z-zeechung

第二点有误，千帆平台不是上限8000token，而是上限8000个字符。这会导致某些插件的request过长而出错 希望在generate_from_baidu_qianfan函数中添加校验，在request过长时压缩或截断

nice work. i'm gonna impl a multi-llm interface for langchain

let's make some infrastructures like llm api, knowledge base, browser, shell env and so on. it harms nothing, at least

autogpt has a knowledge base, so i guess we'll be the same

count me in as well, thanks