fix(moderation): immediately abort retries on 429 Too Many Requests

- In llmModerationClient.ts (inner retry), if OpenAI throws a 429 (or 401/403), throw p-retry's AbortError to immediately exit the 3-attempt inner retry loop.
- In aiAnalyzer.ts (outer retry), propagate the AbortError from runModerationAnalysis so the 2-attempt outer retry loop also aborts immediately.
- This ensures that a burst of 20 concurrent tasks hitting rate limits immediately returns the messages to the DB queue (as 'analysis_incomplete') and rapidly increments the individual circuit breaker, pausing processing and preventing a thundering herd instead of making 12 API calls per stuck message.

src/moderation/aiAnalyzer.ts

@@ -1,5 +1,6 @@
 import { existsSync } from "node:fs";
 import { fileURLToPath } from "node:url";
+import { AbortError } from "p-retry";
 import { Piscina } from "piscina";
 import { config } from "../config.js";
 import { createChildLogger } from "../logger.js";
@@ -227,6 +228,7 @@ async function processIndividualFallback(
 
     const analysisResult = await retryWithBackoff(
       async () => {
+        try {
           const result = await runModerationAnalysis({
             targets: [message],
             contextText: contextLines.join("\n"),
@@ -248,7 +250,22 @@ async function processIndividualFallback(
 
           // Got a real result — clear the incomplete flag.
           exhaustedOnIncomplete = false;
+
           return result;
+        } catch (err: any) {
+          // Propagate AbortError so outer retry is immediately cancelled on 429.
+          if (err instanceof AbortError) {
+            throw err;
+          }
+          if (
+            err?.status === 429 ||
+            err?.status === 401 ||
+            err?.status === 403
+          ) {
+            throw new AbortError(err);
+          }
+          throw err;
+        }
       },
       {
         retries: 2,

src/moderation/llmModerationClient.ts

@@ -1,4 +1,5 @@
 import OpenAI from "openai";
+import { AbortError } from "p-retry";
 import { z } from "zod";
 import { config } from "../config.js";
 import { createChildLogger } from "../logger.js";
@@ -34,7 +35,10 @@ const openai = new OpenAI({
 
     // Override headers to bypass Cloudflare WAF Bot Fight Mode
     const headers = new Headers(init?.headers);
-    headers.set("User-Agent", "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36");
+    headers.set(
+      "User-Agent",
+      "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36",
+    );
     for (const key of Array.from(headers.keys())) {
       if (key.toLowerCase().startsWith("x-stainless")) {
         headers.delete(key);
@@ -610,6 +614,7 @@ CRITICAL: "message_id" HARUS berupa STRING (dibungkus tanda kutip ganda). Jangan
   try {
     const analysis = await retryWithBackoff(
       async () => {
+        try {
           const completion = await openai.chat.completions.create({
             model: config.AI_LLM_MODEL,
             messages: [
@@ -666,6 +671,18 @@ CRITICAL: "message_id" HARUS berupa STRING (dibungkus tanda kutip ganda). Jangan
             );
             throw parseError;
           }
+        } catch (apiError: any) {
+          // Immediately abort retries on rate limits or auth errors so the
+          // message can return to the DB queue instead of bursting retries.
+          if (
+            apiError?.status === 429 ||
+            apiError?.status === 401 ||
+            apiError?.status === 403
+          ) {
+            throw new AbortError(apiError);
+          }
+          throw apiError;
+        }
       },
       {
         retries: 3,