All four image tools (套图, 详情图, 模特图, 爆款图复刻) previously used
setTimeout + static sample images. Now they:
1. Upload product images to OSS via uploadAssetBinary
2. Build contextual prompts including platform/ratio/language/market + user text
3. Call aiGenerationClient.createImageTask with model=gpt-image-2
4. Poll task status via waitForTask (SSE + fallback polling)
5. Display the generated result URL in the preview grid
Key changes:
- Add ServerRequestError + waitForTask imports
- Add imageAbortRef for task cancellation
- Add uploadCloneImages() (generic version of uploadProductImages)
- Add buildEcommerceImagePrompt() with output-type-specific instructions
- Add generateEcommerceImage() orchestrating upload → prompt → API → result
- Replace all 5 mock handlers (handleGenerate, handleSetGenerate,
handleDetailGenerate, handleTryOnGenerate, handleGenerateModel) with
real async API calls
- Handle 402 (Payment Required) with user-friendly error message
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
stringadmin
b07ff439f3