Skip to main content

​
What is Scenario Mode?

In EverMemOS Cloud, you choose a Scenario Mode when you create a Memory Space. The mode defines how EverMemOS extracts and consolidates memories based on the chat-session structure. Specifically, Scenario Mode tells EverMemOS:
  • how to interpret the chat-session structure (human–AI interaction vs multi-participant group chat)
  • the extraction strategy and granularity (single-speaker-centric vs multi-participant in one run)
  • how memory extraction and consolidation are applied in this space
Once a Memory Space contains data, you cannot change its Scenario Mode. This protects memory consistency and retrieval quality.
If you are unsure, start with a new space and validate the behavior with a small test conversation before storing important data.

​
Two Scenario Modes

​
Personal AI Assistant / Companion

This mode is designed for chat sessions where one human user talks with one or more AI assistants/companions over time.
  • Memory subject (per chat session): the human user
  • Best for: personal assistant/companion experiences, including cases where you talk to multiple AIs
  • Extraction focus: preferences, personal facts, and evolving states that help an agent serve you better

​
Key Features

  • Human-Centric Extraction: The system monitors the dialogue and extracts memories, traits, and preferences only for the human user.
  • AI Memory Exclusion: To maintain a clean and focused user profile, EverMemOS does not extract memories for the AI agent, but still keeps the atomic facts of responses from the AI agent.
  • Deep Personalization: By focusing all extraction resources on the human user, the system can capture more nuanced details about their personality, history, and evolving needs.

​
Team Collaboration

This mode is designed for multi-participant group chats. The group chat may include an AI participant, or it may be human-only. In a single extraction run, EverMemOS extracts and updates memories for every participant in the conversation.
  • Memory subject: the group and its participants
  • Best for: multi-participant group chats and team workflows
  • Extraction builds:
    • a group profile — recurring topics, decision makers, shared norms, and team conventions
    • participant profiles — each member’s role, action items, responsibilities, and relevant traits

​
Key Features

  • Granular memory extraction: EverMemOS tracks distinct situational memories for every participant individually — even when multiple topics are discussed in parallel within the same timeframe. Each contribution is attributed to the correct speaker, never merged into a single undifferentiated stream.
  • Scenario-specific profiles: Profile fields are structured for professional contexts. Instead of free-form AI-generated text, profiles capture structured data such as roles, action items, and ownership. The group profile surfaces who drives decisions, what topics recur, and how the team operates.
  • Separated storage: Memories are stored as individual episodic records associated with each user’s unique user_id. This prevents profile contamination where one participant’s context is mistakenly attributed to another.
This scenario is ideal for Discord moderators, team assistants, or any agent operating in a social or professional group setting.
1

Decide what each chat session looks like

If each session is one human user talking with one or more AIs, choose Personal AI Assistant / Companion.If each session is a work-oriented group chat with multiple participants collaborating on professional tasks, choose Team Collaboration.
2

Create a new space with the chosen mode

Set the Scenario Mode at creation time.
Confirm the mode in your space settings before you start storing important memories.
3

Validate with a small test conversation

Add a small, low-stakes conversation and verify retrieval matches your expectation (personal profile vs group + participants).