Now when the token count is too big and I split it, it needs to make multiple LLM calls. Need to store the ones already ran so I don't need to run them again.
Big inputs seem to face a lot of variance, with multiple times not generating a valid object. Need to keep already processed bits of transactions stores and call more retries.
Now when the token count is too big and I split it, it needs to make multiple LLM calls. Need to store the ones already ran so I don't need to run them again.