[Discussion]: 本地上传 pdf 文档时（非知识库，而是通过添加附件的方式）模型无法识别

[laocuhongshaorou]

Issue Checklist

• I understand that issues are meant for feedback and problem-solving, not for venting, and I will provide as much detail as possible to help resolve the issue.
• I confirm that I am here to ask questions and discuss issues, not to report bugs or request features.

Platform

Windows

Version

0.9.27

Your Question

在与模型对话时，类似 R1 和 豆包的 vision pro 这样的模型，都无法正确识别 pdf 文件，所以我在想是不是应该在上传 pdf 这一步，增加一个向量化的过程，这样模型就能识别了？

Context

No response

Additional Information

Priority

Medium (Would like a response soon)

[FinTechCao]

看起来是触发敏感词直接停了。我今晚刚用腾讯云的ds读了3篇PDF论文了，没问题

[henrycoding]

我是用deepseek-r1:70b模型不能用，读取不了pdf

[kangfenmao]

文件的问题全部使用大参数模型，不要用 100b 以下的小模型