feat(note): 添加视频理解功能- 在 GPT 模型中增加 video_img_urls 字段用于存储视频截图

- 在笔记生成请求中添加视频理解相关参数
- 实现视频截图功能，支持按指定间隔生成截图
- 更新笔记生成逻辑，支持视频理解功能- 在前端服务中添加视频理解相关参数

backend/app/gpt/universal_gpt.py

@@ -7,6 +7,7 @@ from app.models.transcriber_model import TranscriptSegment
 from datetime import timedelta
 from typing import List
 
+
 class UniversalGPT(GPT):
     def __init__(self, client, model: str, temperature: float = 0.7):
         self.client = client
@@ -28,21 +29,41 @@ class UniversalGPT(GPT):
     def ensure_segments_type(self, segments) -> List[TranscriptSegment]:
         return [TranscriptSegment(**seg) if isinstance(seg, dict) else seg for seg in segments]
 
-    def create_messages(self, segments: List[TranscriptSegment],**kwargs):
-        print("UniversalGPT",kwargs)
-        content =generate_base_prompt(
+    def create_messages(self, segments: List[TranscriptSegment], **kwargs):
+
+        content_text = generate_base_prompt(
             title=kwargs.get('title'),
             segment_text=self._build_segment_text(segments),
             tags=kwargs.get('tags'),
             _format=kwargs.get('_format'),
             style=kwargs.get('style'),
-            extras=kwargs.get('extras')
+            extras=kwargs.get('extras'),
         )
 
-        return [{"role": "user", "content": content }]
+        # ⛳ 组装 content 数组，支持 text + image_url 混合
+        content = [{"type": "text", "text": content_text}]
+        video_img_urls = kwargs.get('video_img_urls', [])
+
+        for url in video_img_urls:
+            content.append({
+                "type": "image_url",
+                "image_url": {
+                    "url": url,
+                    "detail": "auto"
+                }
+            })
+
+        # ✅ 正确格式：整体包在一个 message 里，role + content array
+        messages = [{
+            "role": "user",
+            "content": content
+        }]
+
+        return messages
 
     def list_models(self):
         return self.client.models.list()
+
     def summarize(self, source: GPTSource) -> str:
         self.screenshot = source.screenshot
         self.link = source.link
@@ -51,8 +72,8 @@ class UniversalGPT(GPT):
         messages = self.create_messages(
             source.segment,
             title=source.title,
-            tags=source.tags
-            ,
+            tags=source.tags,
+            video_img_urls=source.video_img_urls,
             _format=source._format,
             style=source.style,
             extras=source.extras
@@ -63,7 +84,3 @@ class UniversalGPT(GPT):
             temperature=0.7
         )
         return response.choices[0].message.content.strip()
-
-if __name__ == '__main__':
-    print('s')
-