Add SelfMemory memory provider plugin

[shrijayan]

Is this a new feature, an improvement, or a change to existing functionality?

New Feature

How would you describe the priority of this feature request

Low (would be nice)

Please provide a clear description of problem this feature solves

NeMo Agent Toolkit currently supports limited memory backends (Mem0, Redis, Zep). Users who want to use other vector stores (Qdrant, Chroma, Pinecone, Weaviate, etc.) or embedding providers (OpenAI, Ollama, HuggingFace, etc.) need to write their own integration from scratch. SelfMemory provides a unified interface to 29+ vector stores and 15+ embedding providers, reducing this effort to a simple config change.

Describe your ideal solution

A new nvidia-nat-selfmemory package that implements the MemoryEditor interface using SelfMemory as the backend. Users should be able to configure it in YAML like any other memory provider:

memory:
  my_store:
    _type: selfmemory
    vector_store_provider: qdrant
    embedding_provider: openai

The plugin should follow the same patterns as the existing nvidia-nat-mem0ai package — config class extending MemoryBaseConfig, async MemoryEditor implementation, registration via @register_memory, and entry point discovery.

Additional context

SelfMemory is an open-source library: https://github.com/shrijayan/selfmemory. It also supports optional encryption via MASTER_ENCRYPTION_KEY and LLM-based memory extraction. The implementation follows the existing mem0 plugin structure under packages/nvidia_nat_selfmemory/.

Code of Conduct

• I agree to follow this project's Code of Conduct
• I have searched the open feature requests and have found no duplicates for this feature request

[willkill07]

Thank you for opening the issue and putting together the PR — we really appreciate you noticing NeMo Agent Toolkit and identifying a feature gap.

One question from our side: do you have any usage or adoption data for SelfMemory? Each additional plugin or package included in the NeMo Agent Toolkit comes with some ongoing maintenance overhead, so we try to be thoughtful about what we bring directly into the toolkit.

That said, a NeMo Agent Toolkit plugin does not necessarily need to live inside the main repository, since it can also be distributed as its own package.

We’re currently improving our documentation and the visibility of external/community plugins, and we’d be very happy to work with you to make sure any plugin created is:

• easy to install (should always be just pip install)
• well documented within the NeMo Agent Toolkit repository and documentation site (clear table/list of community plugins)

We’d love to collaborate on the best path forward.