end_nodes functionality is bugged (but not feature breaking) because edge processing is done post-node processing

so edge case (get it) is for parent node of end node, it will process all the edges on that layer of the BFS instead of just the one connecting parent to end

actually there’s a bunch of other cases that will probably be buggy since there are no checks on end nodes

e.g. all the end nodes could be upstream of the start nodes if graph is a DAG

could change implementation to backtracking instead. I think that should solve the problems

I think your link to mermaid live editor is borked - it seems to be linking to github diffs

Otherwise, very neat!
Could you add a placeholder/templating to children nodes, it's not clear that answer_question can accept context/question.

added parameter info for all nodes now

cool. beans.

Slack File: image.png

Hm. I was able to get the explicit function signatures when I ran it, not just args and kwargs

:sadge:

For reference

Slack File: image.png

are you able to repro what I got?

Not yet. It's something I saw earlier when I was pulling from the operator insurance and not the class

if this isn't quite resolved we can push back our meeting @abjjabjj