I would like to contribute to py-why

[palguna26]

I built a causal memory layer for ai agents its called causal-os, i aim to make it into a toolkit used to build causally aware ai agents, and would like py-why's help for it.

[kunwuz]

Thanks for your interest! Could you please elaborate a bit more on your idea?

[palguna26]

Hey, thanks for replying.

I’ve been thinking a lot about how agent memory today is mostly just retrieval based. Agents store logs, summaries, traces, and then pull back similar context later.

What I want to explore is something more causal.

Instead of an agent only remembering:
“this action happened and the task failed”

it should understand:
“this action caused this state change, which eventually led to the failure.”

The goal is to experiment with memory systems where agents can connect actions, outcomes, failures, and environmental changes through cause-and-effect relationships rather than just semantic similarity.

I think this could help agents reason about why something failed, what decisions led there, and how to avoid repeating the same mistake again.

I’d love to contribute to PyWhy around this area, especially wherever causal discovery, SCMs, or reasoning over agent traces could fit. I’m still exploring the research direction myself, so I’d also appreciate guidance on where this might align best within the ecosystem.