🐞 fix: 增加错误之后对已解析段落的缓存功能,再次重试时不再重头开始

解析长视频时,当附件大小过大时不再调用后进行报错,而是将附件进行分批次发送在每篇笔记开头默认增加地址来源链接,对模糊处可溯源
2026-06-03 06:41:54 +08:00 · 2026-02-12 18:28:11 +08:00
parent 7b45db2f59
commit d9a7b89e7d
67 changed files with 279293 additions and 64 deletions
--- a/backend/app/models/gpt_model.py
+++ b/backend/app/models/gpt_model.py
@@ -15,4 +15,5 @@ class GPTSource:
    extras: Optional[str] = None
    _format: Optional[list] = None
    video_img_urls:  Optional[list] = None
+    checkpoint_key: Optional[str] = None