Auto-merge PR #27

chore(deps-dev): bump @biomejs/biome from 2.5.8 to 2.5.9 in the biome group

next.config.ts

@@ -5,6 +5,16 @@ const nextConfig: NextConfig = {
   turbopack: {
     root: process.cwd(),
   },
+  // Nix store is read-only at runtime; Next.js tries to mkdir
+  // <distDir>/cache/images for the on-disk image optimizer LRU and
+  // crashes with EACCES unhandledRejection on every optimized image
+  // request. Disable the disk cache (and disable ISR flush which gates it).
+  images: {
+    maximumDiskCacheSize: 0,
+  },
+  experimental: {
+    isrFlushToDisk: false,
+  },
 };
 
 export default nextConfig;

package.json

@@ -10,30 +10,30 @@
     "format": "biome format --write"
   },
   "dependencies": {
-    "@base-ui/react": "^1.6.0",
-    "@shadcn/react": "^0.2.1",
+    "@base-ui/react": "^1.7.0",
+    "@shadcn/react": "^0.3.0",
     "class-variance-authority": "^0.7.1",
     "clsx": "^2.1.1",
     "cmdk": "^1.1.1",
     "date-fns": "^4.4.0",
     "embla-carousel-react": "^8.6.0",
-    "framer-motion": "^12.42.2",
+    "framer-motion": "^13.1.0",
     "input-otp": "^1.4.2",
-    "lucide-react": "^1.26.0",
-    "next": "16.2.12",
+    "lucide-react": "^1.31.0",
+    "next": "16.3.1",
     "next-themes": "^0.4.6",
     "react": "19.2.8",
     "react-day-picker": "^10.0.1",
     "react-dom": "19.2.8",
-    "react-resizable-panels": "^4.12.2",
+    "react-resizable-panels": "^4.12.3",
     "recharts": "3.10.1",
-    "shadcn": "^4.14.1",
-    "sonner": "^2.0.7",
+    "shadcn": "^4.18.0",
+    "sonner": "^2.0.8",
     "tailwind-merge": "^3.6.0",
     "tw-animate-css": "^1.4.0"
   },
   "devDependencies": {
-    "@biomejs/biome": "2.5.5",
+    "@biomejs/biome": "2.5.9",
     "@tailwindcss/postcss": "^4",
     "@types/node": "^26",
     "@types/react": "^19",

src/app/api/dashboard/route.ts

@@ -1,16 +1,10 @@
-import http from "node:http";
+import { execFile } from "node:child_process";
+import { promisify } from "node:util";
 import { NextResponse } from "next/server";
 
-const JAEGER = "http://jaeger:16686";
-const PROMETHEUS = "http://prometheus:9090";
-const DOCKER_SOCK = "/var/run/docker.sock";
+const execFileAsync = promisify(execFile);
 
-interface Container {
-  Names: string[];
-  State: string;
-  Labels: Record<string, string>;
-  NetworkSettings?: { Networks?: Record<string, unknown> };
-}
+const PROMETHEUS = "http://127.0.0.1:9090";
 
 interface Trace {
   service: string;
@@ -26,24 +20,70 @@ interface Service {
   hasWeb: boolean;
 }
 
-function dockerFetch(path: string): Promise<unknown> {
-  return new Promise((resolve, reject) => {
-    http
-      .get({ socketPath: DOCKER_SOCK, path }, (res) => {
-        let data = "";
-        res.on("data", (c: string) => {
-          data += c;
-        });
-        res.on("end", () => {
-          try {
-            resolve(JSON.parse(data));
-          } catch {
-            resolve(null);
-          }
-        });
-      })
-      .on("error", reject);
-  });
+// Service-owning systemd units that should show up on the dashboard.
+// GMW, Booster, etc. are listed even though they are headless (no web).
+const MONITORED_UNITS = [
+  "caddy",
+  "9router",
+  "booster-role",
+  "gmw-backend",
+  "gmw-discord-gateway",
+  "hub",
+  "lidm-backend",
+  "lidm-frontend",
+  "llm-api",
+  "nats",
+  "node-exporter",
+  "otel",
+  "pr-agent-server",
+  "prometheus",
+  "scraper",
+  "teleuploader",
+  "tools-frontend",
+  "tools-gateway",
+  "tools-workers",
+  "zeavis-api",
+  "zeavis-ml-service",
+  "zeavis-web",
+];
+
+// Units with a public HTTPS site behind Caddy.
+const WEB_UNITS = new Set([
+  "caddy",
+  "9router",
+  "hub",
+  "lidm-frontend",
+  "pr-agent-server",
+  "scraper",
+  "teleuploader",
+  "tools-frontend",
+  "zeavis-web",
+]);
+
+async function systemdServices(): Promise<Service[]> {
+  const services: Service[] = [];
+  for (const unit of MONITORED_UNITS) {
+    try {
+      const { stdout } = await execFileAsync("systemctl", [
+        "show",
+        `${unit}.service`,
+        "-p",
+        "ActiveState",
+        "-p",
+        "LoadState",
+        "--no-pager",
+      ]);
+      const state = stdout.match(/ActiveState=(\w+)/)?.[1] ?? "unknown";
+      services.push({
+        name: unit,
+        state: state === "active" ? "running" : state,
+        hasWeb: WEB_UNITS.has(unit),
+      });
+    } catch {
+      // unit doesn't exist — skip silently
+    }
+  }
+  return services;
 }
 
 async function fetchJSON(url: string): Promise<unknown> {
@@ -55,73 +95,6 @@ async function fetchJSON(url: string): Promise<unknown> {
   }
 }
 
-function parseServices(containers: Container[]): Service[] {
-  const project = containers.find((c) => c.Labels["com.docker.compose.project"])
-    ?.Labels["com.docker.compose.project"];
-
-  return containers
-    .filter((c) => {
-      if (!c.NetworkSettings?.Networks?.["app-shared-net"]) return false;
-      if (project && c.Labels["com.docker.compose.project"] !== project)
-        return false;
-      return true;
-    })
-    .map((c) => ({
-      name: c.Names[0].replace(/^\//, ""),
-      state: c.State,
-      hasWeb: Object.keys(c.Labels).some(
-        (k) => k.startsWith("traefik.http.routers.") && k.endsWith(".rule"),
-      ),
-    }));
-}
-
-async function fetchTraces(): Promise<Trace[]> {
-  const svcRes = await fetchJSON(`${JAEGER}/api/services`);
-  if (!svcRes || !Array.isArray((svcRes as { data?: string[] }).data))
-    return [];
-
-  const now = Date.now() * 1000;
-  const start = now - 5 * 60 * 1_000_000;
-  const all: Trace[] = [];
-
-  for (const service of (svcRes as { data: string[] }).data) {
-    const d = await fetchJSON(
-      `${JAEGER}/api/traces?service=${encodeURIComponent(service)}&start=${start}&end=${now}&limit=5&lookback=5m`,
-    );
-    if (!d || !Array.isArray((d as { data?: unknown[] }).data)) continue;
-
-    for (const t of (
-      d as {
-        data: {
-          duration: number;
-          spans: {
-            operationName: string;
-            processID: string;
-            tags: { key: string; value: unknown }[];
-          }[];
-          processes: Record<string, { serviceName: string }>;
-        }[];
-      }
-    ).data) {
-      if (!t.spans?.length) continue;
-      const span = t.spans[0];
-      const svc = t.processes[span.processID]?.serviceName || "unknown";
-      const hasError = t.spans.some((s) =>
-        s.tags?.some((tag) => tag.key === "error" && tag.value === true),
-      );
-      all.push({
-        service: svc,
-        operation: span.operationName,
-        duration: t.duration,
-        spans: t.spans.length,
-        hasError,
-      });
-    }
-  }
-
-  return all.sort((a, b) => b.duration - a.duration).slice(0, 20);
-}
-
 async function promRange(query: string, steps = 20): Promise<number[]> {
   const now = Math.floor(Date.now() / 1000);
   const u = `${PROMETHEUS}/api/v1/query_range?query=${encodeURIComponent(query)}&start=${now - 300}&end=${now}&step=${(300 / steps).toFixed(0)}`;
@@ -164,18 +137,20 @@ interface DashboardData {
   errors: number[];
   traceVolume: number[];
   links: { url: string; label: string }[];
+  llm: {
+    reqRate: number | null;
+    tokRate: number | null;
+    tokSpeed: number | null;
+    uptime: number | null;
+    reqSpark: number[];
+    tokSpark: number[];
+  };
 }
 
 export async function GET() {
-  const [containersRaw] = await Promise.all([
-    dockerFetch("/containers/json?all=true") as Promise<Container[]>,
-  ]);
-
-  const containers = Array.isArray(containersRaw) ? containersRaw : [];
-  const services = parseServices(containers);
+  const services = await systemdServices();
 
   const [
-    traces,
     cpu,
     ram,
     disk,
@@ -184,11 +159,11 @@ export async function GET() {
     load15,
     netIn,
     netOut,
-    rps,
-    latency,
-    errors,
+    netInSpark,
+    loadSpark,
+    netOutSpark,
+    upstreamCount,
   ] = await Promise.all([
-    fetchTraces(),
     promQuery(
       `100 - (avg by(instance)(rate(node_cpu_seconds_total{mode="idle"}[1m])) * 100)`,
     ),
@@ -203,42 +178,78 @@ export async function GET() {
     promQuery("node_load15"),
     promQuery(`rate(node_network_receive_bytes_total{device!="lo"}[1m])`),
     promQuery(`rate(node_network_transmit_bytes_total{device!="lo"}[1m])`),
-    promRange("sum(rate(traefik_service_requests_total[1m]))"),
+    promRange('sum(rate(node_network_receive_bytes_total{device!="lo"}[1m]))'),
+    promRange("avg(node_load1)"),
+    promRange('sum(rate(node_network_transmit_bytes_total{device!="lo"}[1m]))'),
+    promQuery("count(up)"),
+  ]);
+
+  // ── Error rate (llm-api) & request rate (any /metrics service) ──
+  const [errRate, errSpark] = await Promise.all([
+    promQuery(`sum(rate(llm_api_errors_total[5m]))`),
+    promRange(`sum(rate(llm_api_errors_total[1m]))`),
+  ]);
+  const reqRate = await promQuery(
+    `sum(rate(llm_api_requests_total[5m])) + sum(rate(http_server_request_count_total[5m]))`,
+  );
+  const reqSpark = await promRange(
+    `sum(rate(llm_api_requests_total[1m])) + sum(rate(http_server_request_count_total[1m]))`,
+  );
+
+  // ── LLM inference metrics (llm-api /metrics) ──
+  const [
+    llmReqRate,
+    llmTokRate,
+    llmTokSpeed,
+    llmUptime,
+    llmReqSpark,
+    llmTokSpark,
+  ] = await Promise.all([
+    promQuery(
+      `sum(rate(llm_api_requests_total{service="llm-api"}[5m]))`,
+    ),
+    promQuery(
+      `sum(rate(llm_api_completion_tokens_total{service="llm-api"}[5m]))`,
+    ),
+    promQuery(
+      `avg(llm_api_tokens_per_second{service="llm-api"})`,
+    ),
+    promQuery(`llm_api_uptime_seconds{service="llm-api"}`),
     promRange(
-      "avg(traefik_service_request_duration_seconds_sum / traefik_service_request_duration_seconds_count) * 1000",
+      `sum(rate(llm_api_requests_total{service="llm-api"}[1m]))`,
+    ),
+    promRange(
+      `sum(rate(llm_api_completion_tokens_total{service="llm-api"}[1m]))`,
     ),
-    promRange('sum(rate(traefik_service_requests_total{code=~"5.."}[1m]))'),
   ]);
 
   const links: { url: string; label: string }[] = [
-    { url: "/jaeger", label: "Jaeger UI" },
+    { url: "https://github.com/asepharyana/asepharyana-hub", label: "GitHub" },
   ];
-  if (containers.some((c) => c.Names?.some((n) => n.includes("prometheus")))) {
-    links.push({ url: "/api/prometheus/targets", label: "Prometheus" });
-  }
-  links.push({
-    url: "https://github.com/asepharyana/asepharyana-hub",
-    label: "GitHub",
-  });
-
-  const domains = ["asepharyana.my.id", "asepharyana.web.id"];
+  const domains = ["asepharyana.my.id"];
   for (const s of services) {
     if (s.hasWeb && s.state === "running") {
       links.push({ url: `https://${s.name}.${domains[0]}`, label: s.name });
     }
   }
 
-  const traceVolume = [traces.length];
-
   const data: DashboardData = {
     services,
-    traces,
+    traces: [],
     node: { cpu, ram, disk, load1, load5, load15, netIn, netOut },
-    rps,
-    latency,
-    errors,
-    traceVolume,
+    rps: reqSpark,
+    latency: loadSpark,
+    errors: errSpark,
+    traceVolume: netInSpark,
     links,
+    llm: {
+      reqRate: llmReqRate,
+      tokRate: llmTokRate,
+      tokSpeed: llmTokSpeed,
+      uptime: llmUptime,
+      reqSpark: llmReqSpark,
+      tokSpark: llmTokSpark,
+    },
   };
 
   return NextResponse.json(data);

src/app/dashboard/page.tsx

@@ -10,6 +10,7 @@ import { Card, CardContent, CardHeader, CardTitle } from "@/components/ui/card";
 import { AnimatedNumber } from "@/components/ui/motion-primitives";
 import {
   type DashboardData,
+  fmtUptime,
   gaugeColor,
   safeDur,
   serviceIndicator,
@@ -59,8 +60,8 @@ export default function Dashboard() {
     data?.services.filter((s) => s.state === "running").length ?? 0;
   const degraded = (data?.services.length ?? 0) - running;
   const hasNode = data?.node.cpu !== null || data?.node.ram !== null;
-  const hasTraffik =
-    (data?.rps?.length ?? 0) > 0 || (data?.latency?.length ?? 0) > 0;
+  const hasTraffic =
+    (data?.rps?.length ?? 0) > 0 || (data?.errors?.length ?? 0) > 0;
   const node = data?.node;
 
   return (
@@ -146,7 +147,15 @@ export default function Dashboard() {
                   label="Traces"
                   color="text-blue-400"
                 />
-                <StatBox value={0} label="Errors" color="text-red-400" />
+                <StatBox
+                  value={
+                    data?.errors?.length
+                      ? data.errors[data.errors.length - 1].toFixed(2)
+                      : undefined
+                  }
+                  label="LLM err/s"
+                  color="text-red-400"
+                />
               </div>
             </CardContent>
           </Card>
@@ -230,8 +239,75 @@ export default function Dashboard() {
           </motion.div>
         )}
 
+        {/* LLM Inference */}
+        <motion.div variants={cardVariants}>
+          <Card className="transition-shadow duration-300 hover:shadow-md hover:shadow-border/20">
+            <CardHeader className="flex flex-row items-center justify-between">
+              <CardTitle>LLM Inference</CardTitle>
+              <Badge variant="secondary" className="font-mono text-[10px] font-normal">
+                {data?.llm.uptime != null
+                  ? `${fmtUptime(data.llm.uptime)} up`
+                  : "no data"}
+              </Badge>
+            </CardHeader>
+            <CardContent>
+              <div className="grid grid-cols-2 gap-1.5">
+                <StatBox
+                  value={
+                    data?.llm.reqRate != null
+                      ? data.llm.reqRate.toFixed(1)
+                      : undefined
+                  }
+                  label="Req/s (5m)"
+                  color="text-blue-400"
+                />
+                <StatBox
+                  value={
+                    data?.llm.tokRate != null
+                      ? `${data.llm.tokRate.toFixed(0)}`
+                      : undefined
+                  }
+                  label="Tokens/s (5m)"
+                  color="text-green-400"
+                />
+                <StatBox
+                  value={
+                    data?.llm.tokSpeed != null
+                      ? `${data.llm.tokSpeed.toFixed(0)}`
+                      : undefined
+                  }
+                  label="Gen tok/s"
+                  color="text-purple-400"
+                />
+                <StatBox
+                  value={data?.llm.reqSpark?.length ? "live" : undefined}
+                  label="Sparkline"
+                  color="text-yellow-400"
+                />
+              </div>
+              {data?.llm.reqSpark?.length ? (
+                <div className="mt-2">
+                  <Sparkline data={data.llm.reqSpark} color="#58a6ff" />
+                </div>
+              ) : (
+                <p className="mt-2 text-center text-[10px] text-muted-foreground">
+                  No traffic yet &mdash; send a prompt on{" "}
+                  <a
+                    className="text-blue-400 underline"
+                    href="https://ai.asepharyana.my.id"
+                    target="_blank"
+                    rel="noopener noreferrer"
+                  >
+                    ai.asepharyana.my.id
+                  </a>
+                </p>
+              )}
+            </CardContent>
+          </Card>
+        </motion.div>
+
         {/* Request Rate */}
-        {hasTraffik && (
+        {hasTraffic && (
           <motion.div variants={cardVariants}>
             <Card className="transition-shadow duration-300 hover:shadow-md hover:shadow-border/20">
               <CardHeader className="flex flex-row items-center justify-between">
@@ -249,15 +325,15 @@ export default function Dashboard() {
           </motion.div>
         )}
 
-        {/* Latency */}
-        {hasTraffik && (
+        {/* Load Average */}
+        {hasNode && (
           <motion.div variants={cardVariants}>
             <Card className="transition-shadow duration-300 hover:shadow-md hover:shadow-border/20">
               <CardHeader className="flex flex-row items-center justify-between">
-                <CardTitle>Latency</CardTitle>
+                <CardTitle>Load Average</CardTitle>
                 <Badge variant="secondary" className="text-[10px] font-normal">
                   {data?.latency?.length
-                    ? `${data.latency[data.latency.length - 1].toFixed(0)}ms`
+                    ? data.latency[data.latency.length - 1].toFixed(2)
                     : "-"}
                 </Badge>
               </CardHeader>
@@ -269,7 +345,7 @@ export default function Dashboard() {
         )}
 
         {/* Error Rate */}
-        {hasTraffik && (
+        {hasTraffic && (
           <motion.div variants={cardVariants}>
             <Card className="transition-shadow duration-300 hover:shadow-md hover:shadow-border/20">
               <CardHeader className="flex flex-row items-center justify-between">

src/lib/dashboard/types.ts

@@ -30,6 +30,14 @@ export interface DashboardData {
   errors: number[];
   traceVolume: number[];
   links: { url: string; label: string }[];
+  llm: {
+    reqRate: number | null;
+    tokRate: number | null;
+    tokSpeed: number | null;
+    uptime: number | null;
+    reqSpark: number[];
+    tokSpark: number[];
+  };
 }
 
 export function safeDur(us: number): string {
@@ -55,3 +63,14 @@ export function serviceIndicator(state: string): string {
       return "bg-red-400";
   }
 }
+
+export function fmtUptime(seconds: number): string {
+  const s = Math.max(0, Math.floor(seconds));
+  if (s < 60) return `${s}s`;
+  const m = Math.floor(s / 60);
+  if (m < 60) return `${m}m ${s % 60}s`;
+  const h = Math.floor(m / 60);
+  if (h < 24) return `${h}h ${m % 60}m`;
+  const d = Math.floor(h / 24);
+  return `${d}d ${h % 24}h`;
+}