Question on Memory Plugin Access Control and DLS/FLS Integration

Plugins Machine Learning
1

Versions (relevant - OpenSearch/Dashboard/Server OS/Browser):

Describe the issue:

Configuration:

Relevant Logs or Screenshots:

While reviewing the OpenSearch Memory plugin code, I noticed that access control seems limited to the conversation creator: an OpenSearch security user can only view memories from conversations they initiated. Even if a DLS rule granting broader access exists on the memory index, the plugin appears to enforce creator-only visibility, overriding DLS.

Is my understanding correct? If so, what is the reasoning behind not leveraging OpenSearch’s built-in DLS/FLS capabilities for memory access control? Are there plans to align this with the existing security model?

Thanks!

2

Thanks for digging into this and for the detailed explanation.

The current Memory plugin is quite limited in how it handles access control today. In parallel, we’ve been working on a newer Agentic Memory capability in the ML Commons plugin, which is designed with a more flexible model in mind.

You can find the docs here: https://docs.opensearch.org/latest/ml-commons-plugin/agentic-memory/

It might be worth taking a look to see if it better matches your use case, and we’d really appreciate any feedback you have on it.