So I have some ideas for implementing Coda AI by integrating it with a web scraper I have so I can do a lot of text analysis and text organization. The application I have in mind would be making a lot of calls to GPT using Coda’s AI features. Right now, there is no mention of limits to the number of calls that we can make to GPT using AI features, however I find it hard to believe that Coda is just giving unlimited calls to GPT for free (given that it is probably not free for them). So before I spend a whole bunch of time building out something that might be heavily throttled, I would like to get a sense of what the limits are so that I can know if my application would go over or under those limits.
So, does anyone know at what point Coda AI will stop you from sending AI calls?
Hi @Logipedia_Website , excited to hear of your creative ideas for using Coda AI! During this Beta period, we wanted to give makers the opportunity to test how and where they can leverage Coda AI to improve their workflows, for free. We’ve been exploring what limits make sense, and so far only a few testers have hit these-- and they reset daily.
Since this is something we’re continuing to explore, we’d love to hear any further feedback around your use case, and when you might experience the limits. Feel free to message me directly in the community or at katy@coda.io.
Maybe it’s too early, but @Katy_Turner could you talk more about how one can optimize AI calling and when exactly AI is used.
If I have a table with 5k rows, then I create an AI column with auto refresh turned on, I understand it’ll make 5k calls to fill it. How frequent will it refresh? Just when referenced information changes? Does it get recalculated eventually with no trigger?
What if I want AI to be run on some of these rows based on some criteria? Like I want to summarize only the info from cells on rows that match this criteria.
Hey, big thanks to @Logipedia_Website for bringing up this topic! It’s funny because just yesterday I was pondering the exact same thing. I’ve been pushing the limits with my AI experiments, making a whopping 2,000 to 5,000+ API calls, and surprisingly, I never faced any rate limitations. I was under the impression that this was the acceptable range. However, this morning, I hit a roadblock when I got rate limited after making just something around 1,000 AI calls. So now I think the daily limit is somewhere around 1,000 AI calls.
