agent-relay-orchestrator 0.74.0 → 0.76.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agent-relay-orchestrator",
-  "version": "0.74.0",
+  "version": "0.76.0",
   "description": "Agent Relay orchestrator — manages agent lifecycle across hosts",
   "type": "module",
   "bin": {
@@ -16,7 +16,7 @@
     "test": "bun test"
   },
   "dependencies": {
-    "agent-relay-sdk": "0.2.49"
+    "agent-relay-sdk": "0.2.51"
   },
   "devDependencies": {
     "@types/bun": "latest",

package/src/self-upgrade.ts

@@ -1,6 +1,7 @@
 import { join } from "node:path";
 import { existsSync } from "node:fs";
 import { homedir } from "node:os";
+import { runInstallWithRetry } from "agent-relay-sdk";
 import type { OrchestratorConfig } from "./config";
 import type { RelayClient, RelayCommand } from "./relay";
 import { detectSelfSupervision, type SelfSupervision } from "./self-supervision";
@@ -8,16 +9,6 @@ import { detectSelfSupervision, type SelfSupervision } from "./self-supervision"
 const VALID_PROVIDERS = new Set(["auto", "all", "codex", "claude", "orchestrator"]);
 const SEMVER_RE = /^\d+\.\d+\.\d+(?:-[0-9A-Za-z.-]+)?$/;
 
-/**
- * Stale-packument cache race: right after a publish, a remote host's cached npm
- * metadata may not yet list the new version, so `npm install ...@X` fails with
- * ETARGET / "No matching version found" even though it IS published (#211). The
- * packument refreshes within seconds; `--prefer-online` + retry recovers it.
- */
-const CACHE_RACE_RE = /ETARGET|No matching version found|notarget/i;
-const DEFAULT_INSTALL_RETRIES = 4;
-const DEFAULT_INSTALL_RETRY_BASE_MS = 2000;
-
 interface SelfUpgradeOptions {
   /** Sleep between install retries (injectable for tests). */
   sleep?: (ms: number) => Promise<void>;
@@ -29,10 +20,6 @@ interface SelfUpgradeOptions {
   supervision?: SelfSupervision;
 }
 
-function isCacheRaceError(output: string): boolean {
-  return CACHE_RACE_RE.test(output);
-}
-
 export interface SelfUpgradeRunner {
   run(cmd: string[]): Promise<{ exitCode: number; stdout: string; stderr: string }>;
   commandExists(name: string): boolean;
@@ -137,16 +124,21 @@ export async function handleSelfUpgrade(
     unit: plan.unit,
   });
 
-  const install = await runInstallWithRetry(plan, runner, opts, async (attempt, delayMs) => {
-    await relay.updateCommand(command.id, "running", {
-      phase: "installing",
-      targetVersion: plan.targetVersion,
-      unit: plan.unit,
-      retry: attempt,
-      retryDelayMs: delayMs,
-      note: `target ${plan.targetVersion} not yet visible to this host's npm cache; retrying (attempt ${attempt})`,
-    });
-    console.error(`[orchestrator] self-upgrade install hit a stale-cache race for ${plan.targetVersion}; retry ${attempt} in ${delayMs}ms`);
+  const install = await runInstallWithRetry(plan.installCmd, runner.run.bind(runner), {
+    sleep: opts.sleep,
+    retries: opts.installRetries,
+    baseDelayMs: opts.installRetryBaseMs,
+    onRetry: async ({ attempt, delayMs }) => {
+      await relay.updateCommand(command.id, "running", {
+        phase: "installing",
+        targetVersion: plan.targetVersion,
+        unit: plan.unit,
+        retry: attempt,
+        retryDelayMs: delayMs,
+        note: `target ${plan.targetVersion} not yet visible to this host's npm cache; retrying (attempt ${attempt})`,
+      });
+      console.error(`[orchestrator] self-upgrade install hit a stale-cache race for ${plan.targetVersion}; retry ${attempt} in ${delayMs}ms`);
+    },
   });
   if (install.exitCode !== 0) {
     throw new Error(`install failed (exit ${install.exitCode}): ${(install.stderr || install.stdout).trim().slice(-500)}`);
@@ -167,32 +159,6 @@ export async function handleSelfUpgrade(
   console.error(`[orchestrator] self-upgrade to ${plan.targetVersion} installed; restart dispatched for ${plan.unit}`);
 }
 
-/**
- * Run the install, retrying with exponential backoff when it fails on a
- * stale-packument cache race (#211). Non-cache-race failures return immediately
- * so genuine errors aren't masked by retries.
- */
-async function runInstallWithRetry(
-  plan: SelfUpgradePlan,
-  runner: SelfUpgradeRunner,
-  opts: SelfUpgradeOptions,
-  onRetry: (attempt: number, delayMs: number) => Promise<void>,
-): Promise<{ exitCode: number; stdout: string; stderr: string }> {
-  const retries = opts.installRetries ?? DEFAULT_INSTALL_RETRIES;
-  const baseMs = opts.installRetryBaseMs ?? DEFAULT_INSTALL_RETRY_BASE_MS;
-  const sleep = opts.sleep ?? ((ms: number) => new Promise<void>((resolve) => setTimeout(resolve, ms)));
-
-  let result = await runner.run(plan.installCmd);
-  for (let attempt = 1; attempt <= retries; attempt++) {
-    if (result.exitCode === 0 || !isCacheRaceError(result.stderr || result.stdout)) break;
-    const delayMs = baseMs * 2 ** (attempt - 1);
-    await onRetry(attempt, delayMs);
-    await sleep(delayMs);
-    result = await runner.run(plan.installCmd);
-  }
-  return result;
-}
-
 function normalizeProviders(value: unknown): string[] {
   const list = Array.isArray(value)
     ? value.filter((v): v is string => typeof v === "string").map((v) => v.trim()).filter(Boolean)