Memory

User profiles, memories, and session context in the Mentor team.

This walkthrough follows Demo OS at the reviewed revision, which pins Agno 2.8.1. Use the application’s checked-in requirements.

Mentor shares a LearningMachine between its team leader, Mentor member, and Curator member. Its configuration enables user profile, user memory, session context, learned knowledge, and decision logging.

Three stores, three shapes

StoreWhat it retains
User profileInformation about the caller, such as role, goals, and working context
User memoryPreferences and facts reused in later conversations for that user
Session contextThe active conversation’s task and current context

User memory

Mentor enables UserMemoryConfig(mode=LearningMode.ALWAYS) and adds learnings to context. Use the same user identity when testing reuse across sessions. Sharing a database alone does not configure every agent to consume the same learning stores.

User profile

Tell Mentor your role and a preference, then ask a question where that information matters:

I lead a small platform team. Prefer concise rollout checklists.
Help me prepare a review of our next release.

Inspect the stored profile and the next answer. Capture and use of a fact depend on the configured learning process and model behavior.

Session context

Mentor enables SessionContextConfig(mode=LearningMode.ALWAYS) alongside conversation history. Continue the same conversation to test how it carries the active task forward. A new session has a different session context, while user-scoped profile and memory can still be reused.

Vanilla agentic memory vs LearningMachine

The older enable_agentic_memory option remains a different configuration path. This demo uses explicit LearningMachine stores, shared through the _common configuration in teams/coach/team.py. See Learning for its reusable lessons and decision log.

See it in action

Start Demo OS, select the Mentor team, and use a consistent user identity. Inspect memory/profile records after a conversation, then open a new session and ask for help that uses the same preferences.

Next

teams/coach/team.py

Run Demo OS · Current AgentOS documentation