feat(moderation): improve message analysis scheduling and update moderation prompt structure

src/moderation/aiAnalyzer.ts

@@ -91,6 +91,7 @@ async function processBatch(
   if (messages.length === 0) return;
 
   activeRequests++;
+  let shouldScheduleNext = false;
   const processingStartedAt = Date.now();
   conversationProcessing.set(conversationKey, processingStartedAt);
   try {
@@ -123,6 +124,7 @@ async function processBatch(
     }
 
     conversationErrorCooldown.delete(conversationKey);
+    shouldScheduleNext = true;
   } catch (error) {
     lastError = error instanceof Error ? error.message : String(error);
     const errorStack = error instanceof Error ? error.stack : undefined;
@@ -149,6 +151,9 @@ async function processBatch(
     if (conversationProcessing.get(conversationKey) === processingStartedAt) {
       conversationProcessing.delete(conversationKey);
     }
+    if (shouldScheduleNext) {
+      setImmediate(() => scheduleConversationAnalysis(conversationKey));
+    }
   }
 }
 

src/moderation/llmModerationClient.ts

@@ -390,7 +390,13 @@ export async function runModerationAnalysis(
     .map((msg) => `[${msg.id}] ${msg.username}: ${msg.content}`)
     .join("\n");
 
-  const systemPrompt = `You are a content moderation assistant. Analyze messages for policy violations.
+  const moderationPrompt = `You are a content moderation assistant. Analyze messages for policy violations.
+
+Context:
+${contextText}
+
+Messages to analyze:
+${messagesText}
 
 For each message, respond with a JSON object containing a "results" array.
 CRITICAL: You MUST return the "message_id" EXACTLY as provided in the input, and it MUST be wrapped in double quotes as a STRING. Do not treat IDs as numbers.
@@ -405,11 +411,6 @@ Each result must have:
 Do not include reasoning, analysis steps, markdown, prose, XML tags, or comments.
 Return ONLY valid JSON, no other text.`;
 
-  const userPrompt = `Context: ${contextText}
-
-Messages to analyze:
-${messagesText}`;
-
   // Check for image attachments to support multimodal analysis
   const targetIdSet = new Set(targets.map((t) => t.id));
   const getAttachmentImageUrl = (att: AttachmentRecord): string | null => {
@@ -485,11 +486,11 @@ ${messagesText}`;
       ...imageParts.flat(),
       {
         type: "text",
-        text: userPrompt,
+        text: moderationPrompt,
       },
     ];
   } else {
-    messageContent = userPrompt;
+    messageContent = moderationPrompt;
   }
 
   const result = await retryWithBackoff(
@@ -497,10 +498,6 @@ ${messagesText}`;
       openai.chat.completions.create({
         model: config.AI_LLM_MODEL,
         messages: [
-          {
-            role: "system",
-            content: systemPrompt,
-          },
           {
             role: "user",
             content: messageContent,
@@ -577,7 +574,7 @@ ${messagesText}`;
           model: config.AI_LLM_MODEL,
           timestamp: new Date().toISOString(),
         },
-        "Robust Fallback: Failed to parse moderation response. Defaulting all targets to clean.",
+        "Robust Fallback: Failed to parse moderation response. Marking all targets as analysis errors.",
       );
       parsed = targetIds.map((id) => ({
         messageId: id,

src/moderation/messageCapture.ts

@@ -172,6 +172,7 @@ export async function captureMessage(
   // Queue analysis after attachment uploads settle so AI uses stable tele URLs.
   if (!isBacklog) {
     if (attachmentUploadTasks.length > 0) {
+      setTimeout(() => queueMessageAnalysis(message.id), 30000);
       Promise.allSettled(attachmentUploadTasks)
         .then(() => queueMessageAnalysis(message.id))
         .catch((err) => {

tests/moderation/llmModerationClient.test.ts

@@ -459,11 +459,11 @@ describe("runModerationAnalysis", () => {
     });
 
     const requestBody = JSON.parse((global.fetch as any).mock.calls[0][1].body);
-    expect(requestBody.messages[0].role).toBe("system");
-    expect(requestBody.messages[1].role).toBe("user");
-    expect(typeof requestBody.messages[1].content).toBe("string");
-    expect(requestBody.messages[1].content).toContain("test context");
-    expect(requestBody.messages[1].content).not.toContain("data:image/png");
+    expect(requestBody.messages).toHaveLength(1);
+    expect(requestBody.messages[0].role).toBe("user");
+    expect(typeof requestBody.messages[0].content).toBe("string");
+    expect(requestBody.messages[0].content).toContain("test context");
+    expect(requestBody.messages[0].content).not.toContain("data:image/png");
   });
 
   it("throws on non-ok HTTP response", async () => {
@@ -626,8 +626,8 @@ describe("runModerationAnalysis", () => {
     const [, completionsOptions] = fetchCalls[1];
     const body = JSON.parse(completionsOptions.body);
 
-    expect(body.messages[0].role).toBe("system");
-    const userMessage = body.messages[1];
+    expect(body.messages).toHaveLength(1);
+    const userMessage = body.messages[0];
     expect(userMessage.role).toBe("user");
     expect(Array.isArray(userMessage.content)).toBe(true);
     expect(userMessage.content[0].type).toBe("image_url");
@@ -640,7 +640,7 @@ describe("runModerationAnalysis", () => {
     );
     expect(userMessage.content[2].type).toBe("text");
     expect(userMessage.content[2].text).toContain("test context");
-    expect(body.messages[0].content).toContain(
+    expect(userMessage.content[2].text).toContain(
       "You are a content moderation assistant.",
     );
   });
@@ -848,7 +848,7 @@ describe("runModerationAnalysis", () => {
     expect(fetchCalls[1][0]).toBe("https://httpbin.org/image/png");
 
     const requestBody = JSON.parse(fetchCalls[2][1].body);
-    const contentParts = requestBody.messages[1].content;
+    const contentParts = requestBody.messages[0].content;
     expect(
       contentParts.filter((part: any) => part.type === "image_url"),
     ).toHaveLength(2);
@@ -932,7 +932,7 @@ describe("runModerationAnalysis", () => {
     expect((global.fetch as any).mock.calls[0][0]).toContain(
       "/chat/completions",
     );
-    expect(typeof requestBody.messages[1].content).toBe("string");
+    expect(typeof requestBody.messages[0].content).toBe("string");
   });
 
   it("keeps analyzing text when an image URL returns non-OK", async () => {
@@ -1002,8 +1002,8 @@ describe("runModerationAnalysis", () => {
     expect(result.results[0].status).toBe("warn");
 
     const requestBody = JSON.parse((global.fetch as any).mock.calls[1][1].body);
-    expect(requestBody.messages[1].content).toHaveLength(1);
-    expect(requestBody.messages[1].content[0].text).toContain(
+    expect(requestBody.messages[0].content).toHaveLength(1);
+    expect(requestBody.messages[0].content[0].text).toContain(
       "https://example.invalid/claim",
     );
   });