@integrity-labs/agt-cli 0.28.55 → 0.28.56

package/dist/bin/agt.js

@@ -37,7 +37,7 @@ import {
   success,
   table,
   warn
-} from "../chunk-7P2VKTWK.js";
+} from "../chunk-OLP3LORD.js";
 import {
   CHANNEL_REGISTRY,
   DEPLOYMENT_TEMPLATES,
@@ -4777,7 +4777,7 @@ 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 ? "0.28.55" : "dev";
+var cliVersion = true ? "0.28.56" : "dev";
 async function fetchLatestVersion() {
   const host2 = getHost();
   if (!host2) return null;
@@ -5791,7 +5791,7 @@ function handleError(err) {
 }
 
 // src/bin/agt.ts
-var cliVersion2 = true ? "0.28.55" : "dev";
+var cliVersion2 = true ? "0.28.56" : "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-7P2VKTWK.js → chunk-OLP3LORD.js}

@@ -7355,7 +7355,7 @@ function requireHost() {
 }
 
 // src/lib/api-client.ts
-var agtCliVersion = true ? "0.28.55" : "dev";
+var agtCliVersion = true ? "0.28.56" : "dev";
 var lastConfigHash = null;
 function setConfigHash(hash) {
   lastConfigHash = hash && hash.length > 0 ? hash : null;
@@ -8651,4 +8651,4 @@ export {
   managerInstallSystemUnitCommand,
   managerUninstallSystemUnitCommand
 };
-//# sourceMappingURL=chunk-7P2VKTWK.js.map
+//# sourceMappingURL=chunk-OLP3LORD.js.map

package/dist/lib/manager-worker.js

@@ -27,7 +27,7 @@ import {
   requireHost,
   safeWriteJsonAtomic,
   setConfigHash
-} from "../chunk-7P2VKTWK.js";
+} from "../chunk-OLP3LORD.js";
 import {
   getProjectDir as getProjectDir2,
   getReadyTasks,
@@ -444,6 +444,12 @@ function reapStaleMcpChildren(args) {
 import { execFileSync as execFileSync2 } from "child_process";
 var DEFAULT_COLD_START_GRACE_MS = 9e4;
 var MAX_PRESENCE_RESTART_ATTEMPTS = 3;
+var BACKOFF_AFTER_RESTARTS = MAX_PRESENCE_RESTART_ATTEMPTS;
+var BACKOFF_BASE_MS = 6e4;
+var BACKOFF_MAX_MS = 30 * 6e4;
+var BACKOFF_WINDOW_MS = 60 * 6e4;
+var BACKOFF_GIVE_UP_RESTARTS = 2 * MAX_PRESENCE_RESTART_ATTEMPTS;
+var BACKOFF_STABLE_RESET_MS = 5 * 6e4;
 var presenceReaperState = /* @__PURE__ */ new Map();
 function stateKey(codeName, serverKey) {
   return `${codeName}\0${serverKey}`;
@@ -463,7 +469,7 @@ function givenUpMcpServerKeys(codeName) {
   const prefix = `${codeName}\0`;
   const out = /* @__PURE__ */ new Set();
   for (const [key, st] of presenceReaperState) {
-    if (key.startsWith(prefix) && st.attempts >= MAX_PRESENCE_RESTART_ATTEMPTS) {
+    if (key.startsWith(prefix) && (st.gaveUp || st.attempts >= MAX_PRESENCE_RESTART_ATTEMPTS)) {
       out.add(key.slice(prefix.length));
     }
   }
@@ -565,6 +571,11 @@ function reapMissingMcpSessions(args) {
       state7.gaveUpLogged = false;
       state7.firstMissingAt = null;
       state7.quarantineLogged = false;
+      state7.gaveUp = false;
+      if (state7.stableLiveSince === null) state7.stableLiveSince = now();
+      if (now() - state7.stableLiveSince >= BACKOFF_STABLE_RESET_MS) {
+        state7.restartHistory = [];
+      }
     }
   }
   if (missing.length === 0) {
@@ -578,6 +589,8 @@ function reapMissingMcpSessions(args) {
   const nowMs = now();
   const givenUp = [];
   const active = [];
+  const backingOff = [];
+  const newGenerationKeys = /* @__PURE__ */ new Set();
   const quarantined = [];
   const wouldQuarantine = [];
   for (const key of missing) {
@@ -588,19 +601,30 @@ function reapMissingMcpSessions(args) {
       lastAttemptedSessionStartedAt: null,
       gaveUpLogged: false,
       firstMissingAt: null,
-      quarantineLogged: false
+      quarantineLogged: false,
+      restartHistory: [],
+      stableLiveSince: null,
+      gaveUp: false
     };
     if (state7.firstMissingAt === null) state7.firstMissingAt = nowMs;
+    state7.stableLiveSince = null;
     if (state7.lastAttemptedSessionStartedAt !== sessionStartedAt) {
       state7.attempts += 1;
       state7.lastAttemptedSessionStartedAt = sessionStartedAt;
+      newGenerationKeys.add(key);
     }
+    state7.restartHistory = state7.restartHistory.filter((t) => nowMs - t < BACKOFF_WINDOW_MS);
+    const restartsInWindow = state7.restartHistory.length;
     presenceReaperState.set(sk, state7);
-    if (state7.attempts > MAX_PRESENCE_RESTART_ATTEMPTS) {
+    const overAttemptsCap = state7.attempts > MAX_PRESENCE_RESTART_ATTEMPTS;
+    const overRateCap = restartsInWindow >= BACKOFF_GIVE_UP_RESTARTS;
+    if (overAttemptsCap || overRateCap) {
       givenUp.push(key);
+      state7.gaveUp = true;
       if (!state7.gaveUpLogged) {
+        const cause = overAttemptsCap ? `${MAX_PRESENCE_RESTART_ATTEMPTS} consecutive failed restarts \u2014 declared but never recovers (likely orphaned config; see ENG-5279)` : `${restartsInWindow} restarts within ${Math.round(BACKOFF_WINDOW_MS / 6e4)}m \u2014 persistently flapping (crash-loop; ENG-6480)`;
         log2(
-          `[mcp-presence-reaper] giving up on '${codeName}:${key}' after ${MAX_PRESENCE_RESTART_ATTEMPTS} consecutive failed restarts \u2014 declared but never recovers (likely orphaned config; see ENG-5279)`
+          `[mcp-presence-reaper] giving up on '${codeName}:${key}' after ${cause}`
         );
         state7.gaveUpLogged = true;
         if (onGiveUp) {
@@ -640,15 +664,31 @@ function reapMissingMcpSessions(args) {
         }
       }
     } else {
-      active.push(key);
+      const backoffMs = restartsInWindow < BACKOFF_AFTER_RESTARTS ? 0 : Math.min(
+        BACKOFF_BASE_MS * 2 ** (restartsInWindow - BACKOFF_AFTER_RESTARTS),
+        BACKOFF_MAX_MS
+      );
+      const lastRestartAt = state7.restartHistory[state7.restartHistory.length - 1] ?? null;
+      if (lastRestartAt !== null && nowMs - lastRestartAt < backoffMs) {
+        backingOff.push(key);
+      } else {
+        active.push(key);
+      }
     }
   }
   if (active.length === 0) {
+    const reason = backingOff.length > 0 ? "backoff" : "all-keys-over-cap";
+    if (backingOff.length > 0) {
+      log2(
+        `[mcp-presence-reaper] '${codeName}': holding off restart for [${backingOff.join(", ")}] \u2014 escalating backoff interval not yet elapsed (ENG-6480)${givenUp.length > 0 ? ` (given up: [${givenUp.join(", ")}])` : ""}`
+      );
+    }
     return {
       missing,
       restarted: false,
-      reason: "all-keys-over-cap",
+      reason,
       givenUp,
+      backingOff: backingOff.length > 0 ? backingOff : void 0,
       rotationGraced: rotationGraced.length > 0 ? rotationGraced : void 0,
       quarantined: quarantined.length > 0 ? quarantined : void 0,
       wouldQuarantine: wouldQuarantine.length > 0 ? wouldQuarantine : void 0
@@ -675,10 +715,16 @@ function reapMissingMcpSessions(args) {
     }
   }
   stopSession(codeName, { activeKeys: active });
+  for (const key of active) {
+    if (!newGenerationKeys.has(key)) continue;
+    const st = presenceReaperState.get(stateKey(codeName, key));
+    if (st) st.restartHistory.push(nowMs);
+  }
   return {
     missing,
     restarted: true,
     givenUp: givenUp.length > 0 ? givenUp : void 0,
+    backingOff: backingOff.length > 0 ? backingOff : void 0,
     rotationGraced: rotationGraced.length > 0 ? rotationGraced : void 0,
     quarantined: quarantined.length > 0 ? quarantined : void 0,
     wouldQuarantine: wouldQuarantine.length > 0 ? wouldQuarantine : void 0
@@ -6634,7 +6680,7 @@ var cachedMaintenanceWindow = null;
 var lastVersionCheckAt = 0;
 var VERSION_CHECK_INTERVAL_MS = 5 * 60 * 1e3;
 var lastResponsivenessProbeAt = 0;
-var agtCliVersion = true ? "0.28.55" : "dev";
+var agtCliVersion = true ? "0.28.56" : "dev";
 function resolveBrewPath(execFileSync4) {
   try {
     const out = execFileSync4("which", ["brew"], { timeout: 5e3 }).toString().trim();