feat: 支持视频 API 多图参考输入

app/api/v1/chat.py

@@ -625,6 +625,13 @@ def validate_request(request: ChatCompletionRequest):
     # video 验证
     if model_info and model_info.is_video:
         config = request.video_config or VideoConfig()
+        _, image_urls = _extract_prompt_images(request.messages)
+        if len(image_urls) > 7:
+            raise ValidationException(
+                message="Too many image_url references for video. Maximum is 7.",
+                param="messages",
+                code="invalid_image_count",
+            )
         effective_stream = (
             request.stream if request.stream is not None else get_config("app.stream")
         )

app/services/grok/services/video.py

@@ -1092,6 +1092,17 @@ class VideoService:
         parent_post_id = (parent_post_id or "").strip() or None
         source_image_url = (source_image_url or "").strip()
         reference_items = [item for item in (reference_items or []) if isinstance(item, dict)]
+        if image_attachments and not reference_items:
+            reference_items = [
+                {
+                    "parent_post_id": "",
+                    "image_url": str(image_url or "").strip(),
+                    "source_image_url": str(image_url or "").strip(),
+                    "mention_alias": f"Image {index}",
+                }
+                for index, image_url in enumerate(image_attachments, start=1)
+                if str(image_url or "").strip()
+            ]
         used_tokens: set[str] = set()
 
         for attempt in range(max_token_retries):