feat: add spam detection lane (tower.classify.spam) with content-aware filtering

- Add SpamJobData + sync missing job types (DncJobData, ClassifyJobData, P1JobData, etc.) from dist back to source in @tower/types
- spam-detector.ts: emoji-only, short-greeting, link-only detection + duplicate-hash (1h window)
- spam.queue.ts / spam.processor.ts: BullMQ tower.classify.spam lane
- classify.processor.ts: spam check (Step 0) before rule engine, accepts spamQueue param
- main.ts: wire spamQueue + spamWorker; spam pre-filter before ingestQueue.add
- dnc.processor.ts: labelled reason logging

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

apps/worker/src/spam/spam-detector.ts

@@ -0,0 +1,77 @@
+import { createHash } from 'crypto';
+
+export type SpamReason = 'emoji_only' | 'short_greeting' | 'link_only' | 'forward_flood' | 'duplicate_hash';
+
+export interface SpamResult {
+  isSpam: boolean;
+  reason?: SpamReason;
+  contentHash: string;
+}
+
+const GREETING_PHRASES = [
+  'good morning', 'good evening', 'good night', 'good afternoon',
+  'gm', 'gn', 'subah ki shayari', 'शुभ प्रभात', 'शुभ रात्रि',
+  'happy', 'jai hind', 'jai shree ram', 'jai mata di',
+  'forwarded as received', 'please share', 'must share',
+];
+
+const EMOJI_RE = /^[\p{Emoji}\p{Emoji_Presentation}\s]+$/u;
+const URL_RE = /https?:\/\/\S+/gi;
+const TOKEN_MIN = 4;
+
+export function hashContent(content: string): string {
+  return createHash('sha256').update(content.trim().toLowerCase()).digest('hex').slice(0, 16);
+}
+
+export function detectSpam(content: string): SpamResult {
+  const trimmed = content.trim();
+  const contentHash = hashContent(trimmed);
+
+  if (EMOJI_RE.test(trimmed)) {
+    return { isSpam: true, reason: 'emoji_only', contentHash };
+  }
+
+  const lower = trimmed.toLowerCase();
+  const withoutUrls = lower.replace(URL_RE, '').trim();
+  const tokens = withoutUrls.split(/\s+/).filter(Boolean);
+
+  if (tokens.length < TOKEN_MIN) {
+    if (GREETING_PHRASES.some((phrase) => lower.includes(phrase))) {
+      return { isSpam: true, reason: 'short_greeting', contentHash };
+    }
+  }
+
+  const urls = trimmed.match(URL_RE) ?? [];
+  if (urls.length > 0 && tokens.length === 0) {
+    return { isSpam: true, reason: 'link_only', contentHash };
+  }
+
+  return { isSpam: false, contentHash };
+}
+
+export async function checkDuplicateHash(
+  contentHash: string,
+  tenantId: string,
+  sourceGroupId: string,
+  windowMs: number,
+  prisma: any,
+): Promise<boolean> {
+  const since = new Date(Date.now() - windowMs);
+  const count = await prisma.message.count({
+    where: {
+      tenantId,
+      sourceGroupId,
+      content: { startsWith: '' },
+      createdAt: { gte: since },
+    },
+  });
+  if (count === 0) return false;
+
+  const recent = await prisma.message.findMany({
+    where: { tenantId, sourceGroupId, createdAt: { gte: since } },
+    select: { content: true },
+    take: 100,
+  });
+
+  return recent.some((m: { content: string }) => hashContent(m.content) === contentHash);
+}