why is this not turning info from the knowledge ba...
# 🤝help
a
it is supposed to be answering questions of the user based on the info it digested from the knowledge base (PDF file ) that i have added!
a
Are you asking a question whose answer is available in the knowledge base?
a
yes
and it answered it correctly at first
when i refreshed and asked it again it just captures the question to the variable (question) and doesn't respond, the knowledge base is enabled
a
only the node "Standard 2" has knowledge base enabled on it. If you ask a question in "Standard 1", no knowledge search will be triggered. Same for anywhere else in your flow. If you have "Answer on start node" enabled on your knowledge agent, it will answer first and then dive into the flow.
a
Alright thanks for answering
is it a good idea tho to enable it from the start?
also is it true that we only get 5000 messages sent? per month ? would that increase later ?
a
https://botpress.com/pricing has all the details on message quotas. Messages sent & received in the emulator don't count
Enabling/disabling answer on start node depends on your bot and its use case. Which makes for a better conversation?
c
How would I know it's standard 2 , as we do change label of the node . I tried to upload pdf / URL /.text, nothing seems to work. I passed my website and few other, it did not give the answer I passed McKinney website and it worked and started answering. Any specific reason?
And now I did use standard 2 node keeping the same name . It worked a little more but gave wrong answer. I mean content was pulled but not correctly.
I tried again... It's actually hallucinating and not reading the Pdf I have put in the knowledge base
a
Hey Vivek, the node doesn't actually have to be called "Standard 2," it just has to have knowledge answering enabled. In the screenshot you shared, the node with knowledge answering enabled was called "Standard 2," so I used that label to make it clearer what I was talking about.
How are you adding websites? If they are in separate knowledge bases, there's a known bug that affects that. But if they're in a single KB source like the screenshot below, it should work better.
also text-based PDFs work a lot better than image based ones, so make sure your PDF has highlightable text and isn't just an image of text.
c
Hi Gordy, thanks for comming back ! I am trying to use the website in smalltalk and also in quickstart knowledge but it doesnt seem to be taking from anywhere
@acceptable-kangaroo-64719 I did try making the whole website content into the word doc and it worked fairly but that is not something i want I want a simple bot to start with where the bot takes the information just from the website or websites. Also, please help me with PDFs. Usually all the PDFs include images and text and the text is usually stylized. I am working with academic books which might have mathematical equations and chemical formulas. How should I be working on that ?
a
You should try to include as much text-based information as you can. Not sure how well our OCR does with LaTeX or other math formula formatting. There might be a third party tool that could turn it into more OCR friendly text? I'm not sure, you'll need to experiment and let us know how it goes 😁
f
Help please..
Enable Knowledge Answering not showing in the menu ??
Screenshot 2024-03-06 at 9.09.30 AM.png
105 Views