I'm facing a challenge where I'm asking the AI (Navigator, Argon, Neon and Open AI models) to parse through 10 different websites and on each website conduct 15 actions to lookup data.
With all models, it seems to always stop short after doing one or two websites with only one or two of the requested actions. The instructions also explicitly ask it to go through each and every website instead of steps. When looking at steps taken, it runs through the steps correctly for the first two websites and few lookups, but quits out as if the task is complete when it hasnt completed all the steps.
Any suggestions on how to debug this further?