@integrity-labs/agt-cli 0.28.461 → 0.28.462

package/dist/bin/agt.js

@@ -40,7 +40,7 @@ import {
   success,
   table,
   warn
-} from "../chunk-WCYGJANF.js";
+} from "../chunk-UNDFX52P.js";
 import {
   AnchorSessionClient,
   CHANNEL_REGISTRY,
@@ -4830,7 +4830,7 @@ import { execFileSync, execSync } from "child_process";
 import { existsSync as existsSync10, realpathSync as realpathSync2 } from "fs";
 import chalk18 from "chalk";
 import ora16 from "ora";
-var cliVersion = true ? "0.28.461" : "dev";
+var cliVersion = true ? "0.28.462" : "dev";
 async function fetchLatestVersion() {
   const host2 = getHost();
   if (!host2) return null;
@@ -6002,7 +6002,7 @@ function handleError(err) {
 }
 
 // src/bin/agt.ts
-var cliVersion2 = true ? "0.28.461" : "dev";
+var cliVersion2 = true ? "0.28.462" : "dev";
 var program = new Command();
 program.name("agt").description("Augmented CLI \u2014 agent provisioning and management").version(cliVersion2).option("--json", "Emit machine-readable JSON output (suppress spinners and colors)").option("--skip-update-check", "Skip the automatic update check on startup");
 program.hook("preAction", async (thisCommand, actionCommand) => {

package/dist/{chunk-WCYGJANF.js → chunk-UNDFX52P.js}

@@ -4911,7 +4911,7 @@ function requireHost() {
 }
 
 // src/lib/api-client.ts
-var agtCliVersion = true ? "0.28.461" : "dev";
+var agtCliVersion = true ? "0.28.462" : "dev";
 var lastConfigHash = null;
 function setConfigHash(hash) {
   lastConfigHash = hash && hash.length > 0 ? hash : null;
@@ -5401,13 +5401,24 @@ function recordStaleRotation(codeName, serverKeys, now = Date.now) {
     rotationTimestamps.set(rotationKey(codeName, key), ts);
   }
 }
-function wasRecentlyRotated(codeName, serverKey, withinMs, now = Date.now) {
+var ROTATION_TRACKING_RETENTION_MS = 15 * 60 * 1e3;
+function classifyRotation(codeName, serverKey, graceMs, now = Date.now, retentionMs = ROTATION_TRACKING_RETENTION_MS) {
   const k = rotationKey(codeName, serverKey);
   const ts = rotationTimestamps.get(k);
-  if (ts === void 0) return false;
-  if (now() - ts < withinMs) return true;
+  if (ts === void 0) return "none";
+  const age = now() - ts;
+  if (age < graceMs) return "within-grace";
+  if (age < retentionMs) return "expired";
   rotationTimestamps.delete(k);
-  return false;
+  return "none";
+}
+function getRotationRecordedAt(codeName, serverKey) {
+  return rotationTimestamps.get(rotationKey(codeName, serverKey));
+}
+function clearStaleRotation(codeName, serverKeys) {
+  for (const key of serverKeys) {
+    rotationTimestamps.delete(rotationKey(codeName, key));
+  }
 }
 function parseEnvIntegrationsEntries(content) {
   const entries = /* @__PURE__ */ new Map();
@@ -5706,6 +5717,9 @@ var BACKOFF_GIVE_UP_RESTARTS = 2 * MAX_PRESENCE_RESTART_ATTEMPTS;
 var BACKOFF_STABLE_RESET_MS = 5 * 6e4;
 var DEAD_CHANNEL_MS = 15 * 6e4;
 var presenceReaperState = /* @__PURE__ */ new Map();
+var MAX_ROTATION_PROMPT_PROBES = 3;
+var ROTATION_PROMPT_PROBE_INTERVAL_MS = 3e4;
+var rotationStallState = /* @__PURE__ */ new Map();
 function stateKey(codeName, serverKey) {
   return `${codeName}\0${serverKey}`;
 }
@@ -5718,6 +5732,7 @@ function clearPresenceReaperState(codeName) {
 function clearPresenceReaperStateForKeys(codeName, keys) {
   for (const key of keys) {
     presenceReaperState.delete(stateKey(codeName, key));
+    rotationStallState.delete(stateKey(codeName, key));
   }
 }
 function givenUpMcpServerKeys(codeName) {
@@ -5763,6 +5778,14 @@ function reapMissingMcpSessions(args) {
     rotationGraceMs = DEFAULT_ROTATION_GRACE_MS,
     onGiveUp,
     onRestart,
+    // ENG-8225: unwired callbacks make the prompt-probe path a strict no-op —
+    // `onRotationRespawnStalled` absent means "not armed", which falls through to
+    // the pre-ENG-8225 restart. Legacy callers and tests are unchanged.
+    onRotationRespawnStalled,
+    onRotationRespawnGaveUp,
+    onRotationRespawnOutcome,
+    maxRotationPromptProbes = MAX_ROTATION_PROMPT_PROBES,
+    rotationPromptProbeIntervalMs = ROTATION_PROMPT_PROBE_INTERVAL_MS,
     // ENG-5932: defaults make quarantine a strict no-op (classify everything
     // essential, mode off) so legacy callers/tests are unaffected.
     classifyKey = () => "essential",
@@ -5807,19 +5830,112 @@ function reapMissingMcpSessions(args) {
   const rows = parsePsRows(psOutput);
   const missingRaw = findMissingMcpServers({ rows, codeName, mcpJson, inContainer: isolated });
   const rotationGraced = [];
+  const rotationProbeArmed = [];
+  const rotationProbeWaiting = [];
+  const rotationProbeExhausted = [];
+  const rotationRespawnOutcomes = {};
+  const recordRotationOutcome = (key, outcome) => {
+    rotationRespawnOutcomes[key] = outcome;
+    if (!onRotationRespawnOutcome) return;
+    try {
+      onRotationRespawnOutcome(codeName, key, outcome);
+    } catch (err) {
+      log(
+        `[mcp-presence-reaper] onRotationRespawnOutcome callback threw for '${codeName}:${key}' (suppressed; reaper continues): ${err.message}`
+      );
+    }
+  };
   const missing = [];
   for (const key of missingRaw) {
-    if (wasRecentlyRotated(codeName, key, rotationGraceMs, now)) {
+    const rotation = classifyRotation(codeName, key, rotationGraceMs, now);
+    if (rotation === "within-grace") {
       rotationGraced.push(key);
-    } else {
+      continue;
+    }
+    if (rotation === "none") {
+      rotationStallState.delete(stateKey(codeName, key));
+      missing.push(key);
+      continue;
+    }
+    const sk = stateKey(codeName, key);
+    const rotationRecordedAt = getRotationRecordedAt(codeName, key) ?? now();
+    const priorStall = rotationStallState.get(sk);
+    const stall = priorStall && priorStall.rotationRecordedAt === rotationRecordedAt ? priorStall : {
+      rotationRecordedAt,
+      probeAttempts: 0,
+      lastProbeAt: null,
+      escalated: false
+    };
+    if (priorStall && priorStall !== stall) {
+      log(
+        `[mcp-presence-reaper] '${codeName}:${key}' discarding stall state from a previous rotation (rotation=${priorStall.rotationRecordedAt}, probes=${priorStall.probeAttempts}, escalated=${priorStall.escalated}) \u2014 this is a NEW rotation (rotation=${rotationRecordedAt}) and gets a full prompt-probe budget (ENG-8225)`
+      );
+    }
+    rotationStallState.set(sk, stall);
+    if (stall.escalated) {
       missing.push(key);
+      continue;
     }
+    const spacingElapsed = stall.lastProbeAt === null || now() - stall.lastProbeAt >= rotationPromptProbeIntervalMs;
+    if (!spacingElapsed) {
+      rotationProbeWaiting.push(key);
+      continue;
+    }
+    if (stall.probeAttempts >= maxRotationPromptProbes) {
+      stall.escalated = true;
+      rotationProbeExhausted.push(key);
+      recordRotationOutcome(key, "never-respawned");
+      log(
+        `[mcp-presence-reaper] '${codeName}:${key}' rotation outcome=never-respawned \u2014 child SIGTERM'd by stale-mcp-reaper never returned after ${stall.probeAttempts} prompt bind-probe(s); escalating (integration marked unhealthy + reaper_action alert) and releasing to the normal restart path (ENG-8225)`
+      );
+      if (onRotationRespawnGaveUp) {
+        try {
+          onRotationRespawnGaveUp(codeName, key, { attempts: stall.probeAttempts });
+        } catch (err) {
+          log(
+            `[mcp-presence-reaper] onRotationRespawnGaveUp callback threw for '${codeName}:${key}' (suppressed; reaper continues): ${err.message}`
+          );
+        }
+      }
+      missing.push(key);
+      continue;
+    }
+    const attempt = stall.probeAttempts + 1;
+    let armed = false;
+    if (onRotationRespawnStalled) {
+      try {
+        armed = onRotationRespawnStalled(codeName, key, {
+          attempt,
+          maxAttempts: maxRotationPromptProbes
+        });
+      } catch (err) {
+        log(
+          `[mcp-presence-reaper] onRotationRespawnStalled callback threw for '${codeName}:${key}' (suppressed; falling back to the pre-ENG-8225 restart path): ${err.message}`
+        );
+        armed = false;
+      }
+    }
+    if (!armed) {
+      missing.push(key);
+      continue;
+    }
+    stall.probeAttempts = attempt;
+    stall.lastProbeAt = now();
+    rotationProbeArmed.push(key);
+    log(
+      `[mcp-presence-reaper] '${codeName}:${key}' rotation grace expired (${rotationGraceMs}ms) with no live child and no tool call to respawn it \u2014 armed prompt bind-probe ${attempt}/${maxRotationPromptProbes} (ENG-7429) instead of restarting the session (ENG-8225)`
+    );
   }
   if (rotationGraced.length > 0) {
     log(
       `[mcp-presence-reaper] '${codeName}': skipping [${rotationGraced.join(", ")}] within rotation grace window (${rotationGraceMs}ms) \u2014 child(ren) just SIGTERM'd by stale-mcp-reaper, awaiting respawn on next tool call (ENG-5344)`
     );
   }
+  if (rotationProbeWaiting.length > 0) {
+    log(
+      `[mcp-presence-reaper] '${codeName}': skipping [${rotationProbeWaiting.join(", ")}] \u2014 prompt bind-probe already armed for the stalled rotation, waiting out the ${rotationPromptProbeIntervalMs}ms re-arm interval (ENG-8225)`
+    );
+  }
   const declared = mcpJson.mcpServers ?? {};
   const liveKeys = /* @__PURE__ */ new Set();
   for (const key of Object.keys(declared)) {
@@ -5827,6 +5943,24 @@ function reapMissingMcpSessions(args) {
     if (typeof entry?.command !== "string") continue;
     if (!missingRaw.includes(key)) liveKeys.add(key);
   }
+  for (const key of liveKeys) {
+    const sk = stateKey(codeName, key);
+    if (classifyRotation(codeName, key, rotationGraceMs, now) === "none") {
+      rotationStallState.delete(sk);
+      continue;
+    }
+    const stall = rotationStallState.get(sk);
+    const viaProbe = (stall?.probeAttempts ?? 0) > 0;
+    recordRotationOutcome(
+      key,
+      viaProbe ? "respawned-after-prompt-probe" : "respawned-on-tool-call"
+    );
+    log(
+      viaProbe ? `[mcp-presence-reaper] '${codeName}:${key}' rotation outcome=respawned-after-prompt-probe \u2014 child is live again after ${stall?.probeAttempts} prompt bind-probe(s)${stall?.escalated ? " (recovered AFTER escalation \u2014 the earlier never-respawned verdict stands for that window)" : ""}, no session restart needed (ENG-8225)` : `[mcp-presence-reaper] '${codeName}:${key}' rotation outcome=respawned-on-tool-call \u2014 child returned on the agent's own next tool call, ENG-5344's assumption held (ENG-8225)`
+    );
+    clearStaleRotation(codeName, [key]);
+    rotationStallState.delete(sk);
+  }
   for (const key of liveKeys) {
     const state = presenceReaperState.get(stateKey(codeName, key));
     if (state) {
@@ -5850,7 +5984,11 @@ function reapMissingMcpSessions(args) {
       missing,
       restarted: false,
       reason: "healthy",
-      rotationGraced: rotationGraced.length > 0 ? rotationGraced : void 0
+      rotationGraced: rotationGraced.length > 0 ? rotationGraced : void 0,
+      rotationProbeArmed: rotationProbeArmed.length > 0 ? rotationProbeArmed : void 0,
+      rotationProbeWaiting: rotationProbeWaiting.length > 0 ? rotationProbeWaiting : void 0,
+      rotationProbeExhausted: rotationProbeExhausted.length > 0 ? rotationProbeExhausted : void 0,
+      rotationRespawnOutcomes: Object.keys(rotationRespawnOutcomes).length > 0 ? rotationRespawnOutcomes : void 0
     };
   }
   const nowMs = now();
@@ -5978,6 +6116,10 @@ function reapMissingMcpSessions(args) {
       givenUp,
       backingOff: backingOff.length > 0 ? backingOff : void 0,
       rotationGraced: rotationGraced.length > 0 ? rotationGraced : void 0,
+      rotationProbeArmed: rotationProbeArmed.length > 0 ? rotationProbeArmed : void 0,
+      rotationProbeWaiting: rotationProbeWaiting.length > 0 ? rotationProbeWaiting : void 0,
+      rotationProbeExhausted: rotationProbeExhausted.length > 0 ? rotationProbeExhausted : void 0,
+      rotationRespawnOutcomes: Object.keys(rotationRespawnOutcomes).length > 0 ? rotationRespawnOutcomes : void 0,
       quarantined: quarantined.length > 0 ? quarantined : void 0,
       wouldQuarantine: wouldQuarantine.length > 0 ? wouldQuarantine : void 0,
       genuinelyDead: genuinelyDead.length > 0 ? genuinelyDead : void 0
@@ -6015,6 +6157,10 @@ function reapMissingMcpSessions(args) {
     givenUp: givenUp.length > 0 ? givenUp : void 0,
     backingOff: backingOff.length > 0 ? backingOff : void 0,
     rotationGraced: rotationGraced.length > 0 ? rotationGraced : void 0,
+    rotationProbeArmed: rotationProbeArmed.length > 0 ? rotationProbeArmed : void 0,
+    rotationProbeWaiting: rotationProbeWaiting.length > 0 ? rotationProbeWaiting : void 0,
+    rotationProbeExhausted: rotationProbeExhausted.length > 0 ? rotationProbeExhausted : void 0,
+    rotationRespawnOutcomes: Object.keys(rotationRespawnOutcomes).length > 0 ? rotationRespawnOutcomes : void 0,
     quarantined: quarantined.length > 0 ? quarantined : void 0,
     wouldQuarantine: wouldQuarantine.length > 0 ? wouldQuarantine : void 0,
     genuinelyDead: genuinelyDead.length > 0 ? genuinelyDead : void 0
@@ -7449,4 +7595,4 @@ export {
   managerInstallSystemUnitCommand,
   managerUninstallSystemUnitCommand
 };
-//# sourceMappingURL=chunk-WCYGJANF.js.map
+//# sourceMappingURL=chunk-UNDFX52P.js.map