an open-source alternative to the dots, bots and muses of the world, run an office of claude code/codex like agents on your laptop, sandboxes or anywhere, uses your existing subscriptions
HTML agent-harnessagent-orchestrationagentsai-agentsautonomous-agentsclaude-codecodexcoding-agentsdesktop-appdeveloper-toolselectrongemini-cliharnessharness-engineeringlocal-firstmemorymulti-agentopencodeorchestrationtypescript
fix(hive): re-arm message router on system resume so delivery survives sleep
Chaitanya Giria0609eadd536
diff --git a/CHANGELOG.md b/CHANGELOG.mdindex 607bc6e8..4c1b7ba8 100644--- a/CHANGELOG.md+++ b/CHANGELOG.md@@ -4,6 +4,15 @@ All notable changes to this project are documented here. The format is based on[Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and this project adheres to[Semantic Versioning](https://semver.org/spec/v2.0.0.html).+## [0.3.1] — 2026-06-22++A reliability patch for the hive message router. Completes the **wake-reliability hardening**+started in 0.3.0 (which caught up missed *schedules* and auto-revived wedged *terminals* on wake,+but left the *router* out).++### Fixed+- **Re-arm the hive message router on wake (god→worker delivery survives sleep).** The outbox→inbox router is a `setInterval` (`hive.routeOnce` every ~1.5s) that, like the always-on beats, freezes during true macOS system sleep. `onSystemResume()` already re-armed the mission scheduler, the fleet/breaker beats, and keep-awake on `powerMonitor` `resume`/`unlock-screen` — but it never re-armed the router. So after a long sleep (e.g. laptop closed overnight) the scheduler→god path recovered while **every agent's outbox silently stopped draining**: god→worker, worker↔worker, and broadcast mail piled up undelivered, and no `message` event was logged. The resume handler now re-arms the router (clear-then-set, idempotent) **and** immediately drains the accumulated backlog instead of waiting for the first post-wake tick; the renderer's idle inbox-wake nudge then wakes each parked recipient once its mail lands (`src/main/index.ts`). Verified by `scripts/verify-keepalive-catchup.mjs` (now also reproduces the pre-fix backlog stall and proves the re-arm + flush).+## [0.3.0] — 2026-06-21A platform release: the floor stops being Claude-shaped. **Selectable agent engines** makediff --git a/README.md b/README.mdindex 8f67656d..1ac37572 100644--- a/README.md+++ b/README.md@@ -267,7 +267,7 @@ Shipped in **v0.2.0–v0.3.0**:- [x] **Slack-spawned ephemeral workers** — Michael spawns an isolated worker from a Slack request, replies in-thread, then tears it down safely (worktree GC + per-worker token caps), surfaced in a Workers tab.- [x] **Temporal date-range skills + worker capability catalog** — named windows resolve to concrete ISO dates, and each worker can read exactly what tools/integrations it has.- [x] **Agent Gallery + six off-the-shelf hires** — *The Hiring Fair* rebranded, with a visual Provider/Hive picker in onboarding and add-agent and feature-aware onboarding.-- [x] **Wake-reliability hardening** — auto-revive wedged terminals and catch up missed schedules when the machine wakes.+- [x] **Wake-reliability hardening** — auto-revive wedged terminals, catch up missed schedules, and re-arm the hive message router (draining any mail that piled up) when the machine wakes.- [x] **Shareable hires** — import a role-configured agent from a `munderdifflin://hire` deep link or a local manifest; import pre-fills the Add-Agent modal (the human spawns), and the manifest is validated as untrusted input. Ready-made roles live in the [Agent Gallery](https://munderdiffl.in/hires/).- [x] **Heartbeat** — scheduler heartbeat that re-engages the floor when it goes quiet, with last/next-fired times surfaced in the Schedules tab.diff --git a/RELEASE.md b/RELEASE.mdindex 4d39eb66..efa84f7a 100644--- a/RELEASE.md+++ b/RELEASE.md@@ -1,4 +1,4 @@-# Munder Difflin v0.3.0+# Munder Difflin v0.3.1**A local hive of Claude Code, Antigravity & Codex agents that run themselves** — messaging,routing, and remembering, coordinated by a GOD orchestrator you talk to. Local-first and open source.@@ -7,6 +7,18 @@ routing, and remembering, coordinated by a GOD orchestrator you talk to. Local-f---+## What's new in 0.3.1 — *Router survives sleep*++A reliability patch. The hive message router (the loop that drains each agent's `outbox/` into+recipients' `inbox/`) is a timer that **freezes during true macOS system sleep**. 0.3.0's+wake-reliability hardening re-armed the scheduler and revived wedged terminals on wake, but the+router was left out — so after a long sleep (laptop closed overnight) the GOD orchestrator's+standups kept arriving while **agent-to-agent and god-to-worker mail silently stopped delivering**.+On wake the harness now **re-arms the router and immediately drains whatever piled up**, so+delegation keeps flowing after the machine sleeps. (Verified by `scripts/verify-keepalive-catchup.mjs`.)++---+## What's new in 0.3.0 — *Selectable engines, integrations, and Slack-spawned workers*The biggest platform release yet: the floor stops being Claude-shaped. Every hire — and Michaeldiff --git a/docs/index.html b/docs/index.htmlindex b95678e3..003246b0 100644--- a/docs/index.html+++ b/docs/index.html@@ -33,7 +33,7 @@"url": "https://munderdiffl.in/","applicationCategory": "DeveloperApplication","operatingSystem": "macOS, Windows, Linux",- "softwareVersion": "0.3.0",+ "softwareVersion": "0.3.1","downloadUrl": "https://github.com/chaitanyagiri/munder-difflin/releases/latest","softwareHelp": "https://github.com/chaitanyagiri/munder-difflin#readme","license": "https://github.com/chaitanyagiri/munder-difflin/blob/main/LICENSE",@@ -746,7 +746,7 @@<!-- ===================== HERO ===================== --><header class="hero"><div class="wrap">- <div class="chip reveal">⚡ <b>v0.3.0</b> · Multi-provider · Open source · macOS · Windows</div>+ <div class="chip reveal">⚡ <b>v0.3.1</b> · Multi-provider · Open source · macOS · Windows</div><h1 class="hero-title reveal">Run a virtual <span style="color:var(--sky)">office of AI agents</span> on any computer.</h1>@@ -1582,7 +1582,7 @@// download picker — pick macOS / Windows / Linux, click starts the download(function () {var REPO = 'chaitanyagiri/munder-difflin';- var REL = '0.3.0'; // fallback version; live links are resolved from the GitHub API below+ var REL = '0.3.1'; // fallback version; live links are resolved from the GitHub API belowvar BASE = 'https://github.com/' + REPO + '/releases/latest/download/';var PLATFORMS = [{ key: 'mac', icon: '🍎', os: 'macOS', sub: 'Universal · Apple Silicon + Intel', match: 'mac-universal.dmg', file: 'Munder-Difflin-' + REL + '-mac-universal.dmg' },diff --git a/package.json b/package.jsonindex 0f42e05d..d1d97e66 100644--- a/package.json+++ b/package.json@@ -1,6 +1,6 @@{"name": "munder-difflin",- "version": "0.3.0",+ "version": "0.3.1","private": true,"description": "Local multi-agent harness for Claude Code: autonomous agents that message, route, and remember — coordinated by a GOD orchestrator and visualized as avatars at work.","main": "out/main/index.js",diff --git a/scripts/verify-keepalive-catchup.mjs b/scripts/verify-keepalive-catchup.mjsindex b126129f..0e12727f 100644--- a/scripts/verify-keepalive-catchup.mjs+++ b/scripts/verify-keepalive-catchup.mjs@@ -13,6 +13,13 @@// 3. A mission NOT yet due does not fire early on resume.// 4. Overlapping resume + unlock-screen events collapse to ONE catch-up fire// (onSystemResume is idempotent: clear-then-arm).+// 5. PRE-FIX REGRESSION: the hive message router is a setInterval (routeOnce+// every ~1.5s) that freezes across true system sleep just like the beats;+// the old onSystemResume re-armed the scheduler + beats but NOT the router,+// so an outbox that piled up while asleep stayed undelivered after wake.+// 6. FIX: the new onSystemResume re-arms the router AND flushes the backlog+// immediately on wake (no tick wait) — god->worker mail delivers on resume.+// 7. FIX: the re-armed router keeps draining subsequent mail at its cadence.// ---- fake clock + timer queue (libuv-style monotonic; we control advancement) --let nowMs = 1_000_000; // arbitrary "boot" instant@@ -90,5 +97,44 @@ onSystemResume(m); // 'unlock-screen' immediately after — mustadvance(0);ok(fireCount === 1, `resume+unlock collapse to ONE catch-up fire (idempotent) -> fireCount=${fireCount}`);+// ---- router re-arm on wake (the god->worker delivery fix) -----------------------+// Mirror of the hive message router: a setInterval that drains every agent's+// outbox into recipient inboxes (hive.startRouter/stopRouter/routeOnce). Like the+// always-on beats it freezes across true system sleep; the fix re-arms it AND+// flushes the backlog in onSystemResume's new router block.+const ROUTER_MS = 1500;+let queue = 0; // outbox messages awaiting routing (e.g. god's pile-up)+let delivered = 0; // messages routed into recipient inboxes+let routerTimer = null;+const routeOnce = () => { const n = queue; queue = 0; delivered += n; return n; }; // mirrors hive.routeOnce()+const startRouter = () => { if (routerTimer === null) routerTimer = setI(routeOnce, ROUTER_MS); }; // idempotent, like hive.startRouter+const stopRouter = () => { if (routerTimer !== null) { clr(routerTimer); routerTimer = null; } };+const resumeOld = () => { /* BUG: scheduler/beats re-armed, router left frozen */ };+const resumeNew = () => { stopRouter(); startRouter(); routeOnce(); }; // the new onSystemResume router block++reset();+startRouter();+queue = 1; advance(ROUTER_MS);+ok(delivered === 1, `awake: router drains a freshly-written outbox within one tick -> delivered=${delivered}`);++stopRouter(); // true system sleep halts the libuv interval (dead until re-armed)+queue += 3; // god's outbox accumulates while the floor is asleep+advance(100 * ROUTER_MS); // wall-clock passes; the frozen timer never fires+ok(delivered === 1, `asleep: frozen router does NOT drain the backlog -> delivered=${delivered}`);++// (5) PRE-FIX: old resume leaves the router dead -> the backlog never delivers.+resumeOld();+advance(100 * ROUTER_MS);+ok(delivered === 1, `pre-fix resume: backlog STAYS undelivered (reproduces the bug) -> delivered=${delivered}`);++// (6) FIX: new resume re-arms + flushes immediately on wake (no tick wait).+resumeNew();+advance(0);+ok(delivered === 4, `fixed resume: backlog flushed immediately on wake -> delivered=${delivered}`);++// (7) FIX: the re-armed router keeps draining subsequent mail at its cadence.+queue += 2; advance(ROUTER_MS);+ok(delivered === 6, `fixed resume: re-armed router drains subsequent mail -> delivered=${delivered}`);+console.log(failures === 0 ? '\nALL CHECKS PASS' : `\n${failures} CHECK(S) FAILED`);process.exit(failures === 0 ? 0 : 1);diff --git a/src/main/index.ts b/src/main/index.tsindex 712a8c9c..6d2b6a6a 100644--- a/src/main/index.ts+++ b/src/main/index.ts@@ -3127,9 +3127,24 @@ function healthCheckPtys(reason: string, awayMs: number | null): void {* health-check the terminals. Idempotent: overlapping resume+unlock events* collapse safely (clear-then-arm everywhere; at most one catch-up fire). */function onSystemResume(reason: string): void {- console.log(`[power] ${reason} — re-arming scheduler, beats, keep-awake`);+ console.log(`[power] ${reason} — re-arming scheduler, beats, router, keep-awake`);try { syncMissions(); } catch (e) { console.error('[power] syncMissions on resume', e); }try { armAlwaysOnBeats(); } catch (e) { console.error('[power] armAlwaysOnBeats on resume', e); }+ // The hive message router (outbox→inbox drain) is a setInterval that freezes+ // during true system sleep exactly like the beats above — but it was the one+ // always-on timer never re-armed on wake. Symptom: after a long sleep the+ // scheduler→god path recovered (it injects straight into god's inbox), while+ // every agent's outbox silently stopped draining, so god→worker and+ // worker↔worker mail piled up undelivered. Re-arm the poll loop (clear-then-set,+ // idempotent) and immediately drain the backlog that accrued while we were out+ // instead of waiting for the first post-wake tick. The renderer's idle inbox-wake+ // nudge (useHive.ts) then wakes each parked recipient once its mail lands.+ try {+ hive.stopRouter();+ hive.startRouter();+ const drained = hive.routeOnce();+ if (drained > 0) console.log(`[power] ${reason} — flushed ${drained} queued hive message(s)`);+ } catch (e) { console.error('[power] router re-arm on resume', e); }try { syncKeepAwake(); } catch (e) { console.error('[power] syncKeepAwake on resume', e); }const awayMs = lastSuspendAt != null ? Date.now() - lastSuspendAt : null;// Give PTYs a beat to resume their pipes before judging them wedged; reset any