Feature request: discuss optional ClawMem-backed long-term memory integration

[EricZequan]

Summary

I'd like to discuss whether Rasa would be open to an optional ClawMem-backed long-term memory integration for AI agents.

This is not a request to replace Rasa's existing architecture. The goal would be to explore whether ClawMem could live as an optional extension for teams that want durable, auditable memory across sessions.

Why this seems worth discussing

The current Rasa README is explicitly framed around building reliable AI agents, and even calls out memory as part of the broader agent-engineering design space.

That suggests a possible integration shape such as:

1. a custom tracker/memory-related backend
2. an optional extension package outside core
3. a documented integration example for users who want durable memory beyond default conversation state

What ClawMem would add

ClawMem is oriented around:

• long-term memory persistence
• durable records of memory changes
• repo/workspace-scoped memory boundaries
• optional shared memory across human/agent collaborators

Clarification I am looking for

Would maintainers be open to one of these directions?

1. external extension package
2. docs/example integration
3. first-party optional integration
4. not a good fit for Rasa

Minimal scope

The smallest reasonable scope seems to be:

• fully optional integration
• preserve current default Rasa behavior
• target a clearly bounded extension point rather than changing core dialogue behavior

If this is worth pursuing, I can turn this into a concrete implementation brief.