We want to share an important update regarding K-AI (the KNIME AI assistant) in KNIME Analytics Platform.
Starting today, KNIME Analytics Platform users will have access to 20 K-AI requests per month.
If you need more than that, you can upgrade to KNIME Pro or KNIME Team, which includes 500 requests per user per month in addition to other productivity and collaboration features.
Why are we making this change?
Running large language models comes with significant infrastructure costs. To ensure K-AI remains sustainable in the long run, we’ve introduced a fair usage policy that keeps K-AI accessible to everyone, while ensuring that those who rely on it more heavily have the right options available.
For most users, this update won’t affect day-to-day work. Based on our analytics, the vast majority of users fall well below 20 requests per month.
We know KNIME is valued for its openness, and we’ve designed this limit carefully to balance sustainability with accessibility. We also recognize that this is a change from unlimited usage, and we’ll continue listening to your feedback as we evolve these offerings.
Thanks for your understanding and for being part of the KNIME community!
Is “Month” a calendar month or a running 30 days? Also, does “request” mean prompts or separate sessions. Very often I have to fine tune a “request” which results in multiple prompts. I’m not being critical of your decision. I think its very reasonable. Just want to understand the details.
A fair approach indeed!
Would there be an option to disable it (if needed) for the members of a KNIME hub community team’s subscription in case a organization decide not to have it.
Is there an indicator how much requests were “consumed”?
Yep, In the latest AP we show it to the user under the KAI chatbox. Full transparency is very important to us, also for other (future) AI-based features.
How is a request defined (conversation or literally each interaction / submit)?
At this point a simple question in the chat send to our AI Service.
Does a resource exists that transparently describes the advantages of K-AI (curated & consistently trained) vs. standard GenAI chatbot?
Like from a marketing perspective? I think you will get so far with a standard chat bot for Q&A, but K-AI will deliver more value to the end-user beyond simple Q&A (which is obviously very KNIME customized, too). For example we’re planning a “Auto-document” or “auto-annotate” feature, as well as a much improve build mode of course.
K-AI is not “bound” to the release schedule which is why it wasn’t part of the release notes I assume?
Thanks Christian for takign the time to reply About this:
Yes, I meant it exactly as you interpreted it. That might also be a nice incentive to upgrate to the Pro or even Team plan in case some credits could be added as a bonus. Similar to comitting to an upfront payment in AWS CloudWatch where one gets the perk of a few WAF credits.
Yes - we’ll make it possible for Basic and Standard edition customers to setup their own AI backend for K-AI - this enables them to provide unlimited K-AI access to the users of B-Hub as well - no need for a pro subscription for B-Hub users. So far, this was only possible with Enterprise edition of B-Hub. We’re targeting 1.17 B-Hub which is scheduled for End of Nov / Early Dec for this change.
Expressions and E charts nodes use KAi as backend .Very genesis and existence of such robust nodes like Echarts is the KAI .. as single prompt is not sufficient to get the desired result, it takes many iteration to reach the fine tuned result.. limitation of 20 request will limit the usage of such a robust nodes and inturn the outputs switching to basic nodes . Assuming the Pro account.. the limit of 500 too may not be sufficient ( or may be… numbers can tell ) .. considering iterations needed to get the finer results..
@AnilKS thanks for the feedback. We did some digging into our data and 20 interactions seemed to cover ~80% of all user’s use-cases so far - of course if you need more, you can then get KNIME Pro (as described above).
We selected 500 interactions in KNIME Pro because we have to set a limit and thought that’s a high enough limit that the likelihood that someone runs into the limit is extremely low. If we observe many users having meaningful use-cases with 500+ interactions, we can revisit the decision for sure, ofc having to respect our own cost somehow.
The Knime Pro is definitely as answer @christian.birkhold and the current move makes a judicious purpose to move forward.. .. I agree that unlimited usage puts strain on system and resources.. while if we narrow down the sample size to users who use expressions and Echarts ( being recently introduced with KAI functionality ) then may be outcome can be bit different. It may limit the beginners / user who are in dilemma to switch to use only Non KAi enabled Nodes.. while that is good enough trade off .. if it has been thought through .
