctx) {
 				sendJson(res, 403, { ok: false, error: "forbidden origin" });
 				return;
 			}
-			if (run === null || run.exitCode !== null) {
+			const running = [...runs.values()].filter((r) => r.exitCode === null);
+			if (running.length === 0) {
 				sendJson(res, 200, { ok: true, stopped: false });
 				return;
 			}
-			try {
-				if (process.platform === "win32") {
-					// proc.kill() only terminates run_livebench.py itself; its child
-					// (cmd → gen_api_answer.py → …) would survive. taskkill /T /F
-					// takes down the whole tree.
-					spawn("taskkill", ["/PID", String(run.proc.pid), "/T", "/F"], { windowsHide: true });
-				} else {
-					run.proc.kill();
+			for (const record of running) {
+				try {
+					if (process.platform === "win32") {
+						// proc.kill() only terminates run_livebench.py itself; its child
+						// (cmd → gen_api_answer.py → …) would survive. taskkill /T /F
+						// takes down the whole tree.
+						spawn("taskkill", ["/PID", String(record.proc.pid), "/T", "/F"], { windowsHide: true });
+					} else {
+						record.proc.kill();
+					}
+				} catch (error) {
+					sendJson(res, 500, { ok: false, error: String(error.message ?? error) });
+					return;
 				}
-			} catch (error) {
-				sendJson(res, 500, { ok: false, error: String(error.message ?? error) });
-				return;
 			}
-			sendJson(res, 200, { ok: true, stopped: true });
+			sendJson(res, 200, { ok: true, stopped: running.length });
 		},
 	}), `${name}: stop route`);
 
@@ -908,9 +952,4 @@ function apply(ctx) {
 	}), `${name}: delete route`);
 }
 
-/** Count alive runs (at most one today, kept for future parallel lanes). */
-function runsAlive() {
-	return 1;
-}
-
 export { name, inject, apply, writeGeneratedModelConfig, readProviders, validateBenchName };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-livebench-panel",
-  "version": "0.1.11",
+  "version": "0.1.13",
   "description": "DSH web plugin: a LiveBench tab in the Trajectory view (right of 对话/轨迹). Run LiveBench evaluations against every model configured in the DeepSeek Harness — pick provider/model, category, task, release and question range from dropdowns, watch progress, and read scores in place.",
   "license": "MIT",
   "type": "module",