feat: add task lifecycle management and improve generation reliability

Centralize timeout policies, stall detection, and error classification
for image/video/text generation tasks. Improve ecommerce OSS upload flow
and add script evaluation enhancements.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

src/api/aiGenerationClient.ts

@@ -134,6 +134,12 @@ export interface ChatInput {
   temperature?: number;
 }
 
+export interface ChatUsage {
+  promptTokens?: number;
+  completionTokens?: number;
+  totalTokens?: number;
+}
+
 export interface AiTaskStatus {
   taskId: string;
   projectId?: string;
@@ -500,6 +506,7 @@ export const aiGenerationClient = {
     input: ChatInput,
     onChunk: (text: string) => void,
     signal?: AbortSignal,
+    onUsage?: (usage: ChatUsage) => void,
   ): Promise<void> {
     const res = await fetch(buildApiUrl("ai/chat"), {
       method: "POST",
@@ -529,8 +536,24 @@ export const aiGenerationClient = {
         const payload = line.slice(6).trim();
         if (!payload) continue;
         try {
-          const chunk = JSON.parse(payload) as { delta?: string; done?: boolean; error?: string };
+          const chunk = JSON.parse(payload) as {
+            delta?: string;
+            done?: boolean;
+            error?: string;
+            usage?: ChatUsage & {
+              prompt_tokens?: number;
+              completion_tokens?: number;
+              total_tokens?: number;
+            };
+          };
           if (chunk.error) throw new Error(chunk.error);
+          if (chunk.usage) {
+            onUsage?.({
+              promptTokens: chunk.usage.promptTokens ?? chunk.usage.prompt_tokens,
+              completionTokens: chunk.usage.completionTokens ?? chunk.usage.completion_tokens,
+              totalTokens: chunk.usage.totalTokens ?? chunk.usage.total_tokens,
+            });
+          }
           if (chunk.delta) onChunk(chunk.delta);
           if (chunk.done) return;
         } catch (e) {