If a pdf file is uploaded to research assistant, it successfully reads it and can answer first message sent to it. After sending a second message, it returns error "We are sorry, this input is not supported. (Invalid Input)".
Steps to reproduce:
Not get an error as a response
Same experience here. Here's another in case you need it for troubleshooting. It really kills the experience.
I got the same error in my second follow up message to a hyperlinked PDF. I think I’m gonna downgrade my ultimate plan because of this.
Tirion That seems like a knee-jerk reaction.
I'm also seing this bug while using expert assistant mode. Same pattern with image uploads using the GPT4-V API, after 2 follow-up questions the invalid input error appears.
Tirion Would not recommend using products clearly marked as "Beta" if you can't deal with a few bugs.
As far as I understand, Kagi’s assistant is the main additional feature of the ultimate plan.
Meanwhile, GPT Plus is marginally more expensive for an LLM that does search+RAG, but it generally works.
Not trying to crap all over Kagi btw, I’m still a fan. But if it can’t take more than 1 follow up question reliably, I think it’s fair to consider that it’s not for me at the moment.
It has worked in the past without issue. A bug was introduced and it looks like a fix is imminent. That’s all we’re saying.
