dsh-livebench-panel 0.1.10 → 0.1.11

package/lib/index.js

@@ -502,6 +502,22 @@ function asInt(value, min, max, fallback) {
 	return Math.min(max, Math.max(min, Math.trunc(n)));
 }
 
+/**
+ * Validate one bench path: "live_bench", "live_bench/<category>" or
+ * "live_bench/<category>/<task>". Task names are mixed-case in LiveBench
+ * (AMPS_Hard, LCB_generation, …), so segments allow A-Za-z0-9_.
+ * @returns {string|null} error description, or null when valid.
+ */
+function validateBenchName(bn) {
+	const parts = bn.split("/").filter((s) => s.length > 0);
+	if (parts[0] !== "live_bench") return "must start with live_bench";
+	if (parts.length > 3) return "too deep";
+	for (const seg of parts) {
+		if (!/^[A-Za-z0-9_]{1,80}$/.test(seg)) return `bad segment "${seg}"`;
+	}
+	return null;
+}
+
 function apply(ctx) {
 	/** The single run slot. */
 	let run = null;
@@ -545,14 +561,9 @@ function apply(ctx) {
 			benchNames = [];
 			for (const bn of body.benchNames) {
 				if (typeof bn !== "string") return { status: 400, payload: { ok: false, error: "invalid bench name" } };
-				const parts = bn.split("/").filter((s) => s.length > 0);
-				if (parts[0] !== "live_bench" || parts.length > 3) {
-					return { status: 400, payload: { ok: false, error: `invalid bench path: ${bn}` } };
-				}
-				for (const seg of parts) {
-					if (!/^[a-z0-9_]+$/.test(seg)) return { status: 400, payload: { ok: false, error: `invalid bench path: ${bn}` } };
-				}
-				benchNames.push(parts.join("/"));
+				const error = validateBenchName(bn);
+				if (error) return { status: 400, payload: { ok: false, error: `invalid bench path: ${bn} (${error})` } };
+				benchNames.push(bn.split("/").filter((s) => s.length > 0).join("/"));
 			}
 		}
 
@@ -886,7 +897,7 @@ function apply(ctx) {
 				const task = typeof row.task === "string" ? row.task : "";
 				// path-safety: membership in the scanned task list + charset guards
 				if (!/^[A-Za-z0-9._@-]{1,160}$/.test(model)) continue;
-				if (!/^[a-z0-9_]{1,60}$/.test(category) || !/^[a-z0-9_]{1,60}$/.test(task)) continue;
+				if (!/^[A-Za-z0-9_]{1,80}$/.test(category) || !/^[A-Za-z0-9_]{1,80}$/.test(task)) continue;
 				if (!(tasks[category] && tasks[category][task])) continue;
 				const taskDir = join(layout.dataDir, category, task);
 				removed += filterJsonlByModel(join(taskDir, "model_answer", `${model}.jsonl`), "model_id", model);
@@ -902,4 +913,4 @@ function runsAlive() {
 	return 1;
 }
 
-export { name, inject, apply, writeGeneratedModelConfig, readProviders };
+export { name, inject, apply, writeGeneratedModelConfig, readProviders, validateBenchName };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-livebench-panel",
-  "version": "0.1.10",
+  "version": "0.1.11",
   "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",