[Bug]: With max_token disabled, output is still truncated

Open xyk0930 opened this issue 10 months ago • 5 comments

Describe your problem

If the ollama client is directly used, it is not truncated

@KevinHuSh How to solve this problem?

Mar 03 '25 09:03 xyk0930

A truncated chat

Mar 03 '25 09:03 xyk0930

I encountered the same problem

Mar 03 '25 15:03 surmount1

I have tested different versions and this is a new bug introduced in v0.17.0

Mar 04 '25 01:03 xyk0930

It retrieved a lot of content while answering. So, may be caused by context length. Apart from it, it may lose connection to Ollama because of slow computations.

Mar 04 '25 02:03 KevinHuSh

我也遇到了同样的问题，另外，你的这个思考过程为什么可以放到引用块里呢？请问是怎么做到的？可以分享一下吗？

截断的聊天

Mar 04 '25 02:03 bbj677