feat(dashboard): LLM Inference card with live llm-api metrics

- /api/dashboard now queries llm-api Prometheus metrics: req/s, tok/s, gen tok/s, uptime, sparklines
- new LLM Inference card shows gen throughput + uptime badge + traffic sparkline
- fmtUptime helper (human-readable duration)

src/app/api/dashboard/route.ts

@@ -137,6 +137,14 @@ 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() {
@@ -176,6 +184,33 @@ export async function GET() {
     promQuery("count(up)"),
   ]);
 
+  // ── 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(
+      `sum(rate(llm_api_requests_total{service="llm-api"}[1m]))`,
+    ),
+    promRange(
+      `sum(rate(llm_api_completion_tokens_total{service="llm-api"}[1m]))`,
+    ),
+  ]);
+
   const links: { url: string; label: string }[] = [
     { url: "https://github.com/asepharyana/asepharyana-hub", label: "GitHub" },
   ];
@@ -195,6 +230,14 @@ export async function GET() {
     errors,
     traceVolume: traceVolume ? [traceVolume] : [],
     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,
@@ -230,6 +231,73 @@ 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 && (
           <motion.div variants={cardVariants}>

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`;
+}