Merge pull request 'Add #loser command — worst packet delivery record (v0.21.9)' (#52) from fix/loser-command into main

Owner request: 'now add #help with short list of all command'. Registered in
the CommandService constructor next to #ping; returns a SINGLE multi-line
string (one packet, ~203 chars) — not an array, which would drip-feed one
spaced packet per line through the v0.21.5 MESH_MULTI_SEND_DELAY_MS loop.

Also hardened the v0.21.5 multi-send-spacing spec: its wall-clock assertion
('first send < 8 ms after t0') flaked under full-suite load — replaced with
fake timers, which assert deterministically that line 1 is out before any time
elapses and lines 2-3 land exactly one delay apart.

AGENTS.md

@@ -27,5 +27,5 @@
 
 - MESHTASTIC.md — rules for Meshtastic communication (owner authority, language, response protocol); read before acting on mesh messages.
 - Hourly cron weather report job (`src/weather`): fetches Open-Meteo current conditions and broadcasts a one-line report to the mesh on the pinned send channel; env `WEATHER_CRON`/`WEATHER_LAT`/`WEATHER_LON`/`WEATHER_CITY`.
-- Channel-2 auto-responder (`src/commands`): event-driven — `CommandService.onRx()` is called by `MeshService` for every decoded rx text (right after the notify gate) and dispatches exact, case-sensitive `#command` matches on `COMMAND_CHANNEL` (env, default 2) through a single-registration-point registry (handlers registered in the constructor via `register(name, fn)`; handler gets `{ text, from?, id? }`, returns a reply string — or an array of lines, sent as one message per line through the same routing — or void = no reply). First command: `#ping` → replies `#pong!`. Replies go through `MeshService.respond(id, reply)`: auto-routed (rx → DM back to the original sender unless the original was a broadcast; tx → its destination), riding the original record's channel with `replyId = id` so the recipient app renders it as a threaded quote — `COMMAND_CHANNEL` is only the trigger/listen gate. Only when the trigger has no packet id does it fall back to legacy sendText routing on `COMMAND_CHANNEL`. A failing handler/respond only logs a warn. DI: provided by `CommandModule`'s factory (injects `MeshService`, attaches itself via `attachCommands()`) — no constructor-level cycle; `forwardRef` was tried and left one side null in NestJS 12's injector. Prefix dispatch (`registerPrefix(name, fn)`) lets one handler own a whole family: the `#rss` subcommands live in `src/rss` (`RssService.handle`, registered as prefix `'rss'` from the CommandModule factory — exact matches still win over prefixes). Subcommands: `#rss help`; `#rss add <url> <name>` (validates URL, fetches with 15 s timeout, requires ≥1 titled item, rejects duplicate names case-insensitively; LLM generates 2-4 lowercase tags from up to 8 latest titles — on LLM failure the source is added with empty tags and the reply says so); `#rss remove <name>`; `#rss tags` (all distinct tags); `#rss sources`; `#rss get <name|#tag>` (LLM headline summary as 3-5 SEPARATE mesh messages: prompt asks for exactly 3-5 lines, one item per line, `<short header>: <one or two short sentences>` with ≤20-char header, no intro/outro/numbering; reply parsed into items — drop empty/separator-less lines, cap at 5, fewer than 3 → send what we got (no retry), zero → no message; each item sent via its own respond() threaded to the trigger packet id, per-line backstop `fitOneLine()` hard-cuts >185-char lines at a word boundary near 180 so every item fits one packet; maxTokens 2500 is headroom because the LLM server's reasoning tokens count against it — see rss.service.ts comment; `#tag` collects items across ALL sources carrying that tag — one dead feed does not sink the summary). Storage: `rss_sources(name TEXT PRIMARY KEY, url, tags, created_at)` table in the EXISTING MESH_DB_PATH file via a second node:sqlite connection (names matched case-insensitively); feed parsing via `fast-xml-parser` (RSS 2.0 + Atom; malformed/no-items → "not a valid feed"); LLM via injected `LlmService.chat()` with explicit maxTokens (50 tags / 2500 summary headroom).
-- LLM client (`src/llm`): `LlmService` wraps the local OpenAI-compatible server (default `http://192.168.1.70:1111/v1`, no auth) with native fetch — plain feature module, no Mesh/Schedule dependency. Methods: `listModels()` → `data[].id`; `chat(messages, { maxTokens?, temperature? })` → `choices[0].message.content` (non-2xx throws with status + body snippet); `summarize(text)` = chat() with system prompt "Summarize the following text concisely." Env: `LLM_API_BASE`, `LLM_MODEL` (default `qwen3.8-27b`), `LLM_TIMEOUT_MS` (default 60000, via `AbortSignal.timeout`).
+- Channel-2 auto-responder (`src/commands`): event-driven — `CommandService.onRx()` is called by `MeshService` for every decoded rx text (right after the notify gate) and dispatches exact, case-sensitive `#command` matches on `COMMAND_CHANNEL` (env, default 2) through a single-registration-point registry (handlers registered in the constructor via `register(name, fn)`; handler gets `{ text, from?, id? }`, returns a reply string — or an array of lines, sent as one message per line through the same routing, spaced `MESH_MULTI_SEND_DELAY_MS` ms apart (default 5000) so each packet has time to land in RF — or void = no reply). First commands: `#ping` → replies `#pong!`; `#loser` → worst packet delivery record over directed sends only (groups our own tx history by destination, counts delivered vs failed per node; broadcasts excluded — firmware strips want_ack so they stay pending and the broadcast addr is never a real destination; loser = most failures, tie-break worse failure ratio; destination resolved to long/short name with `@!xxxxxxxx` mention fallback when unknown; three reply shapes: `worst delivery: <name> (@!hex) - N failed / M delivered`, "no failed sends yet", or "not enough data yet" explaining only directed sends count); `#help` → short single-message list of all available commands. Replies go through `MeshService.respond(id, reply)`: auto-routed (rx → DM back to the original sender unless the original was a broadcast; tx → its destination), riding the original record's channel with `replyId = id` so the recipient app renders it as a threaded quote — `COMMAND_CHANNEL` is only the trigger/listen gate. Only when the trigger has no packet id does it fall back to legacy sendText routing on `COMMAND_CHANNEL`. A failing handler/respond only logs a warn. DI: provided by `CommandModule`'s factory (injects `MeshService`, attaches itself via `attachCommands()`) — no constructor-level cycle; `forwardRef` was tried and left one side null in NestJS 12's injector. Prefix dispatch (`registerPrefix(name, fn)`) lets one handler own a whole family: the `#rss` subcommands live in `src/rss` (`RssService.handle`, registered as prefix `'rss'` from the CommandModule factory — exact matches still win over prefixes). Subcommands: `#rss help`; `#rss add <url> <name>` (validates URL, fetches with 15 s timeout, requires ≥1 titled item, rejects duplicate names case-insensitively; LLM generates 2-4 lowercase tags from up to 8 latest titles — on LLM failure the source is added with empty tags and the reply says so); `#rss remove <name>`; `#rss tags` (all distinct tags); `#rss sources`; `#rss get <name|#tag>` (LLM headline summary as 3-5 SEPARATE mesh messages: prompt asks for exactly 3-5 lines, one item per line, `<short header>: <one or two short sentences>` with ≤20-char header, no intro/outro/numbering; reply parsed into items — drop empty/separator-less lines, cap at 5, fewer than 3 → send what we got (no retry), zero → no message; each item sent via its own respond() threaded to the trigger packet id, per-line backstop `fitOneLine()` hard-cuts >185-char lines at a word boundary near 180 so every item fits one packet; maxTokens 2500 is headroom because the LLM server's reasoning tokens count against it — see rss.service.ts comment; `#tag` collects items across ALL sources carrying that tag — one dead feed does not sink the summary). Storage: `rss_sources(name TEXT PRIMARY KEY, url, tags, created_at)` table in the EXISTING MESH_DB_PATH file via a second node:sqlite connection (names matched case-insensitively); feed parsing via `fast-xml-parser` (RSS 2.0 + Atom; malformed/no-items → "not a valid feed"); LLM via injected `LlmService.chat()` with explicit maxTokens (50 tags / 2500 summary headroom).
+- LLM client (`src/llm`): `LlmService` wraps the local OpenAI-compatible server (default `http://192.168.1.70:1111/v1`, no auth) with native fetch — plain feature module, no Mesh/Schedule dependency. Methods: `listModels()` → `data[].id`; `chat(messages, { maxTokens?, temperature? })` → `choices[0].message.content` (non-2xx throws with status + body snippet); `summarize(text)` = chat() with system prompt "Summarize the following text concisely." Env: `LLM_API_BASE`, `LLM_MODEL` (default `qwen3.8-27b`), `LLM_TIMEOUT_MS` (default 180000, via `AbortSignal.timeout`).

CHANGELOG.md

@@ -1,5 +1,33 @@
 # Changelog
 
+## v0.21.9 — 2026-09-29 20:56
+
+- New `#loser` command (owner request): replies with which user/node has the worst packet delivery record, over directed sends only. Meshtastic shares no mesh-wide loss stats, so it groups our own tx history by destination and counts delivered vs failed per node; broadcasts are excluded (firmware strips want_ack pre-TX, they stay pending — and the broadcast addr is never a real destination even if a NACK lands on it). Loser = most failures, tie-break worse failure ratio. Node resolved to `longName`/`shortName`, falling back to the `@!xxxxxxxx` mention when unknown. Three reply shapes: worst delivery line (`worst delivery: <name> (@!hex) - N failed / M delivered`), a "no failed sends yet" line, and a "not enough data yet — only directed sends count (broadcasts have no acks)" line. Registered in the CommandService constructor next to `#ping`; listed in `#help`.
+- Tests: 175 → 180 — new `#loser` specs cover all three reply cases, broadcast/pending exclusion, the unknown-node `@!hex` fallback, and the failure-ratio tie-break; the `#help` spec now also asserts `#loser` is listed.
+
+## v0.21.8 — 2026-09-27 07:13
+
+- Fixed `npm test` exiting 1 with "Errors 2" even though all tests pass (every CI run red since v0.21.2): the two watchdog specs that force a reconnect destroy the service's socket while data is still pending in its receive buffer (the overflow spec floods 300 KB and destroys on the first ~64 KB chunk; the garbage spec destroys right after its last write), so the kernel sends RST instead of FIN and the fake bridge's accepted server-side sockets emit `read ECONNRESET`. Those test-owned sockets had no `'error'` handler, so Node threw it as an uncaught exception that vitest counts as a run-level failure. `startBridge()` now attaches a no-op `'error'` handler to every accepted socket (single registration point, covers all four spec files using the fake bridge). Test-only change — `fake-bridge.ts` is imported by specs only; no runtime diff, no deploy needed.
+
+## v0.21.7 — 2026-09-27 06:50
+
+- Fixed double registration of scheduled jobs: `ScheduleModule.forRoot()` was called from both `WeatherModule` and `BatteryAlertModule`, so Nest created two ScheduleExplorers, each scanning all providers — every @Cron/@Interval job (hourly weather broadcast, per-minute battery check) fired twice. Removed the call from `battery.module.ts`; the single registration in `weather.module.ts` remains.
+- Tests: 174 → 175 — new regression spec (`src/schedule-singleton.spec.ts`) asserts `ScheduleModule.forRoot(` appears exactly once across all `src/**/*.module.ts`.
+
+## v0.21.6 — 2026-09-26 10:37
+
+- New `#help` command (owner request): replies with a short list of all available commands as ONE multi-line message (single packet, not one spaced packet per line) — `#ping`, the `#rss help/add/remove/get/sources/tags` family. Registered in the CommandService constructor next to `#ping`.
+- Tests: 173 → 174 — `#help` dispatches through respond() exactly once with text containing both `#ping` and `#rss`. (Also hardened the v0.21.5 multi-send-spacing spec, which flaked under full-suite load: it now uses fake timers to assert deterministically that line 1 goes out before any time elapses and lines 2-3 land exactly one delay apart.)
+
+## v0.21.5 — 2026-09-26 10:26
+
+- Multi-line command replies are now spaced apart: `CommandService.onRx` waits `MESH_MULTI_SEND_DELAY_MS` (new env, default **5000** = 5 s, read at construction) between consecutive lines of an array reply — no delay before the first line. Production incident: `#rss get lublin` sent its items back-to-back within the same second and 3 of 4 were dropped in RF (the owner saw only the last one); each packet now gets time to land.
+- Tests: 172 → 173 — a 3-line reply with an overridden short delay asserts all lines are sent, the first goes out immediately (no pre-delay), and each subsequent line lands ≥ the configured gap after the previous one; existing specs pin `MESH_MULTI_SEND_DELAY_MS=0` in `beforeEach` so they stay fast.
+
+## v0.21.4 — 2026-09-26 09:51
+
+- `LLM_TIMEOUT_MS` default raised from 60000 (60 s) to **180000** (3 min): production incident — `#rss get lublin` at 9:41 AM failed with "The operation was aborted due to timeout" because the local OpenAI-compatible server (qwen3.8-27b, uncontrollable reasoning tokens) was slow/busy on that call; the same endpoint answers a trivial prompt in ~1 s when idle, so this is latency variance, not a dead endpoint. Still env-overridable (`LLM_TIMEOUT_MS`), only the fallback changed.
+
 ## v0.21.3 — 2026-09-26 09:31
 
 - Idle timeout default lowered from 10 min to **1 min** (`MESH_IDLE_TIMEOUT_MS` fallback `600000` → `60000`) — the v0.21.2 frame-based watchdog works in production (it fired today and the reconnect delivered queued packets), but the bridge/radio side at 192.168.1.4 keeps killing inbound RX within ~20 min of a fresh sync, so a shorter detection window is wanted.

package.json

@@ -1,6 +1,6 @@
 {
   "name": "mesh-bot",
-  "version": "0.21.3",
+  "version": "0.21.9",
   "description": "",
   "author": "",
   "private": true,

src/battery/battery.module.ts

@@ -1,10 +1,9 @@
 import { Module } from '@nestjs/common';
-import { ScheduleModule } from '@nestjs/schedule';
 import { MeshModule } from '../mesh/mesh.module.js';
 import { BatteryAlertService } from './battery.service.js';
 
 @Module({
-  imports: [ScheduleModule.forRoot(), MeshModule],
+  imports: [MeshModule],
   providers: [BatteryAlertService],
   exports: [BatteryAlertService],
 })

src/commands/command.service.spec.ts

@@ -13,7 +13,11 @@ import {
   KAMIL_NUM,
 } from '../mesh/fake-bridge.js';
 import { BROADCAST_ADDR } from '../mesh/codec.js';
-import type { MeshService } from '../mesh/mesh.service.js';
+import type {
+  MessageRecord,
+  MeshService,
+  NodeSummary,
+} from '../mesh/mesh.service.js';
 import { MeshService as RealMeshService } from '../mesh/mesh.service.js';
 import { CommandService } from './command.service.js';
 
@@ -30,6 +34,7 @@ function makeService(
 
 beforeEach(() => {
   delete process.env.COMMAND_CHANNEL; // default channel 2 under test
+  process.env.MESH_MULTI_SEND_DELAY_MS = '0'; // no spacing under test unless a test opts in — the real default (5000) would slow the suite down
 });
 
 describe('CommandService.onRx', () => {
@@ -40,6 +45,16 @@ describe('CommandService.onRx', () => {
     expect(mesh.sendText).not.toHaveBeenCalled();
   });
 
+  it('replies #help with a short single-message list of all commands', async () => {
+    const [svc, mesh] = makeService();
+    await svc.onRx(2, '#help', 42, 7);
+    expect(mesh.respond).toHaveBeenCalledTimes(1); // one packet — not one per line
+    const text = mesh.respond.mock.calls[0][1];
+    expect(text).toContain('#ping');
+    expect(text).toContain('#loser');
+    expect(text).toContain('#rss');
+  });
+
   it('falls back to a directed sendText on COMMAND_CHANNEL when the trigger has no packet id (DM)', async () => {
     const [svc, mesh] = makeService();
     await svc.onRx(2, '#ping', 42); // from present but no id — defensive path
@@ -124,6 +139,37 @@ describe('CommandService.onRx', () => {
     expect(mesh.sendText).not.toHaveBeenCalled();
   });
 
+  it('spaces consecutive multi-line sends by MESH_MULTI_SEND_DELAY_MS — no delay before the first line', async () => {
+    process.env.MESH_MULTI_SEND_DELAY_MS = '50'; // short so the test stays fast
+    let sends = 0;
+    const respond = vi.fn(async () => {
+      sends++;
+      return { id: 1 };
+    });
+    vi.useFakeTimers();
+    try {
+      const [svc, mesh] = makeService(undefined, respond);
+      svc.register('multi3', () => ['a: one', 'b: two', 'c: three']);
+
+      const p = svc.onRx(2, '#multi3', 42, 9);
+      // flush microtasks without advancing time — the first line must already be out (no delay before it)
+      await vi.advanceTimersByTimeAsync(0);
+      expect(sends).toBe(1);
+
+      // each subsequent line waits exactly MESH_MULTI_SEND_DELAY_MS after the previous one landed
+      await vi.advanceTimersByTimeAsync(50);
+      expect(sends).toBe(2);
+      await vi.advanceTimersByTimeAsync(50);
+      expect(sends).toBe(3);
+
+      await p;
+      expect(mesh.respond).toHaveBeenCalledTimes(3);
+    } finally {
+      vi.useRealTimers();
+      delete process.env.MESH_MULTI_SEND_DELAY_MS;
+    }
+  });
+
   it('sends each line of an array reply via the legacy sendText fallback when there is no packet id', async () => {
     const [svc, mesh] = makeService();
     svc.register('multi', () => ['one: first item', 'two: second item']);
@@ -164,6 +210,125 @@ describe('CommandService.onRx', () => {
   });
 });
 
+describe('#loser', () => {
+  const tx = (
+    id: number,
+    to: number,
+    status: MessageRecord['status'],
+  ): MessageRecord => ({
+    id,
+    dir: 'tx',
+    to,
+    text: 'hi',
+    channel: 3,
+    at: Date.now(),
+    status,
+  });
+
+  const makeLoserMesh = (
+    messages: MessageRecord[],
+    nodes: NodeSummary[],
+  ): [CommandService, MeshService] => {
+    const mesh = {
+      sendText: vi.fn(),
+      respond: vi.fn().mockResolvedValue({ id: 1 }),
+      messages: () => messages,
+      nodes: async () => nodes,
+    } as unknown as MeshService;
+    return [new CommandService(mesh), mesh];
+  };
+
+  it('reports the node with the most failures — named via longName, broadcasts and pending sends excluded', async () => {
+    const messages = [
+      tx(10, BROADCAST_ADDR, 'failed'), // firmware quirk: a broadcast NACK is never a real destination
+      tx(9, 222, 'failed'), // Alice: 3 failed / 0 delivered → loser
+      tx(8, 111, 'delivered'), // Bob: 1 failed / 2 delivered
+      tx(7, 222, 'failed'),
+      tx(6, 111, 'failed'),
+      tx(5, 333, 'pending'), // no final status yet — ignored
+      tx(4, 222, 'failed'),
+      {
+        id: 3,
+        dir: 'rx',
+        from: 999,
+        text: 'hello',
+        channel: 0,
+        at: Date.now(),
+      },
+    ];
+    const nodes = [
+      { num: 111, shortName: 'Bob', lastHeard: null },
+      { num: 222, longName: 'Alice', lastHeard: null },
+    ];
+    const [svc, mesh] = makeLoserMesh(messages, nodes);
+
+    await svc.onRx(2, '#loser', 42, 7);
+
+    expect(mesh.respond).toHaveBeenCalledWith(
+      7,
+      'worst delivery: Alice (@!000000de) - 3 failed / 0 delivered',
+    );
+    expect(mesh.sendText).not.toHaveBeenCalled();
+  });
+
+  it('says no failed sends when the directed history is all delivered', async () => {
+    const messages = [tx(2, 5, 'delivered'), tx(1, 6, 'delivered')];
+    const [svc, mesh] = makeLoserMesh(messages, []);
+
+    await svc.onRx(2, '#loser', 42, 7);
+
+    expect(mesh.respond).toHaveBeenCalledWith(
+      7,
+      'no failed sends yet - everyone delivers :)',
+    );
+  });
+
+  it('explains the directed-only rule when there is no final-status data at all', async () => {
+    const messages = [tx(2, BROADCAST_ADDR, 'pending'), tx(1, 5, 'pending')];
+    const [svc, mesh] = makeLoserMesh(messages, []);
+
+    await svc.onRx(2, '#loser', 42, 7);
+
+    expect(mesh.respond).toHaveBeenCalledWith(
+      7,
+      'not enough data yet - only directed sends count (broadcasts have no acks)',
+    );
+  });
+
+  it('falls back to the @!hex mention when the worst node has no known name', async () => {
+    const messages = [tx(1, 600, 'failed')];
+    const nodes = [{ num: 5, longName: 'Someone Else', lastHeard: null }];
+    const [svc, mesh] = makeLoserMesh(messages, nodes);
+
+    await svc.onRx(2, '#loser', 42, 7);
+
+    expect(mesh.respond).toHaveBeenCalledWith(
+      7,
+      'worst delivery: @!00000258 - 1 failed / 0 delivered',
+    );
+  });
+
+  it('tie-breaks equal failure counts by the worse failure ratio', async () => {
+    const messages = [
+      // node 7: 1 failed / 9 delivered (ratio 0.1) — not the loser
+      tx(16, 8, 'failed'), // node 8: 1 failed / 1 delivered (ratio 0.5) → loser
+      tx(15, 8, 'delivered'),
+      tx(14, 7, 'failed'),
+      ...Array.from({ length: 9 }, (_, i) =>
+        tx(i + 1, 7, 'delivered' as const),
+      ),
+    ];
+    const [svc, mesh] = makeLoserMesh(messages, []);
+
+    await svc.onRx(2, '#loser', 42, 7);
+
+    expect(mesh.respond).toHaveBeenCalledWith(
+      7,
+      'worst delivery: @!00000008 - 1 failed / 1 delivered',
+    );
+  });
+});
+
 describe('#rss prefix dispatch', () => {
   let dir: string | undefined;
 

src/commands/command.service.ts

@@ -1,4 +1,6 @@
 import { Injectable, Logger } from '@nestjs/common';
+// pure formatting helper (no cycle: battery → mesh only)
+import { mention } from '../battery/battery.service.js';
 import { BROADCAST_ADDR } from '../mesh/codec.js';
 // type-only: the instances come from DI (command.module.ts factory) or direct construction in tests — no runtime import cycles with src/mesh / src/rss
 import type { MeshService } from '../mesh/mesh.service.js';
@@ -27,6 +29,11 @@ export class CommandService {
     process.env.COMMAND_CHANNEL ?? '2',
     10,
   );
+  // spacing between consecutive multi-line reply messages — back-to-back packets get dropped in RF on a real mesh; each packet needs time to land. Read at construction like the other env config.
+  private readonly multiSendDelayMs = parseInt(
+    process.env.MESH_MULTI_SEND_DELAY_MS ?? '5000',
+    10,
+  );
   // single registration point: the constructor registers every built-in command — add the next one there (or via register())
   private readonly commands = new Map<string, CommandHandler>();
   // prefix handlers own a whole family (`#rss`, `#rss add ...`) instead of one exact name
@@ -38,9 +45,68 @@ export class CommandService {
     rss?: RssService,
   ) {
     this.register('ping', () => '#pong!');
+    this.register('loser', () => this.loserReply());
+    // single multi-line string = ONE packet — not an array, which would drip-feed one spaced packet per line
+    this.register('help', () =>
+      [
+        '#ping - alive check',
+        '#loser - worst delivery record',
+        '#rss help - rss subcommands',
+        '#rss add <url> <name> - add feed with AI tags',
+        '#rss remove <name> - drop feed',
+        '#rss get <name|#tag> - news summary',
+        '#rss sources / #rss tags - list feeds/tags',
+      ].join('\n'),
+    );
     if (rss) this.registerPrefix('rss', (ctx) => rss.handle(ctx));
   }
 
+  /**
+   * #loser — worst packet delivery record, over directed sends only. Meshtastic shares no mesh-wide
+   * loss stats, so the only real data is our own tx history: group by destination, count delivered vs
+   * failed per node; broadcasts are excluded (firmware strips want_ack pre-TX, they stay pending — and
+   * the broadcast addr is never a real destination even if a NACK ever lands on it). Loser = most
+   * failures, tie-break worse failure ratio. One reply line, ≤ ~180 chars.
+   */
+  private async loserReply(): Promise<string> {
+    const mesh = this.mesh;
+    if (!mesh)
+      return 'not enough data yet - only directed sends count (broadcasts have no acks)';
+    const stats = new Map<number, { failed: number; delivered: number }>();
+    for (const m of mesh.messages()) {
+      if (m.dir !== 'tx' || m.to === undefined || m.to === BROADCAST_ADDR)
+        continue; // directed sends only
+      if (m.status !== 'delivered' && m.status !== 'failed') continue; // pending = no final status yet
+      const s = stats.get(m.to) ?? { failed: 0, delivered: 0 };
+      if (m.status === 'failed') s.failed += 1;
+      else s.delivered += 1;
+      stats.set(m.to, s);
+    }
+    if (stats.size === 0)
+      return 'not enough data yet - only directed sends count (broadcasts have no acks)';
+    let worst: { num: number; failed: number; total: number } | undefined;
+    for (const [num, s] of stats) {
+      if (s.failed === 0) continue; // zero-failure nodes can't be the loser
+      const total = s.failed + s.delivered;
+      // most failures first, then worse failure ratio on a tie
+      if (
+        !worst ||
+        s.failed > worst.failed ||
+        (s.failed === worst.failed &&
+          s.failed / total > worst.failed / worst.total)
+      ) {
+        worst = { num, failed: s.failed, total };
+      }
+    }
+    if (!worst) return 'no failed sends yet - everyone delivers :)';
+    const names = new Map(
+      (await mesh.nodes()).map((n) => [n.num, n.longName ?? n.shortName]),
+    );
+    const name = names.get(worst.num);
+    const who = name ? `${name} (${mention(worst.num)})` : mention(worst.num);
+    return `worst delivery: ${who} - ${worst.failed} failed / ${worst.total - worst.failed} delivered`;
+  }
+
   /** Register a command handler (names are case-sensitive; the leading `#` is stripped before lookup). */
   register(name: string, fn: CommandHandler): void {
     this.commands.set(name, fn);
@@ -83,8 +149,15 @@ export class CommandService {
     try {
       const reply = await handler({ text: trimmed, from, id });
       if (reply === undefined || reply === null) return; // void = no reply
-      // an array reply sends one message per line, all threaded to the same trigger packet id
-      for (const line of Array.isArray(reply) ? reply : [reply]) {
+      // an array reply sends one message per line, all threaded to the same trigger packet id — spaced so each packet has time to land in RF (no delay before the first line)
+      const lines = Array.isArray(reply) ? reply : [reply];
+      for (let i = 0; i < lines.length; i++) {
+        if (i > 0 && this.multiSendDelayMs > 0) {
+          await new Promise((resolve) =>
+            setTimeout(resolve, this.multiSendDelayMs),
+          );
+        }
+        const line = lines[i];
         if (id !== undefined && this.mesh) {
           // respond() auto-routes each line: rides the original record's channel with replyId = id (threaded quote)
           await this.mesh.respond(id, line);

src/llm/llm.service.ts

@@ -12,9 +12,9 @@ export class LlmService {
   private readonly apiBase =
     process.env.LLM_API_BASE ?? 'http://192.168.1.70:1111/v1';
   private readonly model = process.env.LLM_MODEL ?? 'qwen3.8-27b';
-  // local 27B model can be slow — generous default
+  // local 27B model can be slow (uncontrollable reasoning tokens) — generous default: 3 min headroom for slow reasoning periods
   private readonly timeoutMs = parseInt(
-    process.env.LLM_TIMEOUT_MS ?? '60000',
+    process.env.LLM_TIMEOUT_MS ?? '180000',
     10,
   );
 

src/mesh/fake-bridge.ts

@@ -10,6 +10,8 @@ export function startBridge(
 ): Promise<{ port: number; sockets: net.Socket[]; close: () => void }> {
   const sockets: net.Socket[] = [];
   const server = net.createServer((socket) => {
+    // Test-owned socket: the service destroys its end mid-test with unread data pending (overflow/garbage tests), which makes the kernel send RST and this side emit ECONNRESET. Expected teardown noise — without a handler Node throws it as an uncaught exception and vitest fails the whole run.
+    socket.on('error', () => {});
     onSocket?.(socket); // attached at accept time, before any data can arrive
     sockets.push(socket);
   });

src/schedule-singleton.spec.ts

@@ -0,0 +1,28 @@
+import { readdirSync, readFileSync } from 'node:fs';
+import { join } from 'node:path';
+import { fileURLToPath } from 'node:url';
+import { describe, expect, it } from 'vitest';
+
+const srcDir = fileURLToPath(new URL('.', import.meta.url));
+
+function moduleFiles(dir: string): string[] {
+  return readdirSync(dir, { withFileTypes: true }).flatMap((entry) =>
+    entry.isDirectory()
+      ? moduleFiles(join(dir, entry.name))
+      : entry.name.endsWith('.module.ts')
+        ? [join(dir, entry.name)]
+        : [],
+  );
+}
+
+describe('ScheduleModule.forRoot single registration', () => {
+  // Regression guard (v0.21.7): each forRoot() call creates its own ScheduleExplorer that scans all providers, so a second call made every @Cron/@Interval job fire twice.
+  it('appears exactly once across all src/**/*.module.ts', () => {
+    const hits = moduleFiles(srcDir).flatMap((file) =>
+      (
+        readFileSync(file, 'utf8').match(/ScheduleModule\.forRoot\(/g) ?? []
+      ).map(() => file.slice(srcDir.length)),
+    );
+    expect(hits).toHaveLength(1);
+  });
+});