0000sir

Not fixed after 2 years in v3.1.0.

How to fix this temporary before the team release the official patch? Thank you

> You can use the following methods to temporarily solve the problem. https://github.com/LinyuJupiter/FastGPT-zhipu-api If you want a proxy, lua+openresty will be better. But this proxy idea is not a good...

Found this will be fixed in #804 Wait for new image 🤣👍

果然是文如其人，丝毫没有受过九年义务教育的迫害

导入文件应该是用不到chatglm模型的，如果能将这部分拆分开（应该是在chains/local_doc_qa.py的初始化部分），就能在不影响前端服务的情况下进行数据批量导入

I have the same problem, this project was abandoned, it use ffmpeg for transcoding videos, but not run it directly, instead reference from github.com/riltech/streamer, that's much more difficult to debug....

I can sure direct upload works fine in the original rails forms.

That worth a shot

save different iter_value results to different directory should be a easier way.