@tea-agent/loop-agent 0.20.1 → 0.21.0

package/dist/worker/loop-agent/loop-agent-client.js

@@ -206,11 +206,12 @@ export class LoopAgentClient {
         const stdoutPath = path.join(artifactDir, "stdout.txt");
         const stderrPath = path.join(artifactDir, "stderr.txt");
         const resultPath = path.join(artifactDir, "result.json");
+        const configuredTimeoutMs = options.timeoutMs ?? this.defaultTimeoutMs;
         const { stdout, stdoutBytes, stdoutTruncated, stderr, stderrBytes, stderrTruncated, exitCode, timedOut, } = await spawnCommand({
             command,
             args: commandArgs,
             cwd: options.cwd,
-            timeoutMs: options.timeoutMs ?? this.defaultTimeoutMs,
+            hardTimeoutMs: configuredTimeoutMs > 0 ? configuredTimeoutMs : undefined,
             env: { ...process.env, ...this.env, ...options.env },
             stdoutPath, stderrPath, resultPath,
             heartbeatIntervalMs: this.heartbeatIntervalMs,
@@ -362,23 +363,56 @@ function spawnCommand(input) {
         let stdout = createBoundedOutput();
         let stderr = createBoundedOutput();
         let timedOut = false;
-        const timeout = setTimeout(() => { timedOut = true; child.kill("SIGTERM"); }, input.timeoutMs);
+        let settled = false;
+        const settle = (payload) => {
+            if (settled)
+                return;
+            settled = true;
+            clearTimers();
+            resolve({
+                stdout: formatBoundedOutput(stdout),
+                stdoutBytes: stdout.bytes,
+                stdoutTruncated: stdout.truncated,
+                stderr: formatBoundedOutput(stderr),
+                stderrBytes: stderr.bytes,
+                stderrTruncated: stderr.truncated,
+                exitCode: payload.exitCode,
+                timedOut: payload.timedOut,
+            });
+        };
+        let hardTimeout;
+        if (input.hardTimeoutMs !== undefined && input.hardTimeoutMs > 0) {
+            hardTimeout = setTimeout(() => {
+                timedOut = true;
+                child.kill("SIGTERM");
+            }, input.hardTimeoutMs);
+        }
         let heartbeatTimer;
         if (input.heartbeatIntervalMs > 0) {
             heartbeatTimer = setInterval(() => { invokeHeartbeatCallback(input.onHeartbeat, startedAtMs); }, input.heartbeatIntervalMs);
         }
-        const clearTimers = () => { clearTimeout(timeout); if (heartbeatTimer !== undefined) {
-            clearInterval(heartbeatTimer);
-            heartbeatTimer = undefined;
-        } };
+        const clearTimers = () => {
+            if (hardTimeout !== undefined)
+                clearTimeout(hardTimeout);
+            if (heartbeatTimer !== undefined) {
+                clearInterval(heartbeatTimer);
+                heartbeatTimer = undefined;
+            }
+        };
         child.stdout.setEncoding("utf8");
         child.stderr.setEncoding("utf8");
         child.stdout.on("data", (chunk) => { stdout = appendBoundedOutput(stdout, chunk); appendChunkBestEffort(input.stdoutPath, chunk); invokeChunkCallback(input.onStdout, chunk); });
         child.stderr.on("data", (chunk) => { stderr = appendBoundedOutput(stderr, chunk); appendChunkBestEffort(input.stderrPath, chunk); invokeChunkCallback(input.onStderr, chunk); });
-        child.on("error", (error) => { clearTimers(); reject(error); });
-        child.on("close", (exitCode) => {
+        child.on("error", (error) => { if (!settled) {
+            settled = true;
             clearTimers();
-            resolve({ stdout: formatBoundedOutput(stdout), stdoutBytes: stdout.bytes, stdoutTruncated: stdout.truncated, stderr: formatBoundedOutput(stderr), stderrBytes: stderr.bytes, stderrTruncated: stderr.truncated, exitCode, timedOut });
+            reject(error);
+        } });
+        child.on("close", (exitCode) => {
+            settle({
+                exitCode,
+                timedOut,
+            });
         });
     });
 }

package/dist/worker/observability/read-model.js

@@ -717,6 +717,8 @@ function computeHealth(batches, tasks, dagRuns) {
 const ACTIVE_EFFECTIVE_STATUSES = new Set([
     "running",
     "running-quiet",
+    "running-suspected-stall",
+    "needs-attention",
     "remote-unknown",
     "pending",
 ]);
@@ -796,8 +798,12 @@ function computeDagHealth(dagRuns) {
             pausedRuns++;
         if (dag.effectiveStatus === "failed")
             failedRuns++;
-        if (liveness === "stale" || liveness === "orphaned")
+        if (liveness === "stale"
+            || liveness === "orphaned"
+            || liveness === "suspected-stall"
+            || liveness === "needs-attention") {
             staleRuns++;
+        }
         if (dag.effectiveStatus === "interrupted")
             interruptedRuns++;
         if (dag.stateConsistent === false)

package/dist/worker/observe/static/constants.js

@@ -75,6 +75,9 @@ export const LIVENESS_LABELS = {
 	active: "执行器正常",
 	"node-quiet": "执行器正常，节点长时间无活动",
 	quiet: "无输出（存活）",
+	"suspected-stall": "疑似卡死",
+	"needs-attention": "需人工关注",
+	"needs-reconcile": "需对账",
 	stale: "执行器心跳中断",
 	"timeout-risk": "即将超时",
 	orphaned: "执行器已退出",
@@ -86,6 +89,8 @@ export const DAG_EFFECTIVE_STATUS_LABELS = {
 	pending: "等待执行",
 	running: "正在执行",
 	"running-quiet": "运行可疑",
+	"running-suspected-stall": "疑似卡死",
+	"needs-attention": "需人工关注",
 	paused: "已暂停",
 	interrupted: "执行已中断",
 	"remote-unknown": "远端状态未知",

package/dist/worker/observe/static/format-pool.js

@@ -71,7 +71,18 @@ export function isPoolAttentionTask(task) {
 	const st = (task?.status ?? "").toLowerCase();
 	const live = (task?.liveness ?? "").toLowerCase();
 	if (["failed", "error", "blocked", "stale"].includes(st)) return true;
-	if (["stale", "orphaned", "timeout-risk"].includes(live)) return true;
+	if (
+		[
+			"stale",
+			"orphaned",
+			"timeout-risk",
+			"suspected-stall",
+			"needs-attention",
+			"needs-reconcile",
+		].includes(live)
+	) {
+		return true;
+	}
 	return false;
 }
 
@@ -79,8 +90,16 @@ export function attentionRank(task) {
 	const st = (task?.status ?? "").toLowerCase();
 	const live = (task?.liveness ?? "").toLowerCase();
 	if (st === "failed" || st === "error") return 0;
-	if (live === "timeout-risk") return 1;
-	if (st === "stale" || live === "stale" || live === "orphaned") return 2;
+	if (live === "timeout-risk" || live === "needs-attention") return 1;
+	if (
+		st === "stale"
+		|| live === "stale"
+		|| live === "orphaned"
+		|| live === "suspected-stall"
+		|| live === "needs-reconcile"
+	) {
+		return 2;
+	}
 	if (st === "blocked") return 3;
 	return 9;
 }

package/dist/worker/observe/static/styles.css

@@ -845,10 +845,14 @@ td:first-child,
 .kpi-tip {
 	position: absolute;
 	left: 12px;
-	right: 12px;
+	right: auto;
 	top: calc(100% - 8px);
 	z-index: 10;
-	max-width: 280px;
+	/* Stable readable width independent of narrow KPI cards; viewport-safe cap. */
+	width: min(280px, calc(100vw - 24px));
+	min-width: min(240px, calc(100vw - 24px));
+	max-width: min(280px, calc(100vw - 24px));
+	box-sizing: border-box;
 	padding: 8px 10px;
 	border: 1px solid var(--hairline-strong);
 	border-radius: var(--radius-md);
@@ -863,6 +867,12 @@ td:first-child,
 	pointer-events: none;
 	transition: opacity var(--transition-fast);
 }
+/* Five-column default: right two columns open inward (left) to avoid viewport overflow. */
+.kpi-grid:not(.pool-health-grid) > .kpi-card:nth-child(5n-1) .kpi-tip,
+.kpi-grid:not(.pool-health-grid) > .kpi-card:nth-child(5n) .kpi-tip {
+	left: auto;
+	right: 12px;
+}
 .kpi-help:hover + .kpi-tip,
 .kpi-help:focus-visible + .kpi-tip {
 	opacity: 1;
@@ -2015,7 +2025,7 @@ body.is-resizing-dag-graph {
 .spec-evidence-back { display: inline-flex; align-items: center; gap: 6px; padding: 6px 12px; border: 1px solid var(--hairline); border-radius: var(--radius-pill); background: var(--surface); color: var(--ink); font-size: 12px; font-weight: 650; cursor: pointer; }
 .spec-evidence-back:hover, .spec-evidence-back:focus-visible { background: var(--orange-soft); border-color: var(--orange); color: var(--orange-active); outline: none; }
 .spec-evidence-detail-meta { display: flex; flex-wrap: wrap; align-items: center; gap: 8px; font-size: 12px; color: var(--body); }
-.spec-evidence-detail-content { min-height: 0; max-width: 100%; margin: 0; padding: var(--space-3); border: 1px solid var(--hairline); border-radius: var(--radius-md); background: var(--canvas-soft); white-space: pre-wrap; overflow-wrap: anywhere; font-size: 12px; line-height: 1.6; }
+.spec-evidence-detail-content { min-height: 0; max-width: 100%; margin: 0; padding: var(--space-3); border: 1px solid var(--hairline); border-radius: var(--radius-md); background: var(--canvas-soft); overflow-wrap: anywhere; font-size: 12px; line-height: 1.6; }
 .spec-evidence-detail-error { margin: 0; color: var(--red, #c00); font-size: 12px; }
 .spec-evidence-detail-truncated { margin: 0; color: var(--muted); font-size: 11px; }
 
@@ -2289,6 +2299,16 @@ body.is-resizing-dag-graph {
 	.kpi-grid.pool-health-grid .kpi-card:first-child {
 		border-left: 0;
 	}
+	/* Three-column: reset five-column tip edges, then rightmost column opens inward. */
+	.kpi-grid:not(.pool-health-grid) > .kpi-card:nth-child(5n-1) .kpi-tip,
+	.kpi-grid:not(.pool-health-grid) > .kpi-card:nth-child(5n) .kpi-tip {
+		left: 12px;
+		right: auto;
+	}
+	.kpi-grid:not(.pool-health-grid) > .kpi-card:nth-child(3n) .kpi-tip {
+		left: auto;
+		right: 12px;
+	}
 	.panel-risk {
 		position: static;
 	}
@@ -2378,6 +2398,15 @@ body.is-resizing-dag-graph {
 	.kpi-grid.pool-health-grid .kpi-card:first-child {
 		border-left: 0;
 	}
+	/* Two-column: reset three-column tip edges, then right column opens inward. */
+	.kpi-grid:not(.pool-health-grid) > .kpi-card:nth-child(3n) .kpi-tip {
+		left: 12px;
+		right: auto;
+	}
+	.kpi-grid:not(.pool-health-grid) > .kpi-card:nth-child(2n) .kpi-tip {
+		left: auto;
+		right: 12px;
+	}
 	.dag-inspector {
 		top: 12px;
 		right: 12px;

package/dist/worker/observe/static/views/dag-inspector.js

@@ -515,8 +515,8 @@ function renderSpecEvidenceDetail(content, dagRunId, nodeId) {
 		error.setAttribute("role", "alert");
 		wrapper.appendChild(error);
 	} else if (detail.data) {
-		const body = el("pre", "spec-evidence-detail-content");
-		body.textContent = detail.data.content ?? "";
+		const body = el("div", "spec-evidence-detail-content");
+		body.appendChild(renderMarkdown(detail.data.content ?? ""));
 		wrapper.appendChild(body);
 		if (detail.data.truncated) {
 			const note = el(

package/dist/worker/run-task/run-task.js

@@ -10,7 +10,8 @@ import { getTaskPoolRoot } from "../pool/run-store.js";
 import { checkRequiredOutputs } from "../outcomes/gate.js";
 import { projectOutcome } from "../outcomes/projector.js";
 import { writeOutcome } from "../outcomes/store.js";
-export const DEFAULT_RUN_DAG_TIMEOUT_MS = 1_800_000;
+/** No outer wall-clock by default; the DAG kernel owns liveness supervision. */
+export const DEFAULT_RUN_DAG_TIMEOUT_MS = 0;
 export const MAX_WORKER_TIMEOUT_MS = 7_200_000;
 export async function runTaskSpec(options) {
     let controllerIdentity = resolveControllerIdentity(options.client, options.controllerIdentity);
@@ -346,6 +347,11 @@ async function runRequiredCommand(repoRoot, client, artifactName, args, expectJs
  * not know the flag; their parser rejects before any DAG node can execute, so
  * retrying the exact command without it is safe and preserves the same run id.
  */
+function assertRunDagCommandOk(result, args) {
+    if (!result.ok) {
+        throw new Error(`loop-agent command failed: ${args.join(" ")}`);
+    }
+}
 async function runDagWithEventsFallback(repoRoot, client, args, timeoutMs, eventCtx) {
     const result = await client.run(args, {
         cwd: repoRoot,
@@ -356,7 +362,7 @@ async function runDagWithEventsFallback(repoRoot, client, args, timeoutMs, event
     if (result.ok)
         return result;
     if (!rejectsEventsJsonl(result) || !args.includes("--events-jsonl")) {
-        throw new Error(`loop-agent command failed: ${args.join(" ")}`);
+        assertRunDagCommandOk(result, args);
     }
     const fallbackArgs = withoutFlagAndValue(args, "--events-jsonl");
     eventCtx.progress.note("run-dag controller does not support --events-jsonl; retrying without DAG event stream");
@@ -366,9 +372,7 @@ async function runDagWithEventsFallback(repoRoot, client, args, timeoutMs, event
         expectJson: true,
         timeoutMs,
     });
-    if (!fallback.ok) {
-        throw new Error(`loop-agent command failed: ${fallbackArgs.join(" ")}`);
-    }
+    assertRunDagCommandOk(fallback, fallbackArgs);
     return fallback;
 }
 function rejectsEventsJsonl(result) {
@@ -424,8 +428,14 @@ async function runObservedStep(ctx, stepName, run, options) {
         throw error;
     }
 }
+/**
+ * Explicit worker.timeout_ms remains an operator hard limit. Omission disables
+ * the Worker wall clock so the DAG kernel can supervise active nodes.
+ */
 function resolveRunDagTimeoutMs(taskSpec) {
-    return Math.min(taskSpec.worker.timeout_ms ?? DEFAULT_RUN_DAG_TIMEOUT_MS, MAX_WORKER_TIMEOUT_MS);
+    return taskSpec.worker.timeout_ms === undefined
+        ? DEFAULT_RUN_DAG_TIMEOUT_MS
+        : Math.min(taskSpec.worker.timeout_ms, MAX_WORKER_TIMEOUT_MS);
 }
 async function collectFailureArtifacts(input) {
     const doctor = await runObservedStep(input.eventCtx, "dag-doctor", () => input.client.run(["dag", "doctor", "--run-id", input.workerRunId, "--markdown"], {