From 67e12a47a4bdf55452b98aff69911338c3229640 Mon Sep 17 00:00:00 2001
From: DCCONSTRUCTIONS <dcconstructions@MacBook-Pro-DCCONSTRUCTIONS-2.local>
Date: Mon, 27 Jul 2026 21:06:51 +0300
Subject: [PATCH] feat(system): add Worker 006 telemetry and network profile

---
 apps/control-station/src/App.tsx              |   4 +
 .../src/components/system/TelemetrySeries.tsx |  47 +
 .../components/system/WorkerRuntimeCard.tsx   |  50 ++
 .../src/components/system/systemFormat.ts     |  67 ++
 .../src/core/system/useWorkerTelemetry.ts     |  47 +
 .../src/core/system/workerTelemetry.ts        | 249 ++++++
 apps/control-station/src/productModel.ts      |  14 +-
 apps/control-station/src/styles.css           |   1 +
 .../src/styles/system-telemetry.css           | 544 ++++++++++++
 .../src/workspaces/Workspaces.tsx             |   6 +
 .../system/ComputeModulesWorkspace.tsx        | 226 +++++
 .../workspaces/system/NetworkWorkspace.tsx    | 311 +++++++
 .../test/systemTelemetry.test.mjs             |  40 +
 .../worker/run_e15_shadow_inference.py        |  48 +-
 src/k1link/web/app.py                         |   6 +
 src/k1link/web/system_telemetry_api.py        | 834 ++++++++++++++++++
 tests/test_system_telemetry_api.py            | 237 +++++
 17 files changed, 2723 insertions(+), 8 deletions(-)
 create mode 100644 apps/control-station/src/components/system/TelemetrySeries.tsx
 create mode 100644 apps/control-station/src/components/system/WorkerRuntimeCard.tsx
 create mode 100644 apps/control-station/src/components/system/systemFormat.ts
 create mode 100644 apps/control-station/src/core/system/useWorkerTelemetry.ts
 create mode 100644 apps/control-station/src/core/system/workerTelemetry.ts
 create mode 100644 apps/control-station/src/styles/system-telemetry.css
 create mode 100644 apps/control-station/src/workspaces/system/ComputeModulesWorkspace.tsx
 create mode 100644 apps/control-station/src/workspaces/system/NetworkWorkspace.tsx
 create mode 100644 apps/control-station/test/systemTelemetry.test.mjs
 create mode 100644 src/k1link/web/system_telemetry_api.py
 create mode 100644 tests/test_system_telemetry_api.py

diff --git a/apps/control-station/src/App.tsx b/apps/control-station/src/App.tsx
index 04da62a..0cfc3d6 100644
--- a/apps/control-station/src/App.tsx
+++ b/apps/control-station/src/App.tsx
@@ -795,6 +795,10 @@ export default function App() {
                 <StatusBadge tone="neutral">Offline evaluation</StatusBadge>
               ) : activeDefinition.kind === "lab-archive" ? (
                 null
+              ) : activeDefinition.kind === "compute-modules" ? (
+                <StatusBadge tone="neutral">Живая телеметрия</StatusBadge>
+              ) : activeDefinition.kind === "network-monitor" ? (
+                <StatusBadge tone="neutral">Живая сеть</StatusBadge>
               ) : (
                 <StatusBadge tone="warning">Интерфейс готов</StatusBadge>
               )
diff --git a/apps/control-station/src/components/system/TelemetrySeries.tsx b/apps/control-station/src/components/system/TelemetrySeries.tsx
new file mode 100644
index 0000000..fbbed41
--- /dev/null
+++ b/apps/control-station/src/components/system/TelemetrySeries.tsx
@@ -0,0 +1,47 @@
+interface TelemetrySeriesProps {
+  label: string;
+  values: Array<number | null>;
+  value: string;
+  ceiling?: number;
+}
+
+function linePoints(values: Array<number | null>, ceiling?: number): string {
+  const finite = values.filter((value): value is number =>
+    value !== null && Number.isFinite(value),
+  );
+  if (finite.length === 0) return "";
+  const maximum = Math.max(ceiling ?? 0, ...finite, 1);
+  const denominator = Math.max(1, values.length - 1);
+  return values.map((value, index) => {
+    const x = index / denominator * 100;
+    const normalized = value === null ? 0 : Math.max(0, Math.min(maximum, value));
+    const y = 36 - normalized / maximum * 34;
+    return `${x.toFixed(2)},${y.toFixed(2)}`;
+  }).join(" ");
+}
+
+export function TelemetrySeries({
+  label,
+  values,
+  value,
+  ceiling,
+}: TelemetrySeriesProps) {
+  const points = linePoints(values, ceiling);
+  return (
+    <div className="system-telemetry-series">
+      <div>
+        <span>{label}</span>
+        <strong>{value}</strong>
+      </div>
+      <svg
+        viewBox="0 0 100 38"
+        preserveAspectRatio="none"
+        role="img"
+        aria-label={`${label}: ${value}`}
+      >
+        <path d="M0 37 H100" />
+        {points ? <polyline points={points} /> : null}
+      </svg>
+    </div>
+  );
+}
diff --git a/apps/control-station/src/components/system/WorkerRuntimeCard.tsx b/apps/control-station/src/components/system/WorkerRuntimeCard.tsx
new file mode 100644
index 0000000..7eb0bc4
--- /dev/null
+++ b/apps/control-station/src/components/system/WorkerRuntimeCard.tsx
@@ -0,0 +1,50 @@
+import { GlassSurface, StatusBadge } from "@nodedc/ui-react";
+
+import type { WorkerRuntime } from "../../core/system/workerTelemetry";
+import {
+  formatOptionalPercent,
+  runtimeStateLabel,
+  runtimeTone,
+} from "./systemFormat";
+
+export function WorkerRuntimeCard({ runtime }: { runtime: WorkerRuntime }) {
+  return (
+    <GlassSurface
+      className="worker-runtime-card"
+      padding="md"
+      tone={runtime.external ? "soft" : "default"}
+      data-external={runtime.external ? "true" : undefined}
+    >
+      <header>
+        <div>
+          <span className="section-eyebrow">
+            {runtime.external ? "ВНЕШНЯЯ НАГРУЗКА" : "MISSION CORE RUNTIME"}
+          </span>
+          <h3>{runtime.role}</h3>
+          <code>{runtime.name}</code>
+        </div>
+        <StatusBadge tone={runtimeTone(runtime)}>
+          {runtimeStateLabel(runtime)}
+        </StatusBadge>
+      </header>
+      <dl>
+        <div>
+          <dt>CPU</dt>
+          <dd>{formatOptionalPercent(runtime.cpu_percent)}</dd>
+        </div>
+        <div>
+          <dt>Память</dt>
+          <dd>{runtime.memory_usage ?? "—"}</dd>
+        </div>
+        <div>
+          <dt>Сеть I/O</dt>
+          <dd>{runtime.network_io ?? "—"}</dd>
+        </div>
+        <div>
+          <dt>Процессы</dt>
+          <dd>{runtime.pids ?? "—"}</dd>
+        </div>
+      </dl>
+    </GlassSurface>
+  );
+}
diff --git a/apps/control-station/src/components/system/systemFormat.ts b/apps/control-station/src/components/system/systemFormat.ts
new file mode 100644
index 0000000..056774c
--- /dev/null
+++ b/apps/control-station/src/components/system/systemFormat.ts
@@ -0,0 +1,67 @@
+import type { StatusTone } from "@nodedc/ui-react";
+
+import type { WorkerRuntime } from "../../core/system/workerTelemetry";
+
+export function formatOptionalPercent(value: number | null | undefined): string {
+  return typeof value === "number" && Number.isFinite(value)
+    ? `${new Intl.NumberFormat("ru-RU", { maximumFractionDigits: 1 }).format(value)}%`
+    : "—";
+}
+
+export function formatBytes(value: number | null | undefined): string {
+  if (typeof value !== "number" || !Number.isFinite(value)) return "—";
+  const units = ["Б", "КиБ", "МиБ", "ГиБ", "ТиБ"];
+  let current = Math.max(0, value);
+  let unit = 0;
+  while (current >= 1024 && unit < units.length - 1) {
+    current /= 1024;
+    unit += 1;
+  }
+  return `${new Intl.NumberFormat("ru-RU", {
+    maximumFractionDigits: current >= 100 ? 0 : 1,
+  }).format(current)} ${units[unit]}`;
+}
+
+export function formatRate(value: number | null | undefined): string {
+  const formatted = formatBytes(value);
+  return formatted === "—" ? formatted : `${formatted}/с`;
+}
+
+export function formatBitRate(value: number | null | undefined): string {
+  if (typeof value !== "number" || !Number.isFinite(value)) return "—";
+  const units = ["бит/с", "Кбит/с", "Мбит/с", "Гбит/с"];
+  let current = Math.max(0, value);
+  let unit = 0;
+  while (current >= 1000 && unit < units.length - 1) {
+    current /= 1000;
+    unit += 1;
+  }
+  return `${new Intl.NumberFormat("ru-RU", {
+    maximumFractionDigits: current >= 100 ? 0 : 1,
+  }).format(current)} ${units[unit]}`;
+}
+
+export function formatDuration(value: number | null | undefined): string {
+  if (typeof value !== "number" || !Number.isFinite(value)) return "—";
+  const hours = Math.floor(value / 3600);
+  if (hours < 24) return `${hours} ч`;
+  return `${Math.floor(hours / 24)} д ${hours % 24} ч`;
+}
+
+export function formatLatency(value: number | null | undefined): string {
+  if (typeof value !== "number" || !Number.isFinite(value)) return "—";
+  return `${Math.round(value)} мс`;
+}
+
+export function runtimeTone(runtime: WorkerRuntime): StatusTone {
+  if (runtime.state !== "running") return "danger";
+  if (runtime.health && runtime.health !== "healthy") return "warning";
+  return runtime.external ? "neutral" : "success";
+}
+
+export function runtimeStateLabel(runtime: WorkerRuntime): string {
+  if (runtime.state !== "running") return "Недоступен";
+  if (runtime.health === "healthy") return "Работает";
+  if (runtime.health) return runtime.health;
+  return "Запущен";
+}
diff --git a/apps/control-station/src/core/system/useWorkerTelemetry.ts b/apps/control-station/src/core/system/useWorkerTelemetry.ts
new file mode 100644
index 0000000..794fc34
--- /dev/null
+++ b/apps/control-station/src/core/system/useWorkerTelemetry.ts
@@ -0,0 +1,47 @@
+import { useCallback, useEffect, useState } from "react";
+
+import {
+  fetchWorkerTelemetry,
+  type WorkerTelemetry,
+} from "./workerTelemetry";
+
+export interface WorkerTelemetryState {
+  telemetry: WorkerTelemetry | null;
+  loading: boolean;
+  error: string | null;
+  refresh: () => void;
+}
+
+export function useWorkerTelemetry(pollMilliseconds = 10_000): WorkerTelemetryState {
+  const [telemetry, setTelemetry] = useState<WorkerTelemetry | null>(null);
+  const [loading, setLoading] = useState(true);
+  const [error, setError] = useState<string | null>(null);
+  const [generation, setGeneration] = useState(0);
+  const refresh = useCallback(() => setGeneration((value) => value + 1), []);
+
+  useEffect(() => {
+    const controller = new AbortController();
+    setLoading(true);
+    void fetchWorkerTelemetry(controller.signal)
+      .then((document) => {
+        if (controller.signal.aborted) return;
+        setTelemetry(document);
+        setError(null);
+      })
+      .catch((reason: unknown) => {
+        if (controller.signal.aborted) return;
+        setError(reason instanceof Error ? reason.message : "Worker 006 не ответил.");
+      })
+      .finally(() => {
+        if (!controller.signal.aborted) setLoading(false);
+      });
+    return () => controller.abort();
+  }, [generation]);
+
+  useEffect(() => {
+    const timer = window.setInterval(refresh, pollMilliseconds);
+    return () => window.clearInterval(timer);
+  }, [pollMilliseconds, refresh]);
+
+  return { telemetry, loading, error, refresh };
+}
diff --git a/apps/control-station/src/core/system/workerTelemetry.ts b/apps/control-station/src/core/system/workerTelemetry.ts
new file mode 100644
index 0000000..326663b
--- /dev/null
+++ b/apps/control-station/src/core/system/workerTelemetry.ts
@@ -0,0 +1,249 @@
+export interface WorkerConnectionProfile {
+  schema_version: "missioncore.worker-connection-profile/v1";
+  profile_id: string;
+  display_name: string;
+  expected_node_id: string;
+  ssh_host_alias: string;
+  address: string;
+  port: number;
+  revision: number;
+  updated_at_utc: string | null;
+}
+
+export interface WorkerProbe {
+  schema_version: "missioncore.worker-probe/v1";
+  reachable: boolean;
+  identity_matches: boolean;
+  node_id: string | null;
+  latency_ms: number | null;
+  observed_at_utc: string;
+  error_code: string | null;
+}
+
+export interface WorkerProfileDocument {
+  schema_version: "missioncore.worker-connection-profile/v1";
+  profile: WorkerConnectionProfile;
+  security: {
+    transport: "ssh";
+    host_key_policy: "strict-pinned";
+    credentials_managed_by_ui: false;
+    ssh_host_alias: string;
+  };
+}
+
+export interface WorkerRuntime {
+  name: string;
+  role: string;
+  owner: "mission-core" | "external";
+  external: boolean;
+  image: string | null;
+  state: string;
+  health: string | null;
+  cpu_percent: number | null;
+  memory_percent: number | null;
+  memory_usage: string | null;
+  network_io: string | null;
+  block_io: string | null;
+  pids: number | null;
+}
+
+export interface WorkerPipelineStage {
+  id: string;
+  label: string;
+  state: "active" | "waiting" | "ready" | "unavailable";
+}
+
+export interface WorkerTelemetryHistoryRow {
+  observed_at_utc: string;
+  cpu_percent: number | null;
+  memory_percent: number | null;
+  gpu_percent: number | null;
+  gpu_memory_percent: number | null;
+  network_receive_bytes_per_second: number | null;
+  network_send_bytes_per_second: number | null;
+}
+
+export interface WorkerNetworkInterface {
+  name: string;
+  description: string | null;
+  status: string | null;
+  link_speed_bps: number | null;
+  mac_address: string | null;
+  addresses: string[];
+  received_bytes: number | null;
+  sent_bytes: number | null;
+}
+
+export interface WorkerTelemetry {
+  schema_version: "missioncore.worker-telemetry/v1";
+  profile: WorkerConnectionProfile;
+  connection: WorkerProbe;
+  node: {
+    node_id: string;
+    observed_at_utc: string;
+    os: {
+      caption?: string;
+      version?: string;
+      uptime_seconds?: number;
+    };
+    cpu: {
+      name?: string;
+      logical_processors?: number;
+      load_percent?: number;
+    };
+    memory: {
+      total_bytes: number | null;
+      used_bytes: number | null;
+      free_bytes: number | null;
+      used_percent: number | null;
+    };
+    disks: Array<{
+      name?: string;
+      size_bytes?: number;
+      free_bytes?: number;
+    }>;
+    gpu: {
+      name?: string;
+      utilization_percent?: number;
+      memory_used_mib?: number;
+      memory_total_mib?: number;
+      power_watts?: number;
+      temperature_celsius?: number;
+      memory_used_percent?: number;
+    } | null;
+    triton: {
+      ready: boolean;
+      requests_succeeded: number | null;
+      requests_failed: number | null;
+      inferences: number | null;
+    };
+  } | null;
+  runtimes: WorkerRuntime[];
+  pipeline: {
+    service_state: string;
+    current_stage: string | null;
+    active_request_id: string | null;
+    active_frame_index: number | null;
+    completed_runs: number | null;
+    failed_runs: number | null;
+    model_load_seconds: number | null;
+    stages: WorkerPipelineStage[];
+  };
+  network: {
+    interfaces: WorkerNetworkInterface[];
+    aggregate: {
+      received_bytes: number;
+      sent_bytes: number;
+      receive_bytes_per_second: number | null;
+      send_bytes_per_second: number | null;
+    } | null;
+  };
+  history: WorkerTelemetryHistoryRow[];
+}
+
+export interface WorkerProfileMutation {
+  revision: number;
+  address: string;
+  port: number;
+}
+
+function isRecord(value: unknown): value is Record<string, unknown> {
+  return Boolean(value) && typeof value === "object" && !Array.isArray(value);
+}
+
+function requireJsonRecord(value: unknown, schema: string): Record<string, unknown> {
+  if (!isRecord(value) || value.schema_version !== schema) {
+    throw new Error("Ответ системной телеметрии не соответствует контракту.");
+  }
+  return value;
+}
+
+async function requestJson(
+  url: string,
+  init: RequestInit,
+): Promise<unknown> {
+  const response = await fetch(url, {
+    ...init,
+    headers: {
+      Accept: "application/json",
+      ...(init.body ? { "Content-Type": "application/json" } : {}),
+      ...init.headers,
+    },
+  });
+  if (!response.ok) {
+    throw new Error(`Системный API вернул HTTP ${response.status}.`);
+  }
+  return response.json();
+}
+
+export async function fetchWorkerTelemetry(
+  signal?: AbortSignal,
+): Promise<WorkerTelemetry> {
+  const document = requireJsonRecord(
+    await requestJson("/api/v1/system/worker-telemetry?history=90", {
+      method: "GET",
+      signal,
+    }),
+    "missioncore.worker-telemetry/v1",
+  );
+  if (
+    !isRecord(document.connection)
+    || !isRecord(document.profile)
+    || !Array.isArray(document.runtimes)
+    || !isRecord(document.pipeline)
+    || !isRecord(document.network)
+    || !Array.isArray(document.history)
+  ) {
+    throw new Error("Срез Worker 006 неполон.");
+  }
+  return document as unknown as WorkerTelemetry;
+}
+
+export async function fetchWorkerProfile(
+  signal?: AbortSignal,
+): Promise<WorkerProfileDocument> {
+  const document = requireJsonRecord(
+    await requestJson("/api/v1/system/worker-profile", {
+      method: "GET",
+      signal,
+    }),
+    "missioncore.worker-connection-profile/v1",
+  );
+  if (!isRecord(document.profile) || !isRecord(document.security)) {
+    throw new Error("Профиль Worker 006 неполон.");
+  }
+  return document as unknown as WorkerProfileDocument;
+}
+
+export async function testWorkerProfile(
+  request: WorkerProfileMutation,
+  signal?: AbortSignal,
+): Promise<WorkerProbe> {
+  const document = requireJsonRecord(
+    await requestJson("/api/v1/system/worker-profile/test", {
+      method: "POST",
+      body: JSON.stringify(request),
+      signal,
+    }),
+    "missioncore.worker-probe/v1",
+  );
+  return document as unknown as WorkerProbe;
+}
+
+export async function saveWorkerProfile(
+  request: WorkerProfileMutation,
+  signal?: AbortSignal,
+): Promise<WorkerProfileDocument & { verification: WorkerProbe }> {
+  const document = requireJsonRecord(
+    await requestJson("/api/v1/system/worker-profile", {
+      method: "PUT",
+      body: JSON.stringify(request),
+      signal,
+    }),
+    "missioncore.worker-connection-profile/v1",
+  );
+  if (!isRecord(document.profile) || !isRecord(document.verification)) {
+    throw new Error("Сохранённый профиль Worker 006 неполон.");
+  }
+  return document as unknown as WorkerProfileDocument & { verification: WorkerProbe };
+}
diff --git a/apps/control-station/src/productModel.ts b/apps/control-station/src/productModel.ts
index 890ad8e..e8efb51 100644
--- a/apps/control-station/src/productModel.ts
+++ b/apps/control-station/src/productModel.ts
@@ -18,6 +18,8 @@ export type WorkspaceKind =
   | "missions"
   | "catalog"
   | "contour-health"
+  | "compute-modules"
+  | "network-monitor"
   | "datasets"
   | "lab-archive";
 
@@ -589,16 +591,17 @@ export const workspaces: WorkspaceDefinition[] = [
     label: "Вычислительные модули",
     title: "Вычислительные модули",
     eyebrow: "СИСТЕМА / МОДУЛИ",
-    description: "Контракты будущих вычислительных модулей без их запуска на текущем интерфейсном этапе.",
+    description: "Живая аппаратная и процессинговая телеметрия выделенного Worker 006.",
     icon: "apps",
-    kind: "catalog",
+    kind: "compute-modules",
     groups: [
       {
         title: "Вычислительные роли",
         description: "Каждый модуль публикует стандартные выходы, интерфейс только отображает их.",
         capabilities: [
           active("Адаптер устройства", "Закрытый протокол текущего устройства уже изолирован."),
-          contract("Восприятие", "Рамки, маски, ключевые точки, траектории и семантические подписи."),
+          active("Восприятие", "Persistent perception worker и его текущие стадии обработки."),
+          active("Inference runtime", "Triton health, inference-счётчики и потребление ресурсов."),
           contract("Картирование", "Высота, занятость, сетки и проходимость."),
           contract("Локализация", "Поза, одометрия, TF и неопределённость."),
           contract("Исполнитель миссий", "Состояние графа миссии и действия верхнего уровня."),
@@ -637,13 +640,14 @@ export const workspaces: WorkspaceDefinition[] = [
     eyebrow: "СИСТЕМА / СЕТЬ",
     description: "Локальный контур, бортовая сеть, ячеистая сеть/LTE и серверные маршруты.",
     icon: "globe",
-    kind: "catalog",
+    kind: "network-monitor",
     groups: [
       {
         title: "Сетевые зоны",
         description: "Потоки управления и высокой пропускной способности разделяются архитектурно.",
         capabilities: [
-          active("Локальная сеть", "Устройство и рабочая станция в одной доступной сети."),
+          active("Переносимый профиль", "Адрес и порт Worker 006 проверяются до применения."),
+          active("Сетевая телеметрия", "Интерфейсы, счётчики, throughput и фактический маршрут."),
           active("Локальное управление", "Точка приёма учётных данных и браузерный контур управления ограничены локальным хостом."),
           contract("Бортовая сеть", "Сенсоры, вычислитель и маршрутизатор внутри аппарата."),
           contract("Дальний транспорт", "Ячеистая сеть или LTE передаёт предпросмотр, позу, события и изменения карты."),
diff --git a/apps/control-station/src/styles.css b/apps/control-station/src/styles.css
index cb5c4b1..fd665ac 100644
--- a/apps/control-station/src/styles.css
+++ b/apps/control-station/src/styles.css
@@ -11,3 +11,4 @@
 @import "./styles/responsive.css";
 @import "./styles/observation.css";
 @import "./styles/environment-settings.css";
+@import "./styles/system-telemetry.css";
diff --git a/apps/control-station/src/styles/system-telemetry.css b/apps/control-station/src/styles/system-telemetry.css
new file mode 100644
index 0000000..8966380
--- /dev/null
+++ b/apps/control-station/src/styles/system-telemetry.css
@@ -0,0 +1,544 @@
+.system-workspace {
+  display: grid;
+  min-width: 0;
+  gap: 1rem;
+  padding-bottom: 1rem;
+}
+
+.system-workspace__lead,
+.system-section-heading {
+  display: flex;
+  min-width: 0;
+  align-items: flex-start;
+  justify-content: space-between;
+  gap: 1.2rem;
+}
+
+.system-workspace__lead {
+  padding: 0.25rem 0 0.35rem;
+}
+
+.system-workspace__lead h2,
+.system-section-heading h3 {
+  margin: 0.4rem 0 0;
+  color: var(--nodedc-text-primary);
+  letter-spacing: -0.035em;
+}
+
+.system-workspace__lead h2 {
+  font-size: 1.42rem;
+}
+
+.system-section-heading h3 {
+  font-size: 1.02rem;
+}
+
+.system-workspace__lead p,
+.system-section-heading p {
+  max-width: 48rem;
+  margin: 0.42rem 0 0;
+  color: var(--nodedc-text-muted);
+  font-size: 0.68rem;
+  line-height: 1.48;
+}
+
+.system-workspace__actions {
+  display: flex;
+  flex: 0 0 auto;
+  align-items: center;
+  gap: 0.8rem;
+}
+
+.system-workspace__notice {
+  display: flex;
+  align-items: center;
+}
+
+.system-telemetry-grid,
+.network-overview-grid {
+  display: grid;
+  min-width: 0;
+  grid-template-columns: repeat(4, minmax(0, 1fr));
+  gap: 0.65rem;
+}
+
+.system-telemetry-series,
+.network-stat-card {
+  min-width: 0;
+  border: 1px solid var(--station-hairline);
+  border-radius: var(--nodedc-radius-card);
+  background: var(--station-panel-soft);
+}
+
+.system-telemetry-series {
+  overflow: hidden;
+  padding: 0.82rem 0.82rem 0.32rem;
+}
+
+.system-telemetry-series > div {
+  display: flex;
+  align-items: baseline;
+  justify-content: space-between;
+  gap: 0.75rem;
+}
+
+.system-telemetry-series span,
+.network-stat-card span {
+  color: var(--nodedc-text-muted);
+  font-size: 0.62rem;
+}
+
+.system-telemetry-series strong,
+.network-stat-card strong {
+  color: var(--nodedc-text-primary);
+  font-size: 1.14rem;
+  font-weight: 720;
+  letter-spacing: -0.035em;
+}
+
+.system-telemetry-series svg {
+  display: block;
+  width: 100%;
+  height: 2.45rem;
+  margin-top: 0.45rem;
+}
+
+.system-telemetry-series path {
+  fill: none;
+  stroke: var(--station-hairline);
+  stroke-width: 0.8;
+  vector-effect: non-scaling-stroke;
+}
+
+.system-telemetry-series polyline {
+  fill: none;
+  stroke: var(--nodedc-text-primary);
+  stroke-linecap: round;
+  stroke-linejoin: round;
+  stroke-width: 1.15;
+  vector-effect: non-scaling-stroke;
+}
+
+.worker-hardware,
+.worker-pipeline,
+.network-profile,
+.network-topology {
+  display: grid;
+  min-width: 0;
+  gap: 1rem;
+}
+
+.worker-hardware__facts {
+  display: grid;
+  min-width: 0;
+  grid-template-columns: repeat(2, minmax(0, 1fr));
+  gap: 0.7rem;
+}
+
+.worker-hardware__facts dl,
+.worker-runtime-card dl,
+.network-interface-card dl {
+  display: grid;
+  min-width: 0;
+  margin: 0;
+  gap: 0;
+}
+
+.worker-hardware__facts dl {
+  grid-template-columns: repeat(2, minmax(0, 1fr));
+  border: 1px solid var(--station-hairline);
+  border-radius: var(--nodedc-radius-control);
+}
+
+.worker-hardware__facts dl > div,
+.worker-runtime-card dl > div,
+.network-interface-card dl > div {
+  display: grid;
+  min-width: 0;
+  gap: 0.22rem;
+}
+
+.worker-hardware__facts dl > div {
+  padding: 0.75rem;
+}
+
+.worker-hardware__facts dt,
+.worker-runtime-card dt,
+.network-interface-card dt,
+.network-profile__security dt {
+  color: var(--nodedc-text-muted);
+  font-size: 0.58rem;
+}
+
+.worker-hardware__facts dd,
+.worker-runtime-card dd,
+.network-interface-card dd,
+.network-profile__security dd {
+  min-width: 0;
+  margin: 0;
+  overflow: hidden;
+  color: var(--nodedc-text-primary);
+  font-size: 0.67rem;
+  font-weight: 650;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
+.worker-disk-list {
+  display: grid;
+  grid-template-columns: repeat(auto-fit, minmax(12rem, 1fr));
+  gap: 0.55rem;
+}
+
+.worker-disk-list > div {
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  gap: 0.8rem;
+  padding: 0.62rem 0.72rem;
+  border-radius: var(--nodedc-radius-control);
+  background: var(--station-accent-soft);
+}
+
+.worker-disk-list span {
+  color: var(--nodedc-text-muted);
+  font-size: 0.61rem;
+}
+
+.worker-disk-list strong {
+  color: var(--nodedc-text-primary);
+  font-size: 0.64rem;
+}
+
+.system-runtime-section {
+  display: grid;
+  min-width: 0;
+  gap: 0.75rem;
+}
+
+.system-runtime-section--external {
+  padding-top: 0.25rem;
+  border-top: 1px solid var(--station-hairline);
+}
+
+.worker-runtime-grid {
+  display: grid;
+  min-width: 0;
+  grid-template-columns: repeat(2, minmax(0, 1fr));
+  gap: 0.7rem;
+}
+
+.worker-runtime-card {
+  display: grid;
+  min-width: 0;
+  gap: 0.9rem;
+}
+
+.worker-runtime-card[data-external="true"] {
+  opacity: 0.78;
+}
+
+.worker-runtime-card > header,
+.network-interface-card > header {
+  display: flex;
+  min-width: 0;
+  align-items: flex-start;
+  justify-content: space-between;
+  gap: 0.8rem;
+}
+
+.worker-runtime-card h3 {
+  margin: 0.38rem 0 0.18rem;
+  color: var(--nodedc-text-primary);
+  font-size: 0.9rem;
+  letter-spacing: -0.025em;
+}
+
+.worker-runtime-card code {
+  display: block;
+  max-width: 23rem;
+  overflow: hidden;
+  color: var(--nodedc-text-muted);
+  font-size: 0.56rem;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
+.worker-runtime-card dl {
+  grid-template-columns: repeat(4, minmax(0, 1fr));
+}
+
+.worker-runtime-card dl > div {
+  padding-right: 0.65rem;
+}
+
+.worker-pipeline__summary {
+  display: grid;
+  grid-template-columns: repeat(4, minmax(0, 1fr));
+  gap: 0.55rem;
+}
+
+.worker-pipeline__summary > div {
+  display: grid;
+  gap: 0.32rem;
+  padding: 0.7rem;
+  border-radius: var(--nodedc-radius-control);
+  background: var(--station-panel-soft);
+}
+
+.worker-pipeline__summary span {
+  color: var(--nodedc-text-muted);
+  font-size: 0.57rem;
+}
+
+.worker-pipeline__summary strong {
+  color: var(--nodedc-text-primary);
+  font-size: 0.78rem;
+}
+
+.worker-stage-list {
+  display: grid;
+  min-width: 0;
+  grid-template-columns: repeat(3, minmax(0, 1fr));
+  margin: 0;
+  padding: 0;
+  gap: 0.48rem;
+  list-style: none;
+}
+
+.worker-stage-list li {
+  display: grid;
+  min-width: 0;
+  grid-template-columns: auto minmax(0, 1fr) auto;
+  align-items: center;
+  gap: 0.55rem;
+  padding: 0.66rem;
+  border: 1px solid var(--station-hairline);
+  border-radius: var(--nodedc-radius-control);
+  background: var(--station-panel-soft);
+}
+
+.worker-stage-list li[data-state="active"] {
+  border-color: var(--station-hairline-strong);
+  background: var(--nodedc-focus-surface);
+}
+
+.worker-stage-list li > span,
+.worker-stage-list li > small {
+  color: var(--nodedc-text-muted);
+  font-size: 0.53rem;
+}
+
+.worker-stage-list li > strong {
+  min-width: 0;
+  overflow: hidden;
+  color: var(--nodedc-text-primary);
+  font-size: 0.62rem;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
+.network-stat-card {
+  display: grid;
+  align-content: center;
+  gap: 0.3rem;
+  padding: 0.82rem;
+}
+
+.network-stat-card small {
+  overflow: hidden;
+  color: var(--nodedc-text-muted);
+  font-size: 0.55rem;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
+.network-profile__form {
+  display: grid;
+  min-width: 0;
+  grid-template-columns: minmax(12rem, 1.5fr) minmax(8rem, 0.5fr) auto;
+  align-items: end;
+  gap: 0.7rem;
+}
+
+.network-profile__buttons {
+  display: flex;
+  align-items: center;
+  gap: 0.5rem;
+  padding-bottom: 0.02rem;
+}
+
+.network-profile__security {
+  display: grid;
+  grid-template-columns: repeat(4, minmax(0, 1fr));
+  margin: 0;
+  gap: 0.55rem;
+}
+
+.network-profile__security > div {
+  display: grid;
+  min-width: 0;
+  gap: 0.24rem;
+  padding: 0.64rem 0.7rem;
+  border-radius: var(--nodedc-radius-control);
+  background: var(--station-accent-soft);
+}
+
+.network-route {
+  display: grid;
+  min-width: 0;
+  grid-template-columns: repeat(3, minmax(8rem, 1fr) 2.2rem) minmax(8rem, 1fr);
+  align-items: center;
+}
+
+.network-route > div {
+  display: grid;
+  min-width: 0;
+  gap: 0.3rem;
+  padding: 0.78rem;
+  border: 1px solid var(--station-hairline);
+  border-radius: var(--nodedc-radius-control);
+  background: var(--station-panel-soft);
+}
+
+.network-route > div[data-state="offline"] {
+  opacity: 0.48;
+}
+
+.network-route > div span,
+.network-route > div small {
+  overflow: hidden;
+  color: var(--nodedc-text-muted);
+  font-size: 0.55rem;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
+.network-route > div strong {
+  overflow: hidden;
+  color: var(--nodedc-text-primary);
+  font-size: 0.68rem;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
+.network-route > i {
+  height: 1px;
+  background: var(--station-hairline-strong);
+}
+
+.network-route > i[data-state="offline"] {
+  background: var(--station-hairline);
+  opacity: 0.4;
+}
+
+.network-interface-section {
+  display: grid;
+  gap: 0.75rem;
+}
+
+.network-interface-list {
+  display: grid;
+  min-width: 0;
+  grid-template-columns: repeat(3, minmax(0, 1fr));
+  gap: 0.65rem;
+}
+
+.network-interface-card {
+  display: grid;
+  min-width: 0;
+  gap: 0.85rem;
+}
+
+.network-interface-card header strong,
+.network-interface-card header small {
+  display: block;
+  max-width: 17rem;
+  overflow: hidden;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
+.network-interface-card header strong {
+  color: var(--nodedc-text-primary);
+  font-size: 0.71rem;
+}
+
+.network-interface-card header small {
+  margin-top: 0.25rem;
+  color: var(--nodedc-text-muted);
+  font-size: 0.55rem;
+}
+
+.network-interface-card dl {
+  grid-template-columns: repeat(2, minmax(0, 1fr));
+  gap: 0.65rem;
+}
+
+@media (max-width: 1100px) {
+  .system-telemetry-grid,
+  .network-overview-grid,
+  .worker-pipeline__summary,
+  .network-profile__security {
+    grid-template-columns: repeat(2, minmax(0, 1fr));
+  }
+
+  .worker-stage-list,
+  .network-interface-list {
+    grid-template-columns: repeat(2, minmax(0, 1fr));
+  }
+
+  .network-profile__form {
+    grid-template-columns: minmax(12rem, 1fr) 8rem;
+  }
+
+  .network-profile__buttons {
+    grid-column: 1 / -1;
+  }
+
+  .network-route {
+    grid-template-columns: 1fr;
+    gap: 0.4rem;
+  }
+
+  .network-route > i {
+    width: 1px;
+    height: 1.2rem;
+    margin-left: 1.1rem;
+  }
+}
+
+@media (max-width: 760px) {
+  .system-workspace__lead,
+  .system-section-heading {
+    display: grid;
+  }
+
+  .system-workspace__actions {
+    justify-content: space-between;
+  }
+
+  .system-telemetry-grid,
+  .network-overview-grid,
+  .worker-hardware__facts,
+  .worker-runtime-grid,
+  .worker-stage-list,
+  .network-interface-list,
+  .network-profile__security {
+    grid-template-columns: 1fr;
+  }
+
+  .worker-runtime-card dl,
+  .worker-pipeline__summary {
+    grid-template-columns: repeat(2, minmax(0, 1fr));
+  }
+
+  .network-profile__form {
+    grid-template-columns: 1fr;
+  }
+
+  .network-profile__buttons {
+    grid-column: auto;
+    flex-wrap: wrap;
+  }
+}
diff --git a/apps/control-station/src/workspaces/Workspaces.tsx b/apps/control-station/src/workspaces/Workspaces.tsx
index 87969b3..b885dfc 100644
--- a/apps/control-station/src/workspaces/Workspaces.tsx
+++ b/apps/control-station/src/workspaces/Workspaces.tsx
@@ -46,6 +46,8 @@ import type { WorkspaceRendererProps } from "./contracts";
 import { DatasetGatewayWorkspace } from "./DatasetGatewayWorkspace";
 import { ContourHealthWorkspace } from "./ContourHealthWorkspace";
 import { LaboratoryArchiveWorkspace } from "./laboratory/LaboratoryArchiveWorkspace";
+import { ComputeModulesWorkspace } from "./system/ComputeModulesWorkspace";
+import { NetworkWorkspace } from "./system/NetworkWorkspace";
 
 function statusTone(status: CapabilityStatus): "success" | "accent" | "warning" | "neutral" {
   if (status === "active") return "success";
@@ -1177,6 +1179,10 @@ export function WorkspaceRenderer(props: WorkspaceRendererProps) {
           state={props.state}
         />
       );
+    case "compute-modules":
+      return <ComputeModulesWorkspace />;
+    case "network-monitor":
+      return <NetworkWorkspace />;
     case "datasets":
       return <DatasetGatewayWorkspace />;
     case "lab-archive":
diff --git a/apps/control-station/src/workspaces/system/ComputeModulesWorkspace.tsx b/apps/control-station/src/workspaces/system/ComputeModulesWorkspace.tsx
new file mode 100644
index 0000000..93c5f05
--- /dev/null
+++ b/apps/control-station/src/workspaces/system/ComputeModulesWorkspace.tsx
@@ -0,0 +1,226 @@
+import {
+  Button,
+  GlassSurface,
+  Icon,
+  StatusBadge,
+} from "@nodedc/ui-react";
+
+import { TelemetrySeries } from "../../components/system/TelemetrySeries";
+import { WorkerRuntimeCard } from "../../components/system/WorkerRuntimeCard";
+import {
+  formatBytes,
+  formatDuration,
+  formatOptionalPercent,
+} from "../../components/system/systemFormat";
+import { useWorkerTelemetry } from "../../core/system/useWorkerTelemetry";
+
+function pipelineStateLabel(state: string): string {
+  if (state === "busy") return "Выполняет задачу";
+  if (state === "ready") return "Готов к задаче";
+  return "Нет live-состояния";
+}
+
+export function ComputeModulesWorkspace() {
+  const { telemetry, loading, error, refresh } = useWorkerTelemetry();
+  const node = telemetry?.node ?? null;
+  const missionCoreRuntimes = telemetry?.runtimes.filter((runtime) => !runtime.external) ?? [];
+  const externalRuntimes = telemetry?.runtimes.filter((runtime) => runtime.external) ?? [];
+  const connected = Boolean(
+    telemetry?.connection.reachable && telemetry.connection.identity_matches && node,
+  );
+  const history = telemetry?.history ?? [];
+  const cpuPercent = node?.cpu.load_percent;
+  const memoryPercent = node?.memory.used_percent;
+  const gpuPercent = node?.gpu?.utilization_percent;
+  const gpuMemoryPercent = node?.gpu?.memory_used_percent;
+
+  return (
+    <div className="system-workspace compute-modules-workspace">
+      <section className="system-workspace__lead">
+        <div>
+          <span className="section-eyebrow">СИСТЕМА / ВЫЧИСЛИТЕЛЬНЫЕ МОДУЛИ</span>
+          <h2>Worker 006</h2>
+          <p>
+            Живой аппаратный и процессинговый срез выделенного узла. Нагрузка внешних
+            сервисов отделена от Mission Core и не входит в оценку наших runtime.
+          </p>
+        </div>
+        <div className="system-workspace__actions">
+          <StatusBadge tone={connected ? "success" : "danger"}>
+            {connected ? "Узел доступен" : "Нет связи с узлом"}
+          </StatusBadge>
+          <Button
+            size="compact"
+            variant="secondary"
+            disabled={loading}
+            onClick={refresh}
+            icon={<Icon name="refresh" size={14} />}
+          >
+            {loading ? "Читаем" : "Обновить"}
+          </Button>
+        </div>
+      </section>
+
+      {error ? (
+        <GlassSurface className="system-workspace__notice" padding="md" tone="soft">
+          <StatusBadge tone="warning">{error}</StatusBadge>
+        </GlassSurface>
+      ) : null}
+
+      <section className="system-telemetry-grid" aria-label="Аппаратная телеметрия">
+        <TelemetrySeries
+          label="CPU"
+          value={formatOptionalPercent(cpuPercent)}
+          ceiling={100}
+          values={history.map((item) => item.cpu_percent)}
+        />
+        <TelemetrySeries
+          label="RAM"
+          value={formatOptionalPercent(memoryPercent)}
+          ceiling={100}
+          values={history.map((item) => item.memory_percent)}
+        />
+        <TelemetrySeries
+          label="GPU"
+          value={formatOptionalPercent(gpuPercent)}
+          ceiling={100}
+          values={history.map((item) => item.gpu_percent)}
+        />
+        <TelemetrySeries
+          label="VRAM"
+          value={formatOptionalPercent(gpuMemoryPercent)}
+          ceiling={100}
+          values={history.map((item) => item.gpu_memory_percent)}
+        />
+      </section>
+
+      <GlassSurface className="worker-hardware" padding="lg">
+        <header className="system-section-heading">
+          <div>
+            <span className="section-eyebrow">HARDWARE</span>
+            <h3>{node?.node_id ?? telemetry?.profile.expected_node_id ?? "Worker 006"}</h3>
+            <p>{node?.os.caption ?? "Аппаратный профиль недоступен"}</p>
+          </div>
+          <StatusBadge tone={node?.node_id ? "success" : "danger"}>
+            {node?.node_id ? telemetry?.profile.display_name : "Нет данных"}
+          </StatusBadge>
+        </header>
+        <div className="worker-hardware__facts">
+          <dl>
+            <div><dt>Процессор</dt><dd>{node?.cpu.name ?? "—"}</dd></div>
+            <div><dt>Логические ядра</dt><dd>{node?.cpu.logical_processors ?? "—"}</dd></div>
+            <div><dt>Память занята</dt><dd>{formatBytes(node?.memory.used_bytes)}</dd></div>
+            <div><dt>Uptime</dt><dd>{formatDuration(node?.os.uptime_seconds)}</dd></div>
+          </dl>
+          <dl>
+            <div><dt>GPU</dt><dd>{node?.gpu?.name ?? "—"}</dd></div>
+            <div><dt>VRAM занята</dt><dd>{formatBytes(
+              typeof node?.gpu?.memory_used_mib === "number"
+                ? node.gpu.memory_used_mib * 1024 * 1024
+                : null,
+            )}</dd></div>
+            <div><dt>Температура</dt><dd>{
+              typeof node?.gpu?.temperature_celsius === "number"
+                ? `${node.gpu.temperature_celsius} °C`
+                : "—"
+            }</dd></div>
+            <div><dt>Мощность</dt><dd>{
+              typeof node?.gpu?.power_watts === "number"
+                ? `${node.gpu.power_watts.toFixed(1)} Вт`
+                : "—"
+            }</dd></div>
+          </dl>
+        </div>
+        <div className="worker-disk-list">
+          {(node?.disks ?? []).map((disk) => {
+            const size = typeof disk.size_bytes === "number" ? disk.size_bytes : null;
+            const free = typeof disk.free_bytes === "number" ? disk.free_bytes : null;
+            const used = size !== null && free !== null ? size - free : null;
+            return (
+              <div key={disk.name ?? "disk"}>
+                <span>Диск {disk.name ?? "—"}</span>
+                <strong>{formatBytes(used)} / {formatBytes(size)}</strong>
+              </div>
+            );
+          })}
+        </div>
+      </GlassSurface>
+
+      <section className="system-runtime-section">
+        <header className="system-section-heading">
+          <div>
+            <span className="section-eyebrow">PROCESSING RUNTIME</span>
+            <h3>Контейнеры Mission Core</h3>
+            <p>Два ограниченных runtime: inference server и прикладной perception pipeline.</p>
+          </div>
+          <StatusBadge tone={node?.triton.ready ? "success" : "danger"}>
+            {node?.triton.ready ? "Triton ready" : "Triton недоступен"}
+          </StatusBadge>
+        </header>
+        <div className="worker-runtime-grid">
+          {missionCoreRuntimes.map((runtime) => (
+            <WorkerRuntimeCard key={runtime.name} runtime={runtime} />
+          ))}
+        </div>
+      </section>
+
+      <GlassSurface className="worker-pipeline" padding="lg">
+        <header className="system-section-heading">
+          <div>
+            <span className="section-eyebrow">ТЕКУЩАЯ ЗАДАЧА</span>
+            <h3>{pipelineStateLabel(telemetry?.pipeline.service_state ?? "unavailable")}</h3>
+            <p>
+              {telemetry?.pipeline.active_request_id
+                ? `Run ${telemetry.pipeline.active_request_id}`
+                : "Очередь свободна; модели остаются загруженными в persistent worker."}
+            </p>
+          </div>
+          <StatusBadge tone={
+            telemetry?.pipeline.service_state === "busy" ? "warning"
+              : telemetry?.pipeline.service_state === "ready" ? "success"
+                : "danger"
+          }>
+            {telemetry?.pipeline.service_state ?? "unavailable"}
+          </StatusBadge>
+        </header>
+        <div className="worker-pipeline__summary">
+          <div><span>Завершено запусков</span><strong>{telemetry?.pipeline.completed_runs ?? "—"}</strong></div>
+          <div><span>Ошибок запусков</span><strong>{telemetry?.pipeline.failed_runs ?? "—"}</strong></div>
+          <div><span>Inference success</span><strong>{node?.triton.requests_succeeded ?? "—"}</strong></div>
+          <div><span>Inference failed</span><strong>{node?.triton.requests_failed ?? "—"}</strong></div>
+        </div>
+        <ol className="worker-stage-list">
+          {(telemetry?.pipeline.stages ?? []).map((stage, index) => (
+            <li key={stage.id} data-state={stage.state}>
+              <span>{String(index + 1).padStart(2, "0")}</span>
+              <strong>{stage.label}</strong>
+              <small>{
+                stage.state === "active" ? "выполняется"
+                  : stage.state === "waiting" ? "в очереди"
+                    : stage.state === "ready" ? "готов"
+                      : "нет данных"
+              }</small>
+            </li>
+          ))}
+        </ol>
+      </GlassSurface>
+
+      <section className="system-runtime-section system-runtime-section--external">
+        <header className="system-section-heading">
+          <div>
+            <span className="section-eyebrow">CO-TENANTS / НЕ MISSION CORE</span>
+            <h3>Внешняя нагрузка узла</h3>
+            <p>
+              Эти процессы влияют на общую плату и VRAM, но не считаются расходом Mission Core.
+            </p>
+          </div>
+        </header>
+        <div className="worker-runtime-grid">
+          {externalRuntimes.map((runtime) => (
+            <WorkerRuntimeCard key={runtime.name} runtime={runtime} />
+          ))}
+        </div>
+      </section>
+    </div>
+  );
+}
diff --git a/apps/control-station/src/workspaces/system/NetworkWorkspace.tsx b/apps/control-station/src/workspaces/system/NetworkWorkspace.tsx
new file mode 100644
index 0000000..f088238
--- /dev/null
+++ b/apps/control-station/src/workspaces/system/NetworkWorkspace.tsx
@@ -0,0 +1,311 @@
+import { useEffect, useMemo, useState } from "react";
+import {
+  Button,
+  GlassSurface,
+  Icon,
+  StatusBadge,
+  TextField,
+} from "@nodedc/ui-react";
+
+import { TelemetrySeries } from "../../components/system/TelemetrySeries";
+import {
+  formatBitRate,
+  formatBytes,
+  formatLatency,
+  formatRate,
+} from "../../components/system/systemFormat";
+import {
+  fetchWorkerProfile,
+  saveWorkerProfile,
+  testWorkerProfile,
+  type WorkerConnectionProfile,
+  type WorkerProbe,
+} from "../../core/system/workerTelemetry";
+import { useWorkerTelemetry } from "../../core/system/useWorkerTelemetry";
+
+type ProfileAction = "idle" | "testing" | "saving";
+
+function probeLabel(probe: WorkerProbe | null): string {
+  if (!probe) return "Изменения не проверены";
+  if (!probe.reachable) return "Адрес не отвечает";
+  if (!probe.identity_matches) return "Ответил другой узел";
+  return "Worker 006 подтверждён";
+}
+
+export function NetworkWorkspace() {
+  const { telemetry, loading, error, refresh } = useWorkerTelemetry();
+  const [profile, setProfile] = useState<WorkerConnectionProfile | null>(null);
+  const [address, setAddress] = useState("");
+  const [port, setPort] = useState("22");
+  const [profileAction, setProfileAction] = useState<ProfileAction>("idle");
+  const [probe, setProbe] = useState<WorkerProbe | null>(null);
+  const [profileError, setProfileError] = useState<string | null>(null);
+  const aggregate = telemetry?.network.aggregate ?? null;
+  const connected = Boolean(
+    telemetry?.connection.reachable && telemetry.connection.identity_matches,
+  );
+
+  useEffect(() => {
+    const controller = new AbortController();
+    void fetchWorkerProfile(controller.signal)
+      .then((document) => {
+        if (controller.signal.aborted) return;
+        setProfile(document.profile);
+        setAddress(document.profile.address);
+        setPort(String(document.profile.port));
+        setProfileError(null);
+      })
+      .catch((reason: unknown) => {
+        if (controller.signal.aborted) return;
+        setProfileError(reason instanceof Error ? reason.message : "Профиль не прочитан.");
+      });
+    return () => controller.abort();
+  }, []);
+
+  const mutation = useMemo(() => {
+    const numericPort = Number(port);
+    if (!profile || !Number.isInteger(numericPort) || numericPort < 1 || numericPort > 65535) {
+      return null;
+    }
+    return {
+      revision: profile.revision,
+      address: address.trim(),
+      port: numericPort,
+    };
+  }, [address, port, profile]);
+
+  const runTest = () => {
+    if (!mutation) return;
+    const controller = new AbortController();
+    setProfileAction("testing");
+    setProfileError(null);
+    setProbe(null);
+    void testWorkerProfile(mutation, controller.signal)
+      .then(setProbe)
+      .catch((reason: unknown) => {
+        setProfileError(reason instanceof Error ? reason.message : "Проверка не выполнена.");
+      })
+      .finally(() => setProfileAction("idle"));
+  };
+
+  const saveProfile = () => {
+    if (!mutation) return;
+    const controller = new AbortController();
+    setProfileAction("saving");
+    setProfileError(null);
+    void saveWorkerProfile(mutation, controller.signal)
+      .then((document) => {
+        setProfile(document.profile);
+        setAddress(document.profile.address);
+        setPort(String(document.profile.port));
+        setProbe(document.verification);
+        refresh();
+      })
+      .catch((reason: unknown) => {
+        setProfileError(reason instanceof Error ? reason.message : "Профиль не сохранён.");
+      })
+      .finally(() => setProfileAction("idle"));
+  };
+
+  return (
+    <div className="system-workspace network-workspace">
+      <section className="system-workspace__lead">
+        <div>
+          <span className="section-eyebrow">СИСТЕМА / СЕТЬ</span>
+          <h2>Локальный вычислительный контур</h2>
+          <p>
+            Переносимый профиль связи между Mission Core и Worker 006. Адрес можно заменить
+            при переходе в другую локальную сеть; идентичность узла проверяется до сохранения.
+          </p>
+        </div>
+        <div className="system-workspace__actions">
+          <StatusBadge tone={connected ? "success" : "danger"}>
+            {connected ? "Маршрут доступен" : "Маршрут недоступен"}
+          </StatusBadge>
+          <Button
+            size="compact"
+            variant="secondary"
+            disabled={loading}
+            onClick={refresh}
+            icon={<Icon name="refresh" size={14} />}
+          >
+            {loading ? "Читаем" : "Обновить"}
+          </Button>
+        </div>
+      </section>
+
+      {(error || profileError) ? (
+        <GlassSurface className="system-workspace__notice" padding="md" tone="soft">
+          <StatusBadge tone="warning">{profileError ?? error}</StatusBadge>
+        </GlassSurface>
+      ) : null}
+
+      <section className="network-overview-grid" aria-label="Сетевая телеметрия">
+        <TelemetrySeries
+          label="Приём"
+          value={formatRate(aggregate?.receive_bytes_per_second)}
+          values={(telemetry?.history ?? []).map(
+            (item) => item.network_receive_bytes_per_second,
+          )}
+        />
+        <TelemetrySeries
+          label="Передача"
+          value={formatRate(aggregate?.send_bytes_per_second)}
+          values={(telemetry?.history ?? []).map(
+            (item) => item.network_send_bytes_per_second,
+          )}
+        />
+        <div className="network-stat-card">
+          <span>Сбор телеметрии</span>
+          <strong>{formatLatency(telemetry?.connection.latency_ms)}</strong>
+          <small>полный SSH probe Worker 006</small>
+        </div>
+        <div className="network-stat-card">
+          <span>Активные интерфейсы</span>
+          <strong>{telemetry?.network.interfaces.length ?? "—"}</strong>
+          <small>по Windows network counters</small>
+        </div>
+      </section>
+
+      <GlassSurface className="network-profile" padding="lg">
+        <header className="system-section-heading">
+          <div>
+            <span className="section-eyebrow">ПРОФИЛЬ ПОДКЛЮЧЕНИЯ</span>
+            <h3>Worker 006</h3>
+            <p>
+              Реальный Node ID: <code>{profile?.expected_node_id ?? "DESKTOP-OPJ8J04"}</code>.
+              Пустой адрес использует закреплённый локальный SSH-профиль.
+            </p>
+          </div>
+          <StatusBadge tone={
+            probe?.reachable && probe.identity_matches ? "success"
+              : probe ? "danger"
+                : connected ? "success" : "neutral"
+          }>
+            {probe ? probeLabel(probe) : connected ? "Текущий профиль работает" : "Не проверено"}
+          </StatusBadge>
+        </header>
+        <div className="network-profile__form">
+          <TextField
+            label="Адрес Worker 006"
+            hint="IP или hostname"
+            value={address}
+            placeholder="из SSH-профиля mission-gpu"
+            autoComplete="off"
+            spellCheck={false}
+            onChange={(event) => {
+              setAddress(event.target.value);
+              setProbe(null);
+            }}
+          />
+          <TextField
+            label="SSH-порт"
+            hint="1–65535"
+            type="number"
+            min={1}
+            max={65535}
+            value={port}
+            onChange={(event) => {
+              setPort(event.target.value);
+              setProbe(null);
+            }}
+          />
+          <div className="network-profile__buttons">
+            <Button
+              variant="secondary"
+              disabled={!mutation || profileAction !== "idle"}
+              onClick={runTest}
+            >
+              {profileAction === "testing" ? "Проверяем" : "Проверить"}
+            </Button>
+            <Button
+              variant="primary"
+              disabled={!mutation || profileAction !== "idle"}
+              onClick={saveProfile}
+            >
+              {profileAction === "saving" ? "Сохраняем" : "Проверить и применить"}
+            </Button>
+          </div>
+        </div>
+        <dl className="network-profile__security">
+          <div><dt>Транспорт</dt><dd>SSH · key-only</dd></div>
+          <div><dt>Host key</dt><dd>strict · pinned</dd></div>
+          <div><dt>Профиль</dt><dd>{profile?.ssh_host_alias ?? "mission-gpu"}</dd></div>
+          <div><dt>Учётные данные</dt><dd>не доступны интерфейсу</dd></div>
+        </dl>
+      </GlassSurface>
+
+      <GlassSurface className="network-topology" padding="lg">
+        <header className="system-section-heading">
+          <div>
+            <span className="section-eyebrow">ФАКТИЧЕСКИЙ МАРШРУТ</span>
+            <h3>Поток управления и обработки</h3>
+            <p>Схема собрана из текущего профиля и live health runtime, без demo-узлов.</p>
+          </div>
+        </header>
+        <div className="network-route">
+          <div data-state="online">
+            <span>Операторский UI</span>
+            <strong>Browser</strong>
+            <small>127.0.0.1:8000</small>
+          </div>
+          <i aria-hidden="true" data-state="online" />
+          <div data-state="online">
+            <span>Control Plane</span>
+            <strong>Mission Core</strong>
+            <small>локальный API</small>
+          </div>
+          <i aria-hidden="true" data-state={connected ? "online" : "offline"} />
+          <div data-state={connected ? "online" : "offline"}>
+            <span>Вычислительный узел</span>
+            <strong>Worker 006</strong>
+            <small>{profile?.address || profile?.ssh_host_alias || "mission-gpu"}:{profile?.port ?? 22}</small>
+          </div>
+          <i aria-hidden="true" data-state={
+            telemetry?.runtimes.some((runtime) => !runtime.external && runtime.state === "running")
+              ? "online" : "offline"
+          } />
+          <div data-state={
+            telemetry?.runtimes.some((runtime) => !runtime.external && runtime.state === "running")
+              ? "online" : "offline"
+          }>
+            <span>Runtime</span>
+            <strong>Triton + Pipeline</strong>
+            <small>внутренний Docker-контур</small>
+          </div>
+        </div>
+      </GlassSurface>
+
+      <section className="network-interface-section">
+        <header className="system-section-heading">
+          <div>
+            <span className="section-eyebrow">ИНТЕРФЕЙСЫ WORKER 006</span>
+            <h3>Адаптеры и счётчики</h3>
+            <p>Только активные интерфейсы, которые вернул сам узел.</p>
+          </div>
+        </header>
+        <div className="network-interface-list">
+          {(telemetry?.network.interfaces ?? []).map((adapter) => (
+            <GlassSurface className="network-interface-card" padding="md" key={adapter.name}>
+              <header>
+                <div>
+                  <strong>{adapter.name}</strong>
+                  <small>{adapter.description ?? "Описание недоступно"}</small>
+                </div>
+                <StatusBadge tone={adapter.status === "Up" ? "success" : "warning"}>
+                  {adapter.status ?? "unknown"}
+                </StatusBadge>
+              </header>
+              <dl>
+                <div><dt>Адреса</dt><dd>{adapter.addresses.join(", ") || "—"}</dd></div>
+                <div><dt>Link</dt><dd>{formatBitRate(adapter.link_speed_bps)}</dd></div>
+                <div><dt>Получено</dt><dd>{formatBytes(adapter.received_bytes)}</dd></div>
+                <div><dt>Отправлено</dt><dd>{formatBytes(adapter.sent_bytes)}</dd></div>
+              </dl>
+            </GlassSurface>
+          ))}
+        </div>
+      </section>
+    </div>
+  );
+}
diff --git a/apps/control-station/test/systemTelemetry.test.mjs b/apps/control-station/test/systemTelemetry.test.mjs
new file mode 100644
index 0000000..b8ee1c3
--- /dev/null
+++ b/apps/control-station/test/systemTelemetry.test.mjs
@@ -0,0 +1,40 @@
+import assert from "node:assert/strict";
+import { readFile } from "node:fs/promises";
+import { test } from "node:test";
+
+const sourceRoot = new URL("../src/", import.meta.url);
+
+async function read(relativePath) {
+  return readFile(new URL(relativePath, sourceRoot), "utf8");
+}
+
+test("Worker 006 telemetry remains a bounded system feature slice", async () => {
+  const [
+    productModel,
+    workspaceHub,
+    core,
+    computeWorkspace,
+    networkWorkspace,
+    styles,
+  ] = await Promise.all([
+    read("productModel.ts"),
+    read("workspaces/Workspaces.tsx"),
+    read("core/system/workerTelemetry.ts"),
+    read("workspaces/system/ComputeModulesWorkspace.tsx"),
+    read("workspaces/system/NetworkWorkspace.tsx"),
+    read("styles.css"),
+  ]);
+
+  assert.match(productModel, /kind: "compute-modules"/);
+  assert.match(productModel, /kind: "network-monitor"/);
+  assert.match(workspaceHub, /<ComputeModulesWorkspace \/>/);
+  assert.match(workspaceHub, /<NetworkWorkspace \/>/);
+  assert.doesNotMatch(workspaceHub, /worker-telemetry|worker-profile|DESKTOP-OPJ8J04/);
+  assert.match(core, /\/api\/v1\/system\/worker-telemetry/);
+  assert.match(core, /\/api\/v1\/system\/worker-profile/);
+  assert.doesNotMatch(core, /@nodedc\/ui-react/);
+  assert.match(computeWorkspace, /useWorkerTelemetry/);
+  assert.match(networkWorkspace, /127\.0\.0\.1:8000/);
+  assert.doesNotMatch(networkWorkspace, /8765/);
+  assert.match(styles, /system-telemetry\.css/);
+});
diff --git a/experiments/perception/worker/run_e15_shadow_inference.py b/experiments/perception/worker/run_e15_shadow_inference.py
index 7193dc5..72c5c17 100644
--- a/experiments/perception/worker/run_e15_shadow_inference.py
+++ b/experiments/perception/worker/run_e15_shadow_inference.py
@@ -1138,7 +1138,11 @@ def _load_models(args: argparse.Namespace, common: dict[str, Any]) -> _LoadedMod
     )
 
 
-def run(args: argparse.Namespace, loaded: _LoadedModels | None = None) -> int:
+def run(
+    args: argparse.Namespace,
+    loaded: _LoadedModels | None = None,
+    runtime_state: dict[str, Any] | None = None,
+) -> int:
     import av
     import torch
     import transformers
@@ -1149,6 +1153,18 @@ def run(args: argparse.Namespace, loaded: _LoadedModels | None = None) -> int:
     token = sys.stdin.readline().strip() if args.token_stdin else args.token
     if not token or len(token) < 40:
         raise RuntimeError("LAB E15 shadow token is missing")
+
+    def report_stage(stage: str, frame_index: int | None = None) -> None:
+        if runtime_state is None:
+            return
+        runtime_state["current_stage"] = stage
+        runtime_state["stage_observed_at_utc"] = (
+            datetime.now(UTC).isoformat().replace("+00:00", "Z")
+        )
+        if frame_index is not None:
+            runtime_state["active_frame_index"] = frame_index
+
+    report_stage("preprocessing")
     common = _common(args) if loaded is None else loaded.common
     live = common["live"]
     e14 = common["e14"]
@@ -1215,6 +1231,7 @@ def run(args: argparse.Namespace, loaded: _LoadedModels | None = None) -> int:
     def on_decoded(frame: DecodedCameraFrame) -> None:
         nonlocal decoded_frame_count
         try:
+            report_stage("camera-decode", frame.frame_index)
             if not first_camera_epoch_ns:
                 first_camera_epoch_ns.append(frame.metadata.captured_at_epoch_ns)
             last_camera_epoch_ns[:] = [frame.metadata.captured_at_epoch_ns]
@@ -1383,6 +1400,7 @@ def run(args: argparse.Namespace, loaded: _LoadedModels | None = None) -> int:
         )
         semantic_thread.start()
         decoder.start()
+        report_stage("source-ingress")
         receiver_thread = threading.Thread(
             target=_receiver,
             kwargs={
@@ -1429,10 +1447,13 @@ def run(args: argparse.Namespace, loaded: _LoadedModels | None = None) -> int:
             latency["decode_age_ms"].append(float(envelope.decode_ms))
             latency["queue_wait_ms"].append(max(0.0, (started - envelope.decoded_monotonic) * 1000))
             try:
+                report_stage("preprocessing", envelope.frame_index)
                 detector_started = time.perf_counter()
                 tensor = _preprocess(envelope.image, valid_mask, detector)
+                report_stage("detector", envelope.frame_index)
                 output_tensor, _request_ms = _infer(args.triton_url, detector["model"], tensor)
                 detections, _rejected = _detections(output_tensor, detector, valid_mask)
+                report_stage("tracking", envelope.frame_index)
                 tracks = tracker.update(detections, envelope.frame_index)
                 latency["detector_ms"].append((time.perf_counter() - detector_started) * 1000)
 
@@ -1463,6 +1484,7 @@ def run(args: argparse.Namespace, loaded: _LoadedModels | None = None) -> int:
                     ).reshape((-1, 3))
                     position = binding.pose.position_xyz
                     quaternion = binding.pose.orientation_xyzw
+                    report_stage("sensor-fusion", envelope.frame_index)
                     projection_started = time.perf_counter()
                     pixels, depths, source_indices, points_lidar = project_points(
                         points_map,
@@ -1533,6 +1555,7 @@ def run(args: argparse.Namespace, loaded: _LoadedModels | None = None) -> int:
                 if temporal_stabilizer is None:
                     fusion_objects = raw_fusion_objects
                 else:
+                    report_stage("temporal-state", envelope.frame_index)
                     temporal_started = time.perf_counter()
                     fusion_objects = temporal_stabilizer.update(
                         frame_index=envelope.frame_index,
@@ -1619,6 +1642,7 @@ def run(args: argparse.Namespace, loaded: _LoadedModels | None = None) -> int:
                     objects=fusion_objects,
                     delivery=world["delivery"],
                 )
+                report_stage("result-publication", envelope.frame_index)
                 try:
                     result_queue.put_nowait(live_result)
                 except queue.Full:
@@ -1627,6 +1651,7 @@ def run(args: argparse.Namespace, loaded: _LoadedModels | None = None) -> int:
                         result_queue.task_done()
                         transport.results_dropped += 1
                     result_queue.put_nowait(live_result)
+                report_stage("source-ingress", envelope.frame_index)
             except Exception:
                 detector_failures += 1
                 raise
@@ -2063,7 +2088,15 @@ def serve(args: argparse.Namespace) -> int:
     loaded = _load_models(args, common)
     model_load_seconds = time.perf_counter() - load_started
     run_lock = threading.Lock()
-    state = {"busy": False, "completed_runs": 0, "failed_runs": 0}
+    state: dict[str, Any] = {
+        "busy": False,
+        "completed_runs": 0,
+        "failed_runs": 0,
+        "active_request_id": None,
+        "active_frame_index": None,
+        "current_stage": None,
+        "stage_observed_at_utc": None,
+    }
 
     class Handler(BaseHTTPRequestHandler):
         server_version = "MissionCorePersistentPerception/1"
@@ -2100,6 +2133,10 @@ def serve(args: argparse.Namespace) -> int:
                     "model_load_seconds": model_load_seconds,
                     "completed_runs": state["completed_runs"],
                     "failed_runs": state["failed_runs"],
+                    "active_request_id": state["active_request_id"],
+                    "active_frame_index": state["active_frame_index"],
+                    "current_stage": state["current_stage"],
+                    "stage_observed_at_utc": state["stage_observed_at_utc"],
                     "authority": common["live"]["authority"],
                     "gpu": torch.cuda.get_device_name(),
                 },
@@ -2129,7 +2166,8 @@ def serve(args: argparse.Namespace) -> int:
                 request_id = str(document.get("request_id", "invalid"))
                 run_args = _persistent_run_arguments(args, document)
                 document["token"] = None
-                exit_code = run(run_args, loaded)
+                state["active_request_id"] = request_id
+                exit_code = run(run_args, loaded, state)
                 state["completed_runs"] += 1
                 self._send(
                     200,
@@ -2164,6 +2202,10 @@ def serve(args: argparse.Namespace) -> int:
                 )
             finally:
                 state["busy"] = False
+                state["active_request_id"] = None
+                state["active_frame_index"] = None
+                state["current_stage"] = None
+                state["stage_observed_at_utc"] = None
                 run_lock.release()
 
     server = ThreadingHTTPServer((args.listen_host, args.listen_port), Handler)
diff --git a/src/k1link/web/app.py b/src/k1link/web/app.py
index 24acba2..6f36e2a 100644
--- a/src/k1link/web/app.py
+++ b/src/k1link/web/app.py
@@ -52,6 +52,7 @@ from k1link.web.plugin_runtime import (
 )
 from k1link.web.polygon_api import build_polygon_router, configured_polygon_runs_root
 from k1link.web.session_api import build_session_router
+from k1link.web.system_telemetry_api import build_system_telemetry_router
 
 REPOSITORY_ROOT = Path(__file__).resolve().parents[3]
 INVALID_REQUEST_DETAIL = "Некорректные параметры запроса."
@@ -618,6 +619,11 @@ app.include_router(
         ),
     )
 )
+app.include_router(
+    build_system_telemetry_router(
+        root_provider=lambda: REPOSITORY_ROOT / ".runtime" / "system",
+    )
+)
 
 
 frontend_dist = REPOSITORY_ROOT / "apps" / "control-station" / "dist"
diff --git a/src/k1link/web/system_telemetry_api.py b/src/k1link/web/system_telemetry_api.py
new file mode 100644
index 0000000..5fd4af4
--- /dev/null
+++ b/src/k1link/web/system_telemetry_api.py
@@ -0,0 +1,834 @@
+from __future__ import annotations
+
+import base64
+import ipaddress
+import json
+import math
+import os
+import re
+import subprocess
+import tempfile
+import threading
+import time
+from collections import deque
+from collections.abc import Callable
+from datetime import UTC, datetime
+from pathlib import Path
+from typing import Any, Final
+
+from fastapi import APIRouter, HTTPException, Query
+from pydantic import BaseModel, ConfigDict, Field, field_validator
+
+PROFILE_SCHEMA: Final = "missioncore.worker-connection-profile/v1"
+TELEMETRY_SCHEMA: Final = "missioncore.worker-telemetry/v1"
+PROBE_SCHEMA: Final = "missioncore.worker-probe/v1"
+DEFAULT_PROFILE_ID: Final = "worker-006"
+DEFAULT_DISPLAY_NAME: Final = "Worker 006"
+EXPECTED_NODE_ID: Final = "DESKTOP-OPJ8J04"
+SSH_HOST_ALIAS: Final = "mission-gpu"
+PROFILE_FILE_NAME: Final = "worker-006.json"
+CONTAINER_NAMES: Final = (
+    "mission-core-triton",
+    "mission-core-perception-worker",
+    "sentinel-frigate",
+    "sentinel-ollama",
+)
+SAFE_HOSTNAME = re.compile(
+    r"^(?=.{1,253}\.?$)(?:[A-Za-z0-9](?:[A-Za-z0-9-]{0,61}[A-Za-z0-9])?\.)*"
+    r"[A-Za-z0-9](?:[A-Za-z0-9-]{0,61}[A-Za-z0-9])?\.?$"
+)
+PROMETHEUS_SAMPLE = re.compile(
+    r"^(?P<name>nv_inference_(?:request_success|request_failure|count))"
+    r"(?:\{[^}]*\})?\s+(?P<value>[0-9.eE+-]+)$"
+)
+
+RootProvider = Callable[[], Path]
+ProbeRunner = Callable[["WorkerConnectionProfile"], dict[str, Any]]
+
+
+class WorkerConnectionProfile(BaseModel):
+    model_config = ConfigDict(extra="forbid", frozen=True)
+
+    schema_version: str = PROFILE_SCHEMA
+    profile_id: str = DEFAULT_PROFILE_ID
+    display_name: str = DEFAULT_DISPLAY_NAME
+    expected_node_id: str = EXPECTED_NODE_ID
+    ssh_host_alias: str = SSH_HOST_ALIAS
+    address: str = ""
+    port: int = Field(default=22, ge=1, le=65535)
+    revision: int = Field(default=0, ge=0)
+    updated_at_utc: str | None = None
+
+    @field_validator("address")
+    @classmethod
+    def validate_address(cls, value: str) -> str:
+        normalized = value.strip()
+        if not normalized:
+            return ""
+        if any(character.isspace() or ord(character) < 32 for character in normalized):
+            raise ValueError("worker address contains whitespace or control characters")
+        try:
+            ipaddress.ip_address(normalized)
+        except ValueError:
+            if SAFE_HOSTNAME.fullmatch(normalized) is None:
+                raise ValueError("worker address is not a valid IP address or hostname") from None
+        return normalized.rstrip(".")
+
+
+class WorkerConnectionProfilePut(BaseModel):
+    model_config = ConfigDict(extra="forbid", frozen=True)
+
+    revision: int = Field(ge=0)
+    address: str = Field(default="", max_length=253)
+    port: int = Field(default=22, ge=1, le=65535)
+
+    @field_validator("address")
+    @classmethod
+    def validate_address(cls, value: str) -> str:
+        return WorkerConnectionProfile.validate_address(value)
+
+
+def _utc_now() -> str:
+    return datetime.now(UTC).isoformat().replace("+00:00", "Z")
+
+
+class WorkerProfileStore:
+    def __init__(self, root: Path) -> None:
+        self.root = root
+        self.path = root / PROFILE_FILE_NAME
+        self._lock = threading.Lock()
+
+    def read(self) -> WorkerConnectionProfile:
+        with self._lock:
+            if not self.path.is_file():
+                return WorkerConnectionProfile()
+            try:
+                document = json.loads(self.path.read_text(encoding="utf-8"))
+                return WorkerConnectionProfile.model_validate(document)
+            except (OSError, ValueError, TypeError, json.JSONDecodeError):
+                return WorkerConnectionProfile()
+
+    def save(self, request: WorkerConnectionProfilePut) -> WorkerConnectionProfile:
+        with self._lock:
+            current = self._read_unlocked()
+            if current.revision != request.revision:
+                raise RuntimeError("worker profile revision changed")
+            profile = current.model_copy(
+                update={
+                    "address": request.address,
+                    "port": request.port,
+                    "revision": current.revision + 1,
+                    "updated_at_utc": _utc_now(),
+                }
+            )
+            self.root.mkdir(mode=0o700, parents=True, exist_ok=True)
+            descriptor, temporary_name = tempfile.mkstemp(
+                prefix=".worker-006.",
+                suffix=".tmp",
+                dir=self.root,
+            )
+            temporary = Path(temporary_name)
+            try:
+                with os.fdopen(descriptor, "w", encoding="utf-8", newline="\n") as stream:
+                    json.dump(
+                        profile.model_dump(mode="json"),
+                        stream,
+                        ensure_ascii=False,
+                        sort_keys=True,
+                        separators=(",", ":"),
+                    )
+                    stream.write("\n")
+                    stream.flush()
+                    os.fsync(stream.fileno())
+                os.chmod(temporary, 0o600)
+                os.replace(temporary, self.path)
+            finally:
+                temporary.unlink(missing_ok=True)
+            return profile
+
+    def _read_unlocked(self) -> WorkerConnectionProfile:
+        if not self.path.is_file():
+            return WorkerConnectionProfile()
+        try:
+            return WorkerConnectionProfile.model_validate_json(
+                self.path.read_text(encoding="utf-8")
+            )
+        except (OSError, ValueError):
+            return WorkerConnectionProfile()
+
+
+WORKER_PROBE_POWERSHELL: Final = r"""
+$ErrorActionPreference = "Stop"
+$containers = @(
+  "mission-core-triton",
+  "mission-core-perception-worker",
+  "sentinel-frigate",
+  "sentinel-ollama"
+)
+$os = Get-CimInstance Win32_OperatingSystem
+$cpu = Get-CimInstance Win32_Processor
+$computer = Get-CimInstance Win32_ComputerSystem
+$dockerStats = @{}
+try {
+  $statsLines = @(docker stats --no-stream --format "{{json .}}" $containers)
+  foreach ($line in $statsLines) {
+    if ($line) {
+      $row = $line | ConvertFrom-Json
+      $dockerStats[$row.Name] = $row
+    }
+  }
+} catch {}
+$containerStates = @{}
+foreach ($name in $containers) {
+  try {
+    $state = (docker inspect --format "{{json .State}}" $name) | ConvertFrom-Json
+    $image = docker inspect --format "{{.Config.Image}}" $name
+    $containerStates[$name] = [ordered]@{state=$state; image=$image}
+  } catch {
+    $containerStates[$name] = $null
+  }
+}
+$gpu = $null
+try {
+  $gpuQuery = "name,utilization.gpu,memory.used,memory.total,power.draw,temperature.gpu"
+  $gpuLine = nvidia-smi "--query-gpu=$gpuQuery" --format=csv,noheader,nounits |
+    Select-Object -First 1
+  $gpuParts = @($gpuLine -split ",\s*")
+  if ($gpuParts.Count -ge 6) {
+    $gpu = [ordered]@{
+      name=$gpuParts[0]
+      utilization_percent=[double]$gpuParts[1]
+      memory_used_mib=[double]$gpuParts[2]
+      memory_total_mib=[double]$gpuParts[3]
+      power_watts=[double]$gpuParts[4]
+      temperature_celsius=[double]$gpuParts[5]
+    }
+  }
+} catch {}
+$network = @()
+foreach ($adapter in @(Get-NetAdapter | Where-Object Status -eq "Up")) {
+  try {
+    $statistics = Get-NetAdapterStatistics -Name $adapter.Name
+    $addresses = @(
+      Get-NetIPAddress -InterfaceIndex $adapter.ifIndex -AddressFamily IPv4 `
+        -ErrorAction SilentlyContinue |
+        Where-Object { $_.IPAddress -notlike "169.254.*" } |
+        Select-Object -ExpandProperty IPAddress
+    )
+    $network += [ordered]@{
+      name=$adapter.Name
+      description=$adapter.InterfaceDescription
+      status=$adapter.Status
+      link_speed_bps=[double]$adapter.Speed
+      mac_address=$adapter.MacAddress
+      addresses=$addresses
+      received_bytes=[double]$statistics.ReceivedBytes
+      sent_bytes=[double]$statistics.SentBytes
+    }
+  } catch {}
+}
+$tritonReady = $false
+try {
+  $tritonResponse = Invoke-WebRequest -UseBasicParsing `
+    -Uri "http://127.0.0.1:8000/v2/health/ready" -TimeoutSec 2
+  $tritonReady = $tritonResponse.StatusCode -eq 200
+} catch {}
+$tritonMetrics = @()
+try {
+  $metricsResponse = Invoke-WebRequest -UseBasicParsing `
+    -Uri "http://127.0.0.1:8002/metrics" -TimeoutSec 2
+  $tritonMetrics = @(
+    $metricsResponse.Content -split "`n" |
+      Where-Object { $_ -match "^nv_inference_(request_success|request_failure|count)" }
+  )
+} catch {}
+$perceptionHealth = $null
+try {
+  $healthJson = docker exec mission-core-perception-worker python3 -c "import urllib.request;print(urllib.request.urlopen('http://127.0.0.1:18020/health',timeout=2).read().decode())"
+  $perceptionHealth = $healthJson | ConvertFrom-Json
+} catch {}
+$disks = @(
+  Get-CimInstance Win32_LogicalDisk -Filter "DriveType=3" |
+    ForEach-Object {
+      [ordered]@{
+        name=$_.DeviceID
+        size_bytes=[double]$_.Size
+        free_bytes=[double]$_.FreeSpace
+      }
+    }
+)
+[ordered]@{
+  schema_version="missioncore.worker-probe-raw/v1"
+  observed_at_utc=[DateTime]::UtcNow.ToString("o")
+  node_id=$env:COMPUTERNAME
+  os=[ordered]@{
+    caption=$os.Caption
+    version=$os.Version
+    uptime_seconds=([DateTime]::UtcNow - $os.LastBootUpTime.ToUniversalTime()).TotalSeconds
+  }
+  cpu=[ordered]@{
+    name=(@($cpu | Select-Object -ExpandProperty Name) -join " + ")
+    logical_processors=[int]$computer.NumberOfLogicalProcessors
+    load_percent=[double](($cpu | Measure-Object LoadPercentage -Average).Average)
+  }
+  memory=[ordered]@{
+    total_bytes=[double]$os.TotalVisibleMemorySize * 1024
+    free_bytes=[double]$os.FreePhysicalMemory * 1024
+  }
+  disks=$disks
+  gpu=$gpu
+  network=$network
+  docker_stats=$dockerStats
+  container_states=$containerStates
+  triton=[ordered]@{ready=$tritonReady; metrics=$tritonMetrics}
+  perception=$perceptionHealth
+} | ConvertTo-Json -Depth 12 -Compress
+""".strip()
+
+
+def _ssh_arguments(profile: WorkerConnectionProfile) -> list[str]:
+    arguments = [
+        "ssh",
+        "-o",
+        "BatchMode=yes",
+        "-o",
+        "StrictHostKeyChecking=yes",
+        "-o",
+        f"HostKeyAlias={profile.ssh_host_alias}",
+        "-o",
+        "ConnectTimeout=5",
+    ]
+    if profile.address:
+        arguments.extend(["-o", f"HostName={profile.address}"])
+    arguments.extend(
+        [
+            "-p",
+            str(profile.port),
+            profile.ssh_host_alias,
+            "powershell.exe",
+            "-NoLogo",
+            "-NoProfile",
+            "-NonInteractive",
+            "-Command",
+            (
+                "$encoded=[Console]::In.ReadToEnd();"
+                "Invoke-Expression "
+                "([Text.Encoding]::Unicode.GetString([Convert]::FromBase64String($encoded)))"
+            ),
+        ]
+    )
+    return arguments
+
+
+def run_worker_probe(profile: WorkerConnectionProfile) -> dict[str, Any]:
+    started = time.perf_counter()
+    try:
+        completed = subprocess.run(
+            _ssh_arguments(profile),
+            check=False,
+            capture_output=True,
+            input=base64.b64encode(WORKER_PROBE_POWERSHELL.encode("utf-16le")),
+            timeout=12,
+        )
+    except (OSError, subprocess.TimeoutExpired):
+        return _failed_probe(profile, "worker-unreachable", started)
+    if completed.returncode != 0:
+        return _failed_probe(profile, "worker-unreachable", started)
+    try:
+        raw = completed.stdout.decode("cp866")
+        document = json.loads(raw.strip())
+    except (UnicodeDecodeError, json.JSONDecodeError, TypeError):
+        return _failed_probe(profile, "invalid-worker-response", started)
+    if not isinstance(document, dict):
+        return _failed_probe(profile, "invalid-worker-response", started)
+    node_id = document.get("node_id")
+    if node_id != profile.expected_node_id:
+        return {
+            "schema_version": PROBE_SCHEMA,
+            "reachable": True,
+            "identity_matches": False,
+            "node_id": node_id if isinstance(node_id, str) else None,
+            "latency_ms": (time.perf_counter() - started) * 1000,
+            "observed_at_utc": _utc_now(),
+            "error_code": "worker-identity-mismatch",
+            "raw": None,
+        }
+    return {
+        "schema_version": PROBE_SCHEMA,
+        "reachable": True,
+        "identity_matches": True,
+        "node_id": node_id,
+        "latency_ms": (time.perf_counter() - started) * 1000,
+        "observed_at_utc": _utc_now(),
+        "error_code": None,
+        "raw": document,
+    }
+
+
+def _failed_probe(
+    profile: WorkerConnectionProfile,
+    error_code: str,
+    started: float,
+) -> dict[str, Any]:
+    return {
+        "schema_version": PROBE_SCHEMA,
+        "reachable": False,
+        "identity_matches": False,
+        "node_id": None,
+        "expected_node_id": profile.expected_node_id,
+        "latency_ms": (time.perf_counter() - started) * 1000,
+        "observed_at_utc": _utc_now(),
+        "error_code": error_code,
+        "raw": None,
+    }
+
+
+def _number(value: object) -> float | None:
+    if isinstance(value, bool) or not isinstance(value, int | float):
+        return None
+    result = float(value)
+    return result if math.isfinite(result) else None
+
+
+def _mapping(value: object) -> dict[str, Any]:
+    return value if isinstance(value, dict) else {}
+
+
+def _items(value: object) -> list[Any]:
+    return value if isinstance(value, list) else []
+
+
+def _percent(used: float | None, total: float | None) -> float | None:
+    if used is None or total is None or total <= 0:
+        return None
+    return max(0.0, min(100.0, used / total * 100))
+
+
+def _container_document(
+    name: str,
+    raw_stats: dict[str, Any],
+    raw_states: dict[str, Any],
+) -> dict[str, Any]:
+    stats = _mapping(raw_stats.get(name))
+    descriptor = _mapping(raw_states.get(name))
+    state = _mapping(descriptor.get("state"))
+    health = _mapping(state.get("Health"))
+    roles = {
+        "mission-core-triton": ("Inference Runtime", "mission-core", False),
+        "mission-core-perception-worker": ("Perception Pipeline", "mission-core", False),
+        "sentinel-frigate": ("Sentinel Frigate", "external", True),
+        "sentinel-ollama": ("Sentinel Ollama", "external", True),
+    }
+    role, owner, external = roles[name]
+    return {
+        "name": name,
+        "role": role,
+        "owner": owner,
+        "external": external,
+        "image": descriptor.get("image") if isinstance(descriptor.get("image"), str) else None,
+        "state": state.get("Status") if isinstance(state.get("Status"), str) else "unavailable",
+        "health": health.get("Status") if isinstance(health.get("Status"), str) else None,
+        "cpu_percent": _parse_percent(stats.get("CPUPerc")),
+        "memory_percent": _parse_percent(stats.get("MemPerc")),
+        "memory_usage": stats.get("MemUsage") if isinstance(stats.get("MemUsage"), str) else None,
+        "network_io": stats.get("NetIO") if isinstance(stats.get("NetIO"), str) else None,
+        "block_io": stats.get("BlockIO") if isinstance(stats.get("BlockIO"), str) else None,
+        "pids": _parse_integer(stats.get("PIDs")),
+    }
+
+
+def _parse_percent(value: object) -> float | None:
+    if not isinstance(value, str):
+        return None
+    try:
+        number = float(value.rstrip("%"))
+    except ValueError:
+        return None
+    return number if math.isfinite(number) else None
+
+
+def _parse_integer(value: object) -> int | None:
+    if not isinstance(value, str):
+        return None
+    try:
+        return int(value)
+    except ValueError:
+        return None
+
+
+def _triton_metrics(lines: object) -> dict[str, float | None]:
+    totals: dict[str, float] = {
+        "nv_inference_request_success": 0,
+        "nv_inference_request_failure": 0,
+        "nv_inference_count": 0,
+    }
+    matched: set[str] = set()
+    for line in _items(lines):
+        if not isinstance(line, str):
+            continue
+        match = PROMETHEUS_SAMPLE.fullmatch(line.strip())
+        if match is None:
+            continue
+        try:
+            value = float(match.group("value"))
+        except ValueError:
+            continue
+        if math.isfinite(value):
+            name = match.group("name")
+            totals[name] += value
+            matched.add(name)
+    return {
+        "requests_succeeded": (
+            totals["nv_inference_request_success"]
+            if "nv_inference_request_success" in matched
+            else None
+        ),
+        "requests_failed": (
+            totals["nv_inference_request_failure"]
+            if "nv_inference_request_failure" in matched
+            else None
+        ),
+        "inferences": (
+            totals["nv_inference_count"] if "nv_inference_count" in matched else None
+        ),
+    }
+
+
+def _pipeline_document(raw: dict[str, Any]) -> dict[str, Any]:
+    perception = _mapping(raw.get("perception"))
+    current_stage = perception.get("current_stage")
+    if not isinstance(current_stage, str):
+        current_stage = None
+    busy = perception.get("state") == "busy"
+    stages = (
+        ("source-ingress", "Приём сенсорного потока"),
+        ("camera-decode", "Декодирование камеры"),
+        ("preprocessing", "Предобработка"),
+        ("detector", "Детектор объектов"),
+        ("semantic-model", "Семантическая модель"),
+        ("sensor-fusion", "Camera ↔ LiDAR fusion"),
+        ("tracking", "Трекинг"),
+        ("temporal-state", "Временное состояние"),
+        ("result-publication", "Публикация результата"),
+    )
+
+    def stage_state(stage_id: str) -> str:
+        if not perception:
+            return "unavailable"
+        if not busy:
+            return "ready"
+        if stage_id == current_stage or stage_id in {
+            "source-ingress",
+            "camera-decode",
+            "semantic-model",
+        }:
+            return "active"
+        return "waiting"
+
+    return {
+        "service_state": (
+            perception.get("state") if isinstance(perception.get("state"), str) else "unavailable"
+        ),
+        "current_stage": current_stage,
+        "active_request_id": (
+            perception.get("active_request_id")
+            if isinstance(perception.get("active_request_id"), str)
+            else None
+        ),
+        "active_frame_index": (
+            int(perception["active_frame_index"])
+            if isinstance(perception.get("active_frame_index"), int)
+            else None
+        ),
+        "completed_runs": (
+            int(perception["completed_runs"])
+            if isinstance(perception.get("completed_runs"), int)
+            else None
+        ),
+        "failed_runs": (
+            int(perception["failed_runs"])
+            if isinstance(perception.get("failed_runs"), int)
+            else None
+        ),
+        "model_load_seconds": _number(perception.get("model_load_seconds")),
+        "stages": [
+            {
+                "id": stage_id,
+                "label": label,
+                "state": stage_state(stage_id),
+            }
+            for stage_id, label in stages
+        ],
+    }
+
+
+class WorkerTelemetryService:
+    def __init__(
+        self,
+        store: WorkerProfileStore,
+        probe_runner: ProbeRunner = run_worker_probe,
+        *,
+        cache_seconds: float = 1.5,
+    ) -> None:
+        self.store = store
+        self.probe_runner = probe_runner
+        self.cache_seconds = cache_seconds
+        self._lock = threading.Lock()
+        self._cached_at = 0.0
+        self._cached: dict[str, Any] | None = None
+        self._previous_network: tuple[float, float, float] | None = None
+        self._history: deque[dict[str, Any]] = deque(maxlen=300)
+
+    def profile_document(self) -> dict[str, Any]:
+        return {
+            "schema_version": PROFILE_SCHEMA,
+            "profile": self.store.read().model_dump(mode="json"),
+            "security": {
+                "transport": "ssh",
+                "host_key_policy": "strict-pinned",
+                "credentials_managed_by_ui": False,
+                "ssh_host_alias": SSH_HOST_ALIAS,
+            },
+        }
+
+    def test_profile(self, request: WorkerConnectionProfilePut) -> dict[str, Any]:
+        current = self.store.read()
+        candidate = current.model_copy(
+            update={"address": request.address, "port": request.port}
+        )
+        return self._public_probe(self.probe_runner(candidate))
+
+    def apply_profile(self, request: WorkerConnectionProfilePut) -> dict[str, Any]:
+        current = self.store.read()
+        if request.revision != current.revision:
+            raise RuntimeError("worker profile revision changed")
+        candidate = current.model_copy(
+            update={"address": request.address, "port": request.port}
+        )
+        probe = self.probe_runner(candidate)
+        if not probe.get("reachable") or not probe.get("identity_matches"):
+            raise ConnectionError(str(probe.get("error_code") or "worker-unreachable"))
+        saved = self.store.save(request)
+        with self._lock:
+            self._cached = None
+            self._cached_at = 0
+            self._previous_network = None
+            self._history.clear()
+        return {
+            **self.profile_document(),
+            "verification": self._public_probe(probe),
+            "profile": saved.model_dump(mode="json"),
+        }
+
+    def snapshot(self, history_limit: int) -> dict[str, Any]:
+        with self._lock:
+            now = time.monotonic()
+            if self._cached is not None and now - self._cached_at < self.cache_seconds:
+                return {
+                    **self._cached,
+                    "history": list(self._history)[-history_limit:],
+                }
+            profile = self.store.read()
+            probe = self.probe_runner(profile)
+            document = self._telemetry_document(profile, probe, now)
+            self._cached = document
+            self._cached_at = now
+            return {
+                **document,
+                "history": list(self._history)[-history_limit:],
+            }
+
+    def _telemetry_document(
+        self,
+        profile: WorkerConnectionProfile,
+        probe: dict[str, Any],
+        monotonic_now: float,
+    ) -> dict[str, Any]:
+        public_probe = self._public_probe(probe)
+        if not probe.get("reachable") or not probe.get("identity_matches"):
+            return {
+                "schema_version": TELEMETRY_SCHEMA,
+                "profile": profile.model_dump(mode="json"),
+                "connection": public_probe,
+                "node": None,
+                "runtimes": [],
+                "pipeline": _pipeline_document({}),
+                "network": {"interfaces": [], "aggregate": None},
+            }
+        raw = _mapping(probe.get("raw"))
+        memory = _mapping(raw.get("memory"))
+        memory_total = _number(memory.get("total_bytes"))
+        memory_free = _number(memory.get("free_bytes"))
+        memory_used = (
+            memory_total - memory_free
+            if memory_total is not None and memory_free is not None
+            else None
+        )
+        gpu = _mapping(raw.get("gpu"))
+        gpu_used = _number(gpu.get("memory_used_mib"))
+        gpu_total = _number(gpu.get("memory_total_mib"))
+        interfaces = [
+            self._network_interface(item)
+            for item in _items(raw.get("network"))
+            if isinstance(item, dict)
+        ]
+        received = sum(
+            value
+            for item in interfaces
+            if (value := _number(item.get("received_bytes"))) is not None
+        )
+        sent = sum(
+            value
+            for item in interfaces
+            if (value := _number(item.get("sent_bytes"))) is not None
+        )
+        receive_rate: float | None = None
+        send_rate: float | None = None
+        if self._previous_network is not None:
+            previous_at, previous_received, previous_sent = self._previous_network
+            elapsed = monotonic_now - previous_at
+            if elapsed > 0 and received >= previous_received and sent >= previous_sent:
+                receive_rate = (received - previous_received) / elapsed
+                send_rate = (sent - previous_sent) / elapsed
+        self._previous_network = (monotonic_now, received, sent)
+        raw_stats = _mapping(raw.get("docker_stats"))
+        raw_states = _mapping(raw.get("container_states"))
+        runtimes = [
+            _container_document(name, raw_stats, raw_states) for name in CONTAINER_NAMES
+        ]
+        triton = _mapping(raw.get("triton"))
+        triton_document = {
+            "ready": triton.get("ready") is True,
+            **_triton_metrics(triton.get("metrics")),
+        }
+        memory_used_percent = _percent(memory_used, memory_total)
+        gpu_memory_used_percent = _percent(gpu_used, gpu_total)
+        node: dict[str, Any] = {
+            "node_id": raw.get("node_id"),
+            "observed_at_utc": raw.get("observed_at_utc"),
+            "os": _mapping(raw.get("os")),
+            "cpu": _mapping(raw.get("cpu")),
+            "memory": {
+                "total_bytes": memory_total,
+                "used_bytes": memory_used,
+                "free_bytes": memory_free,
+                "used_percent": memory_used_percent,
+            },
+            "disks": _items(raw.get("disks")),
+            "gpu": {
+                **gpu,
+                "memory_used_percent": gpu_memory_used_percent,
+            }
+            if gpu
+            else None,
+            "triton": triton_document,
+        }
+        network = {
+            "interfaces": interfaces,
+            "aggregate": {
+                "received_bytes": received,
+                "sent_bytes": sent,
+                "receive_bytes_per_second": receive_rate,
+                "send_bytes_per_second": send_rate,
+            },
+        }
+        history_row = {
+            "observed_at_utc": raw.get("observed_at_utc") or _utc_now(),
+            "cpu_percent": _number(_mapping(raw.get("cpu")).get("load_percent")),
+            "memory_percent": memory_used_percent,
+            "gpu_percent": _number(gpu.get("utilization_percent")),
+            "gpu_memory_percent": gpu_memory_used_percent,
+            "network_receive_bytes_per_second": receive_rate,
+            "network_send_bytes_per_second": send_rate,
+        }
+        self._history.append(history_row)
+        return {
+            "schema_version": TELEMETRY_SCHEMA,
+            "profile": profile.model_dump(mode="json"),
+            "connection": public_probe,
+            "node": node,
+            "runtimes": runtimes,
+            "pipeline": _pipeline_document(raw),
+            "network": network,
+        }
+
+    @staticmethod
+    def _network_interface(item: dict[str, Any]) -> dict[str, Any]:
+        addresses = [
+            address for address in _items(item.get("addresses")) if isinstance(address, str)
+        ]
+        return {
+            "name": item.get("name") if isinstance(item.get("name"), str) else "unknown",
+            "description": (
+                item.get("description") if isinstance(item.get("description"), str) else None
+            ),
+            "status": item.get("status") if isinstance(item.get("status"), str) else None,
+            "link_speed_bps": _number(item.get("link_speed_bps")),
+            "mac_address": (
+                item.get("mac_address") if isinstance(item.get("mac_address"), str) else None
+            ),
+            "addresses": addresses,
+            "received_bytes": _number(item.get("received_bytes")),
+            "sent_bytes": _number(item.get("sent_bytes")),
+        }
+
+    @staticmethod
+    def _public_probe(probe: dict[str, Any]) -> dict[str, Any]:
+        return {
+            "schema_version": PROBE_SCHEMA,
+            "reachable": probe.get("reachable") is True,
+            "identity_matches": probe.get("identity_matches") is True,
+            "node_id": probe.get("node_id") if isinstance(probe.get("node_id"), str) else None,
+            "latency_ms": _number(probe.get("latency_ms")),
+            "observed_at_utc": (
+                probe.get("observed_at_utc")
+                if isinstance(probe.get("observed_at_utc"), str)
+                else _utc_now()
+            ),
+            "error_code": (
+                probe.get("error_code")
+                if isinstance(probe.get("error_code"), str)
+                else None
+            ),
+        }
+
+
+def build_system_telemetry_router(
+    *,
+    root_provider: RootProvider,
+    probe_runner: ProbeRunner = run_worker_probe,
+) -> APIRouter:
+    store = WorkerProfileStore(root_provider())
+    service = WorkerTelemetryService(store, probe_runner)
+    router = APIRouter(prefix="/api/v1/system", tags=["system"])
+
+    @router.get("/worker-profile")
+    def get_worker_profile() -> dict[str, Any]:
+        return service.profile_document()
+
+    @router.post("/worker-profile/test")
+    def test_worker_profile(request: WorkerConnectionProfilePut) -> dict[str, Any]:
+        return service.test_profile(request)
+
+    @router.put("/worker-profile")
+    def put_worker_profile(request: WorkerConnectionProfilePut) -> dict[str, Any]:
+        try:
+            return service.apply_profile(request)
+        except RuntimeError as exc:
+            raise HTTPException(
+                status_code=409,
+                detail="Профиль Worker 006 был изменён в другой сессии.",
+            ) from exc
+        except ConnectionError as exc:
+            raise HTTPException(
+                status_code=409,
+                detail="Новый адрес не прошёл проверку узла Worker 006.",
+            ) from exc
+
+    @router.get("/worker-telemetry")
+    def get_worker_telemetry(
+        history: int = Query(default=90, ge=1, le=300),
+    ) -> dict[str, Any]:
+        return service.snapshot(history)
+
+    return router
diff --git a/tests/test_system_telemetry_api.py b/tests/test_system_telemetry_api.py
new file mode 100644
index 0000000..cf54ab1
--- /dev/null
+++ b/tests/test_system_telemetry_api.py
@@ -0,0 +1,237 @@
+from __future__ import annotations
+
+import stat
+from collections.abc import Callable
+from pathlib import Path
+from typing import Any
+
+import pytest
+from fastapi import APIRouter, HTTPException
+from fastapi.routing import APIRoute
+from pydantic import ValidationError
+
+from k1link.web.system_telemetry_api import (
+    EXPECTED_NODE_ID,
+    WorkerConnectionProfile,
+    WorkerConnectionProfilePut,
+    WorkerProfileStore,
+    WorkerTelemetryService,
+    _ssh_arguments,
+    build_system_telemetry_router,
+)
+
+
+def _endpoint(router: APIRouter, path: str, method: str) -> Callable[..., Any]:
+    for route in router.routes:
+        if (
+            isinstance(route, APIRoute)
+            and route.path == path
+            and route.methods is not None
+            and method in route.methods
+        ):
+            return route.endpoint
+    raise AssertionError(f"{method} {path} route is missing")
+
+
+def _probe(
+    *,
+    node_id: str = EXPECTED_NODE_ID,
+    received_bytes: float = 1_000,
+    sent_bytes: float = 500,
+) -> dict[str, Any]:
+    return {
+        "reachable": True,
+        "identity_matches": node_id == EXPECTED_NODE_ID,
+        "node_id": node_id,
+        "latency_ms": 12.5,
+        "observed_at_utc": "2026-07-27T12:00:00Z",
+        "error_code": None,
+        "raw": {
+            "node_id": node_id,
+            "observed_at_utc": "2026-07-27T12:00:00Z",
+            "os": {
+                "caption": "Windows 11 Pro",
+                "version": "10.0",
+                "uptime_seconds": 100,
+            },
+            "cpu": {
+                "name": "CPU",
+                "logical_processors": 32,
+                "load_percent": 25,
+            },
+            "memory": {
+                "total_bytes": 1000,
+                "free_bytes": 400,
+            },
+            "disks": [],
+            "gpu": {
+                "name": "GPU",
+                "utilization_percent": 30,
+                "memory_used_mib": 100,
+                "memory_total_mib": 200,
+            },
+            "network": [
+                {
+                    "name": "LAN",
+                    "description": "Ethernet",
+                    "status": "Up",
+                    "link_speed_bps": 1_000_000_000,
+                    "addresses": ["192.0.2.10"],
+                    "received_bytes": received_bytes,
+                    "sent_bytes": sent_bytes,
+                }
+            ],
+            "docker_stats": {
+                "mission-core-triton": {
+                    "Name": "mission-core-triton",
+                    "CPUPerc": "2.5%",
+                    "MemPerc": "3.5%",
+                    "MemUsage": "1GiB / 32GiB",
+                    "NetIO": "1MB / 2MB",
+                    "BlockIO": "0B / 0B",
+                    "PIDs": "12",
+                },
+                "sentinel-frigate": {
+                    "Name": "sentinel-frigate",
+                    "CPUPerc": "150%",
+                    "MemPerc": "20%",
+                    "MemUsage": "6GiB / 32GiB",
+                    "NetIO": "3GB / 1GB",
+                    "BlockIO": "1GB / 1GB",
+                    "PIDs": "200",
+                },
+            },
+            "container_states": {
+                "mission-core-triton": {
+                    "state": {"Status": "running", "Health": {"Status": "healthy"}},
+                    "image": "triton@sha256:accepted",
+                },
+                "sentinel-frigate": {
+                    "state": {"Status": "running", "Health": {"Status": "healthy"}},
+                    "image": "frigate@sha256:external",
+                },
+            },
+            "triton": {
+                "ready": True,
+                "metrics": [
+                    'nv_inference_request_success{model="detector"} 12',
+                    'nv_inference_request_failure{model="detector"} 0',
+                ],
+            },
+            "perception": {
+                "state": "busy",
+                "current_stage": "detector",
+                "active_request_id": "run-001",
+                "active_frame_index": 42,
+                "completed_runs": 3,
+                "failed_runs": 0,
+                "model_load_seconds": 10.5,
+            },
+        },
+    }
+
+
+def test_worker_profile_rejects_ssh_option_injection() -> None:
+    for unsafe in (
+        "-oProxyCommand=touch /tmp/unsafe",
+        "worker.local -p 2200",
+        "worker.local/../../unsafe",
+        "worker.local\nHost evil",
+    ):
+        with pytest.raises(ValidationError):
+            WorkerConnectionProfile(address=unsafe)
+
+    assert WorkerConnectionProfile(address="192.0.2.15").address == "192.0.2.15"
+    assert WorkerConnectionProfile(address="worker-006.local").address == "worker-006.local"
+
+
+def test_worker_profile_is_atomic_versioned_and_private(tmp_path: Path) -> None:
+    store = WorkerProfileStore(tmp_path / "system")
+    initial = store.read()
+    assert initial.revision == 0
+    saved = store.save(
+        WorkerConnectionProfilePut(
+            revision=0,
+            address="192.0.2.15",
+            port=2200,
+        )
+    )
+
+    assert saved.revision == 1
+    assert store.read() == saved
+    assert stat.S_IMODE(store.path.stat().st_mode) == 0o600
+    with pytest.raises(RuntimeError, match="revision changed"):
+        store.save(
+            WorkerConnectionProfilePut(
+                revision=0,
+                address="192.0.2.16",
+                port=22,
+            )
+        )
+
+
+def test_ssh_command_keeps_identity_pinned_and_values_as_arguments() -> None:
+    arguments = _ssh_arguments(
+        WorkerConnectionProfile(address="192.0.2.15", port=2200)
+    )
+    assert "BatchMode=yes" in arguments
+    assert "StrictHostKeyChecking=yes" in arguments
+    assert "HostKeyAlias=mission-gpu" in arguments
+    assert "HostName=192.0.2.15" in arguments
+    assert arguments[arguments.index("-p") + 1] == "2200"
+    assert "mission-gpu" in arguments
+    assert "192.0.2.15; touch unsafe" not in arguments
+
+
+def test_worker_telemetry_separates_mission_core_and_external_load(
+    tmp_path: Path,
+) -> None:
+    service = WorkerTelemetryService(
+        WorkerProfileStore(tmp_path / "system"),
+        lambda _: _probe(),
+        cache_seconds=0,
+    )
+    document = service.snapshot(10)
+    runtimes = {runtime["name"]: runtime for runtime in document["runtimes"]}
+
+    assert document["connection"]["identity_matches"] is True
+    assert document["node"]["memory"]["used_percent"] == 60
+    assert document["node"]["gpu"]["memory_used_percent"] == 50
+    assert document["node"]["triton"]["requests_succeeded"] == 12
+    assert runtimes["mission-core-triton"]["external"] is False
+    assert runtimes["mission-core-triton"]["cpu_percent"] == 2.5
+    assert runtimes["sentinel-frigate"]["external"] is True
+    assert runtimes["sentinel-frigate"]["cpu_percent"] == 150
+    assert document["pipeline"]["active_request_id"] == "run-001"
+    assert next(
+        stage
+        for stage in document["pipeline"]["stages"]
+        if stage["id"] == "detector"
+    )["state"] == "active"
+    assert next(
+        stage
+        for stage in document["pipeline"]["stages"]
+        if stage["id"] == "semantic-model"
+    )["state"] == "active"
+
+
+def test_profile_apply_fails_closed_on_wrong_node_and_keeps_old_profile(
+    tmp_path: Path,
+) -> None:
+    router = build_system_telemetry_router(
+        root_provider=lambda: tmp_path / "system",
+        probe_runner=lambda _: _probe(node_id="OTHER-NODE"),
+    )
+    apply_profile = _endpoint(router, "/api/v1/system/worker-profile", "PUT")
+
+    with pytest.raises(HTTPException) as error:
+        apply_profile(
+            WorkerConnectionProfilePut(
+                revision=0,
+                address="192.0.2.20",
+                port=22,
+            )
+        )
+    assert error.value.status_code == 409
+    assert WorkerProfileStore(tmp_path / "system").read().revision == 0
+    assert not WorkerProfileStore(tmp_path / "system").path.exists()