Preserve daemon sessions when workers stall (#3263)

Stop treating heartbeat silence as proof that a live daemon worker should be destroyed. Crash recovery, explicit lifecycle requests, and supervisor-loss cleanup remain unchanged.

packages/server/scripts/supervisor.logging.test.ts

@@ -10,15 +10,9 @@ import { resolveSupervisorLogFile } from "./supervisor-log-config.js";
 const repoRoot = path.resolve(fileURLToPath(new URL("../../..", import.meta.url)));
 const supervisorPath = fileURLToPath(new URL("./supervisor.ts", import.meta.url));
 
-interface SupervisorSuspension {
-  afterMs: number;
-  durationMs: number;
-}
-
 async function runSupervisorFixture(options: {
   workerSource: string;
   restartOnCrash?: boolean;
-  suspension?: SupervisorSuspension;
   timeoutMs?: number;
 }): Promise<{
   code: number | null;
@@ -58,23 +52,10 @@ async function runSupervisorFixture(options: {
   const startedAt = Date.now();
   const child = spawn(process.execPath, ["--import", "tsx", runnerPath], {
     cwd: repoRoot,
-    detached: options.suspension !== undefined,
     env: { ...process.env },
     stdio: ["ignore", "pipe", "pipe"],
   });
 
-  if (options.suspension) {
-    const supervisorPid = child.pid;
-    if (supervisorPid === undefined) {
-      throw new Error("Supervisor fixture did not start");
-    }
-    const { afterMs, durationMs } = options.suspension;
-    setTimeout(() => {
-      process.kill(-supervisorPid, "SIGSTOP");
-      setTimeout(() => process.kill(-supervisorPid, "SIGCONT"), durationMs);
-    }, afterMs);
-  }
-
   let stdout = "";
   let stderr = "";
   child.stdout.setEncoding("utf8");
@@ -208,46 +189,21 @@ describe("supervisor durable logging", () => {
     expect(result.log).toContain('"workerPid":');
   });
 
-  test("keeps a worker alive while heartbeats continue", async () => {
+  test("does not restart a worker based on heartbeat absence", async () => {
     const result = await runSupervisorFixture({
-      timeoutMs: 10_000,
+      timeoutMs: 20_000,
-      workerSource: `
-        setInterval(() => {
-          process.send?.({ type: "paseo:worker-heartbeat" });
-        }, 250);
-        setTimeout(() => {
-          process.send?.({ type: "paseo:shutdown", reason: "healthy_heartbeat_test_complete" });
-        }, 6000);
-      `,
-    });
-
-    expect(result.code).toBe(0);
-    expect(result.signal).toBeNull();
-    expect(result.log).toContain('"reason":"healthy_heartbeat_test_complete"');
-    expect(result.log).not.toContain('"msg":"Worker heartbeat timed out; restarting worker"');
-  }, 10_000);
-
-  test("tolerates a transient seven-second heartbeat pause", async () => {
-    const result = await runSupervisorFixture({
-      timeoutMs: 10_000,
       workerSource: `
         import { existsSync, writeFileSync } from "node:fs";
 
         const marker = process.argv[1] + ".started";
         if (!existsSync(marker)) {
           writeFileSync(marker, "started");
-          let heartbeatCount = 0;
-          const heartbeat = setInterval(() => {
-            process.send?.({ type: "paseo:worker-heartbeat" });
-            heartbeatCount += 1;
-            if (heartbeatCount === 3) clearInterval(heartbeat);
-          }, 100);
           setTimeout(() => {
-            process.send?.({ type: "paseo:shutdown", reason: "heartbeat_pause_tolerated" });
+            process.send?.({ type: "paseo:shutdown", reason: "silent_worker_test_complete" });
-          }, 7_000);
+          }, 16_000);
           setInterval(() => {}, 1_000);
         } else {
-          process.send?.({ type: "paseo:shutdown", reason: "unexpected_heartbeat_restart" });
+          process.send?.({ type: "paseo:shutdown", reason: "unexpected_silent_worker_restart" });
           setInterval(() => {}, 1_000);
         }
       `,
@@ -255,44 +211,10 @@ describe("supervisor durable logging", () => {
 
     expect(result.code).toBe(0);
     expect(result.signal).toBeNull();
-    expect(result.log).toContain('"reason":"heartbeat_pause_tolerated"');
+    expect(result.log).toContain('"reason":"silent_worker_test_complete"');
-    expect(result.log).not.toContain('"reason":"unexpected_heartbeat_restart"');
+    expect(result.log).not.toContain('"reason":"unexpected_silent_worker_restart"');
     expect(result.log).not.toContain('"msg":"Worker heartbeat timed out; restarting worker"');
-  }, 10_000);
+  }, 25_000);
-
-  test.skipIf(isPlatform("win32"))(
-    "keeps a healthy worker alive after the host resumes from sleep",
-    async () => {
-      const result = await runSupervisorFixture({
-        suspension: { afterMs: 1_000, durationMs: 16_000 },
-        timeoutMs: 25_000,
-        workerSource: `
-          import { existsSync, writeFileSync } from "node:fs";
-
-          const marker = process.argv[1] + ".started";
-          if (!existsSync(marker)) {
-            writeFileSync(marker, "started");
-            setInterval(() => {
-              process.send?.({ type: "paseo:worker-heartbeat" });
-            }, 250);
-            setTimeout(() => {
-              process.send?.({ type: "paseo:shutdown", reason: "sleep_resume_test_complete" });
-            }, 17_000);
-          } else {
-            process.send?.({ type: "paseo:shutdown", reason: "unexpected_sleep_restart" });
-          }
-          setInterval(() => {}, 1_000);
-        `,
-      });
-
-      expect(result.code).toBe(0);
-      expect(result.signal).toBeNull();
-      expect(result.log).toContain('"reason":"sleep_resume_test_complete"');
-      expect(result.log).not.toContain('"reason":"unexpected_sleep_restart"');
-      expect(result.log).not.toContain('"msg":"Worker heartbeat timed out; restarting worker"');
-    },
-    30_000,
-  );
 
   test.skipIf(isPlatform("win32"))(
     "forces shutdown when a worker ignores SIGTERM",
@@ -315,42 +237,6 @@ describe("supervisor durable logging", () => {
     20_000,
   );
 
-  // POSIX-only: the watchdog uses SIGKILL after its graceful shutdown window.
-  test.skipIf(isPlatform("win32"))(
-    "restarts a worker that stops heartbeating",
-    async () => {
-      const result = await runSupervisorFixture({
-        timeoutMs: 35_000,
-        workerSource: `
-          import { existsSync, writeFileSync } from "node:fs";
-
-          const marker = process.argv[1] + ".started";
-          if (!existsSync(marker)) {
-            writeFileSync(marker, "started");
-            let heartbeatCount = 0;
-            const heartbeat = setInterval(() => {
-              process.send?.({ type: "paseo:worker-heartbeat" });
-              heartbeatCount += 1;
-              if (heartbeatCount === 3) clearInterval(heartbeat);
-            }, 100);
-            process.on("SIGTERM", () => {});
-            setInterval(() => {}, 1000);
-          } else {
-            process.send?.({ type: "paseo:shutdown", reason: "watchdog_test_complete" });
-            setInterval(() => {}, 1000);
-          }
-        `,
-      });
-
-      expect(result.code).toBe(0);
-      expect(result.signal).toBeNull();
-      expect(result.log).toContain('"msg":"Worker heartbeat timed out; restarting worker"');
-      expect(result.log).toContain('"msg":"Worker did not exit after SIGTERM; forcing SIGKILL"');
-      expect(result.log).toContain('"signal":"SIGKILL"');
-    },
-    40_000,
-  );
-
   test.skipIf(isPlatform("win32"))(
     "restarts after worker exit while a descendant retains the worker stdio",
     async () => {

packages/server/scripts/supervisor.ts

@@ -5,7 +5,6 @@ import { createStream as createRotatingFileStream } from "rotating-file-stream";
 import { signalProcessTree } from "../src/utils/tree-kill.js";
 
 const WORKER_HEARTBEAT_INTERVAL_MS = 1_000;
-const WORKER_HEARTBEAT_MISSED_CHECK_LIMIT = 15;
 const WORKER_TERMINATION_GRACE_MS = 10_000;
 
 interface SupervisorLogFileOptions {
@@ -34,10 +33,6 @@ interface SupervisorHeartbeatMessage {
   type: "paseo:supervisor-heartbeat";
 }
 
-interface WorkerHeartbeatMessage {
-  type: "paseo:worker-heartbeat";
-}
-
 interface SupervisorOptions {
   name: string;
   startupMessage: string;
@@ -93,15 +88,6 @@ function parseLifecycleMessage(msg: unknown): WorkerLifecycleMessage | null {
   return null;
 }
 
-function isWorkerHeartbeatMessage(msg: unknown): msg is WorkerHeartbeatMessage {
-  return (
-    typeof msg === "object" &&
-    msg !== null &&
-    "type" in msg &&
-    (msg as { type?: unknown }).type === "paseo:worker-heartbeat"
-  );
-}
-
 function toRotatingFileStreamSize(size: string): string {
   const trimmed = size.trim();
   const match = trimmed.match(/^(\d+)\s*([bBkKmMgG])?$/);
@@ -250,7 +236,6 @@ export function runSupervisor(options: SupervisorOptions): SupervisorController
     }
 
     const currentChild = child;
-    let missedWorkerHeartbeatChecks = 0;
     const heartbeat = setInterval(() => {
       const message: SupervisorHeartbeatMessage = { type: "paseo:supervisor-heartbeat" };
       if (currentChild.connected) {
@@ -267,23 +252,6 @@ export function runSupervisor(options: SupervisorOptions): SupervisorController
     }, WORKER_HEARTBEAT_INTERVAL_MS);
     heartbeat.unref();
 
-    const workerWatchdog = setInterval(() => {
-      if (child !== currentChild || restarting || shuttingDown) {
-        return;
-      }
-      missedWorkerHeartbeatChecks += 1;
-      if (missedWorkerHeartbeatChecks < WORKER_HEARTBEAT_MISSED_CHECK_LIMIT) {
-        return;
-      }
-      writeLifecycleLog("Worker heartbeat timed out; restarting worker", {
-        missedHeartbeatChecks: missedWorkerHeartbeatChecks,
-        supervisorPid: process.pid,
-        workerPid: currentChild.pid ?? null,
-      });
-      requestRestart("worker_heartbeat_timeout");
-    }, WORKER_HEARTBEAT_INTERVAL_MS);
-    workerWatchdog.unref();
-
     child.on("disconnect", () => {
       writeLifecycleLog("Worker IPC channel disconnected");
     });
@@ -299,10 +267,6 @@ export function runSupervisor(options: SupervisorOptions): SupervisorController
     });
 
     child.on("message", (msg: unknown) => {
-      if (isWorkerHeartbeatMessage(msg)) {
-        missedWorkerHeartbeatChecks = 0;
-        return;
-      }
       const lifecycleMessage = parseLifecycleMessage(msg);
       if (!lifecycleMessage) {
         return;
@@ -333,7 +297,6 @@ export function runSupervisor(options: SupervisorOptions): SupervisorController
 
     child.on("exit", (code, signal) => {
       clearInterval(heartbeat);
-      clearInterval(workerWatchdog);
       clearForceKillTimer();
       const exitDescriptor = describeExit(code, signal);
       writeLifecycleLog("Worker exited", { code, signal, exit: exitDescriptor });

packages/server/src/server/daemon-worker.ts

@@ -27,10 +27,6 @@ interface SupervisorHeartbeatMessage {
   type: "paseo:supervisor-heartbeat";
 }
 
-interface WorkerHeartbeatMessage {
-  type: "paseo:worker-heartbeat";
-}
-
 interface BootstrapResult {
   paseoHome: string;
   logger: ReturnType<typeof createRootLogger>;
@@ -232,13 +228,6 @@ async function main() {
     const supervisorPid = process.ppid;
     let lastSupervisorHeartbeatAt = Date.now();
     let supervisorExitRequested = false;
-    const sendWorkerHeartbeat = () => {
-      try {
-        process.send?.({ type: "paseo:worker-heartbeat" } satisfies WorkerHeartbeatMessage);
-      } catch {
-        // The disconnect handler below owns supervisor-loss shutdown.
-      }
-    };
     const exitAfterSupervisorLoss = (reason: string) => {
       if (supervisorExitRequested) {
         return;
@@ -272,10 +261,6 @@ async function main() {
     });
     process.on("disconnect", () => exitAfterSupervisorLoss("ipc_disconnect_event"));
 
-    sendWorkerHeartbeat();
-    const workerHeartbeat = setInterval(sendWorkerHeartbeat, 1_000);
-    workerHeartbeat.unref();
-
     const timer = setInterval(() => {
       const ipcConnected = typeof process.connected === "boolean" ? process.connected : true;
       const heartbeatExpired = Date.now() - lastSupervisorHeartbeatAt > 3500;

skills/paseo-committee/SKILL.md

@@ -6,9 +6,7 @@ user-invocable: true
 
 # Committee Skill
 
-Two agents from contrasting profiles, fresh context, planning a solution in parallel. They stay alive for review after implementation.
+Two agents from contrasting profiles, fresh context, planning a solution in parallel.
-
-The purpose is to step back, not double down. The committee may propose a completely different approach.
 
 **User's additional context:** $ARGUMENTS
 
@@ -35,49 +33,14 @@ If the user names profiles, use those. If fewer than two suitable profiles are c
   This is analysis only. Do NOT edit, create, or delete any files. Do NOT write code.
   ```
 
-- **Trust the wait.** Do not poll, send hurry-ups, or interrupt. GPT-5.4 can reason 15–30 minutes; Opus does extended thinking. Long waits mean it found something worth thinking about.
+- **Trust the finish notification.** Do not poll, send hurry-ups, or interrupt. Models can reason for 15–30 minutes. You can go idle and Paseo will notify you.
-- **You are the middleman.** Drive plan → implement → review without yielding to the user, except for divergences that need their call.
 
-## Phase 1: Plan
+## Workflow
 
-Write a problem-level prompt:
+1. Write a problem-level prompt
+2. Create both agents in parallel via Paseo with `[Committee] <task>` titles and the same prompt
+3. Wait for both responses
+4. Resolve disagreements by passing their arguments between each other
+5. Keep going until they converge into a response
 
-- High-level goal and acceptance criteria
+Share the consensus with the user. Summarize where the agents diverged and how they resolved it.
-- Constraints
-- Symptoms (if a bug)
-- What you tried and why it failed
-- Explicit: "do root cause analysis"
-- Explicit: "state assumptions, ask why three levels deep, check whether you're patching a symptom or removing the problem"
-
-Create both agents in parallel via Paseo with `[Committee] <task>` titles and the same prompt. Wait for both — not just whichever finishes first.
-
-Read both responses. Challenge them — do not accept at face value:
-
-- "Why does <underlying thing> happen? Symptom or cause?"
-- Verify any assumption the plan makes about the code.
-- "What did you consider and reject?"
-
-Send follow-ups until the plan addresses root cause.
-
-Synthesize:
-
-- Convergence → unified plan.
-- Significant divergence → involve the user.
-
-Confirm the merged plan with both members. Multi-turn until consensus.
-
-## Phase 2: Implement
-
-Default: implement yourself. If the user said **"delegate"**, launch one impl agent and pass the merged plan.
-
-The committee stays clean — not involved in implementation.
-
-## Phase 3: Review
-
-Send the diff to the committee:
-
-> Implementation is done. Review changes against the plan. Flag drift or missing pieces. <no-edits suffix>
-
-Apply feedback yourself, or send to the impl agent. Repeat 2 → 3 until consensus.
-
-After ~10 iterations without convergence, start a fresh committee with the full history of what was tried — the current committee's context may have drifted too far.

skills/paseo-handoff/SKILL.md

@@ -62,6 +62,4 @@ Prepare the handoff in a dedicated workspace:
 
 Do not encode independence as a create mode and do not invoke CLI or wire-level detach operations. Detach is a user gesture in the subagents track.
 
-Leave `notifyOnFinish` omitted unless the user explicitly wants no callback.
+Do not wait or poll for the agent to finish.
-
-Don't wait by default — the user decides whether to follow along or move on. Tell them the agent ID and how to follow along (the paseo skill explains).

skills/paseo/SKILL.md

@@ -3,7 +3,7 @@ name: paseo
 description: Paseo reference for managing workspaces, workspace scripts, agents, schedules, and heartbeats.
 ---
 
-Paseo is a daemon that supervises AI coding agents on your machine. Control it through tools or a CLI.
+Paseo is a remote daemon that manages coding agents, terminals. Control it through MCP tools or the CLI.
 
 ## Workspaces
 
@@ -13,7 +13,7 @@ Paseo is a daemon that supervises AI coding agents on your machine. Control it t
 
 **`archive_workspace`** — `{ workspaceId }`. Archives the workspace, its agents, and its terminals. Local directories remain; Paseo removes an owned worktree only after its final active workspace reference is archived.
 
-Worktree creation and reference accounting are implementation details of `isolation: "worktree"`.
+**`rename_workspace`** — `{ workspaceId, name }`. Rename workspace.
 
 ## Workspace scripts