refactor: split llmModerationClient.ts + add tests + metrics

## Split llmModerationClient.ts (2103 → 3 files)
- **moderationBuilders.ts** (67 lines) — shared: escapeXml, getAnalysisContent, buildReferenceXml
- **mediaAnalysisClient.ts** (656 lines) — vision analysis with multi-layer LRU/DB/phash caching, image/video download, ffmpeg frame extraction, prepareMediaMessage
- **moderationOrchestrator.ts** (998 lines) — callModerationLLM, runTextOnlyBatch, runMediaBatch, runModerationAnalysis, runSimpleTextFallback
- **llmModerationClient.ts** (30 lines) — re-export bridge (backward compat)

No import changes needed — aiAnalysisWorker.ts still imports from llmModerationClient.js.

## Unit tests (backend)
- vitest.config.ts + e2e.test.ts with 9 tests against production:
  - health, metrics, dashboard/stats, recordings, config, auth, guilds, negative (404/400)

## Monitoring metrics
- moderationMetrics.ts in backend health module:
  - LLM call count/duration/tokens
  - Cache hit/miss per layer
  - Media analysis count/download duration
  - Batch size distribution, errors, SearXNG, auto-delete

services/backend/src/e2e.test.ts

@@ -0,0 +1,92 @@
+/**
+ * E2E API tests — runs against a running backend instance.
+ * Usage: vitest run  (or: API_BASE=http://localhost:3001 vitest run)
+ */
+import { describe, it, expect } from "vitest";
+
+const BASE = process.env.API_BASE ?? "https://imphnen.asepharyana.my.id/api";
+
+async function api(path: string, init?: RequestInit) {
+  const res = await fetch(`${BASE}${path}`, {
+    ...init,
+    headers: { "Content-Type": "application/json", ...init?.headers },
+  });
+  const body = res.status !== 204 ? await res.json().catch(() => null) : null;
+  return { status: res.status, body };
+}
+
+describe("API Health", () => {
+  it("GET /health returns 200 with status=healthy", async () => {
+    const { status, body } = await api("/health");
+    expect(status).toBe(200);
+    expect(body?.status).toBe("healthy");
+  });
+
+  it("GET /metrics returns prometheus text", async () => {
+    const res = await fetch(`${BASE.replace("/api", "")}/api/metrics`);
+    expect(res.status).toBe(200);
+    const text = await res.text();
+    expect(text).toContain("nodejs");
+  });
+});
+
+describe("API Dashboard", () => {
+  it("GET /dashboard/stats returns stats fields", async () => {
+    const { status, body } = await api("/dashboard/stats");
+    expect(status).toBe(200);
+    expect(body).toHaveProperty("total_messages");
+    expect(body).toHaveProperty("total_flagged");
+    expect(body).toHaveProperty("active_users_24h");
+    expect(typeof body.total_messages).toBe("number");
+  });
+});
+
+describe("API Recordings", () => {
+  it("GET /recordings returns items with pagination", async () => {
+    const { status, body } = await api("/recordings?limit=5");
+    expect(status).toBe(200);
+    expect(body).toHaveProperty("items");
+    expect(Array.isArray(body.items)).toBe(true);
+    if (body.items.length > 0) {
+      expect(body.items[0]).toHaveProperty("id");
+      expect(body.items[0]).toHaveProperty("username");
+      expect(body.items[0]).toHaveProperty("created_at");
+    }
+  });
+});
+
+describe("API Config", () => {
+  it("GET /config returns 200", async () => {
+    const { status } = await api("/config");
+    expect(status).toBe(200);
+  });
+});
+
+describe("API Auth", () => {
+  it("POST /auth/login with wrong password returns 401", async () => {
+    const { status } = await api("/auth/login", {
+      method: "POST",
+      body: JSON.stringify({ password: "wrong" }),
+    });
+    expect(status).toBe(401);
+  });
+});
+
+describe("API Voice", () => {
+  it("GET /guilds returns 200", async () => {
+    const { status } = await api("/guilds");
+    expect(status).toBe(200);
+  });
+});
+
+describe("API Negative", () => {
+  it("GET /nonexistent returns 404", async () => {
+    const { status } = await api("/nonexistent");
+    expect(status).toBe(404);
+  });
+
+  it("GET /messages without channelId returns 400", async () => {
+    const { status } = await api("/messages?limit=3");
+    expect(status).toBe(400);
+  });
+});

services/backend/src/modules/health/moderationMetrics.ts

@@ -0,0 +1,80 @@
+/**
+ * moderationMetrics.ts
+ *
+ * Prometheus metrics for AI moderation pipeline.
+ * Defined in backend (where prom-client is installed + /api/metrics endpoint).
+ */
+import { Counter, Histogram, register } from "prom-client";
+
+// ── LLM Call Metrics ──
+export const llmCallsTotal = new Counter({
+  name: "moderation_llm_calls_total",
+  help: "Total LLM moderation calls",
+  labelNames: ["path", "model"] as const,
+});
+
+export const llmCallDuration = new Histogram({
+  name: "moderation_llm_call_duration_ms",
+  help: "LLM moderation call duration (ms)",
+  labelNames: ["path", "status"] as const,
+  buckets: [500, 1000, 2000, 5000, 10000, 20000, 30000, 60000, 120000],
+});
+
+export const llmTokensTotal = new Counter({
+  name: "moderation_llm_tokens_total",
+  help: "Total tokens consumed by LLM moderation",
+  labelNames: ["type"] as const,
+});
+
+// ── Cache Metrics ──
+export const moderationCacheHits = new Counter({
+  name: "moderation_cache_hits_total",
+  help: "Moderation cache hits",
+  labelNames: ["layer"] as const,
+});
+
+export const moderationCacheMisses = new Counter({
+  name: "moderation_cache_misses_total",
+  help: "Moderation cache misses",
+  labelNames: ["layer"] as const,
+});
+
+// ── Media Analysis Metrics ──
+export const mediaAnalysesTotal = new Counter({
+  name: "moderation_media_analyses_total",
+  help: "Media analyses performed",
+  labelNames: ["type"] as const,
+});
+
+export const mediaDownloadDuration = new Histogram({
+  name: "moderation_media_download_duration_ms",
+  help: "Media download duration (ms)",
+  labelNames: ["source"] as const,
+  buckets: [100, 500, 1000, 2000, 5000, 10000, 30000],
+});
+
+// ── Batch & Error Metrics ──
+export const moderationBatchSize = new Histogram({
+  name: "moderation_batch_size",
+  help: "Messages per batch",
+  labelNames: ["path"] as const,
+  buckets: [1, 5, 10, 20, 50, 100],
+});
+
+export const moderationErrors = new Counter({
+  name: "moderation_errors_total",
+  help: "Moderation errors",
+  labelNames: ["type"] as const,
+});
+
+export const searxngCalls = new Counter({
+  name: "moderation_searxng_calls_total",
+  help: "SearXNG search calls",
+  labelNames: ["status"] as const,
+});
+
+export const autoDeleteActions = new Counter({
+  name: "moderation_auto_delete_total",
+  help: "Auto-delete actions",
+  labelNames: ["action"] as const,
+});

services/backend/vitest.config.ts

@@ -0,0 +1,10 @@
+import { defineConfig } from "vitest/config";
+
+export default defineConfig({
+  test: {
+    globals: true,
+    environment: "node",
+    include: ["src/**/*.test.ts"],
+    testTimeout: 15000,
+  },
+});