- model list: any model with max_reference_images>0 shows 首尾帧 N (frame mode)
or 参考图 N (asset/other); drop the count-less "图生图" label
- ModelFormModal / TestModal: capability labels use 参考图 (not 图生图)
- grok-video uses reference_mode=asset (参考图), runway/adobe video stay frame (首尾帧)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Full Go backend + Vue 3 frontend, OpenAI-compatible API, multi-provider
account pools, billing/admin, Docker one-command deploy with auto HTTPS.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>