无法识图 #893
CastleApollo
started this conversation in
General
无法识图
#893
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
纯文档模型无法识图,但是我想自行添加MCP在模型之前先进行识图,而后可将结果发送模型再做处理,并且制定规则:要求存在图片的情况下,一定要先识图,再模型调用,但是结果是被DSH的硬性门控给拦截,此前这一套方案已在ZCODE上实践过,并且其他的agent例如qoder也都是先agent本身进行处理再去调用模型。

All reactions