Should Rovo Agents support persistent memory independent of the LLM context window?

[AGASTYA ANOOP SHARMA]

Current Rovo Agents primarily rely on prompts and retrieved context during execution.

As enterprise workflows become longer and more complex, should Rovo introduce a persistent memory architecture that stores verified organizational knowledge independently of each conversation?

How should Atlassian balance:

Context Window

Retrieval (RAG)

Long-term Memory

Organizational Knowledge Graph

without introducing hallucinations or stale memory?

[Rebekka Heilmann _viadee_]

That's what the instructions are for. You provide the right knowledge, add understanding about relations between it and give the Agent a task.

What else do you need?