Outdated AI models/tokens + poor bulk automation handling

First issue - outdated AI models and token limits

For a long time now many more OpenAI models have been available for selection, with much higher token limits than currently allowed in Fibery. Since users are using their own OpenAI key, there is no reason to not let people choose these new models that have much much higher token limits. Currently Fibery caps at 4097 total tokens. There are now models with over 100k tokens even, but 16k tokens atleast has been available for a very long time now.

Second issue - bad bulk automation handling

The scenario is that I tried to run a bulk automation using a button on for example 200 records. In this case, a few of those records were over the 4097 token limit for the currently available to choose models in Fibery. So, ALL records failed… The expected result should be 197 records are successful, 3 failed. This is how for example “GPT For Sheets” works with Google Sheets, which I did the whole job (2500 records!) in one go. There were also some errors there, but it handles it much better, it just fails on individual records and the rest continues. To make things even worse, when there’s an error, Fibery doesn’t even tell you which records had the error. But regardless… just allow the ones which are fine to be successful, and SKIP ones that have an issue. This is how other tools like this work.

Otherwise I love Fibery (although I do wish I could rename the default ‘Name’ field), and slowly transitioning my whole business over to it, so these suggestions are based on things I’m having to use other tools for instead of Fibery - I want to do it in Fibery so I can move everything there.

2 Likes

Hi, we will fix tokens issue ASAP.
As for automation rules, it should process all records that can be processed and skip records with errors, it should not work like ALL records failed. But we will try to reproduce this behavior, maybe the rules just turns-off after many fails.