Why did this workflow not include internal data?

[michaelraspuzzi]

this workflow is supposed to 1) read a notion db, and 2) pick the top choices from that notion database.

this is the first run (out of ~50 tests) where it completely hallucinated answers that were not in the notion database.

when looking at the run, i can’t actually see inputs and outputs at each step because they are blank.

can you help check to see what went wrong?

1. was it a bad prompt?
2. was there one place in the workflow that information didn’t get passed forward because of a limit? maybe context window limit?

here’s the run: https://www.gumloop.com/pipeline?run_id=EsfawUzySKQN3SfK6jFXiC&workbook_id=aN9pL3VU3iYkemY644BHS1

and here’s the first ask ai node, the output looks blank:

[Gumloop_Bot]

Hey @michaelraspuzzi! If you’re reporting an issue with a flow or an error in a run, please include the run link and make sure it’s shareable so we can take a look.

1. Find your run link on the history page. Format: https://www.gumloop.com/pipeline?run_id={{your_run_id}}&workbook_id={{workbook_id}}

2. Make it shareable by clicking “Share” → ‘Anyone with the link can view’ in the top-left corner of the flow screen.
   

3. Provide details about the issue—more context helps us troubleshoot faster.

You can find your run history here: https://www.gumloop.com/history

[Wasay-Gumloop]

Hey @michaelraspuzzi – Thank you for sharing. You’re not seeing the output in window there for Ask AI node since its too large to be displayed in the modal but it did produce an output within the workflow. I’m cusrious to understand how this information was a hallcuiantion since the users it mentioned in its response do in fact exist in the notion database, I checked both the outputs.

Was it just a poor quality output? In which case some prompt engineering and bumping to a smarter model should help.