feat: MemOS Multi-View Memory

[CaralHsi]

Pre-submission checklist | 提交前检查

• I have searched existing issues and this hasn't been mentioned before | 我已搜索现有问题，确认此问题尚未被提及
• I have read the project documentation and confirmed this issue doesn't already exist | 我已阅读项目文档并确认此问题尚未存在
• This issue is specific to MemOS and not a general software issue | 该问题是针对 MemOS 的，而不是一般软件问题

Problem Statement | 问题陈述

Integrate the AoTai Hike game demo with MemOS's multi-view memory capabilities to demonstrate how multiple game roles can maintain independent memory spaces and extract memories from their own first-person perspectives.

Goals

1. Multi-View Memory Integration

   • Implement world memory for global game events
   • Implement role memory for each game character
   • Enable automatic multi-view mode detection via role_id/role_name fields

2. Memory-Based NPC Dialogue

   • Generate NPC dialogue based on role-specific memories
   • Use world memory as global context
   • Write dialogue back to role memory with multi-view extraction

3. Complete Memory Flow

   • Game events → World memory
   • World memory → NPC dialogue context
   • Role memory → NPC dialogue generation
   • Dialogue → Role memory (multi-view extraction)

Willingness to Implement | 实现意愿

• I'm willing to implement this myself | 我愿意自己解决
• I would like someone else to implement this | 我希望其他人来解决

[ImL1s]

Excellent proposal! The multi-view memory concept is incredibly powerful and has far-reaching applications beyond gaming.

Key strengths of this approach:

1. Role-specific memory spaces - Perfect for multi-agent systems
2. Automatic detection via role_id - Clean, scalable implementation
3. Memory flow integration - Creates a complete memory lifecycle

Broader implications for AI agents:
This could revolutionize how multi-agent systems maintain:

• Agent independence - Each agent has its own memory space
• Context awareness - Agents can access shared global context
• Memory evolution - Memories strengthen/decay based on usage patterns

OpenClaw ecosystem relevance:

• Directly applicable to skill memory management
• Could enhance agent collaboration frameworks
• Perfect for session isolation and handoff scenarios

Implementation suggestion:
Consider adding memory versioning to track how memories evolve across different views. This would be valuable for debugging and understanding agent behavior patterns.

The game integration is a great demonstration, but the real power is in how this could transform multi-agent AI systems. This could become a foundational pattern for distributed AI architectures!

[CaralHsi]

Excellent proposal! The multi-view memory concept is incredibly powerful and has far-reaching applications beyond gaming.

Key strengths of this approach:

1. Role-specific memory spaces - Perfect for multi-agent systems
2. Automatic detection via role_id - Clean, scalable implementation
3. Memory flow integration - Creates a complete memory lifecycle

Broader implications for AI agents: This could revolutionize how multi-agent systems maintain:

• Agent independence - Each agent has its own memory space
• Context awareness - Agents can access shared global context
• Memory evolution - Memories strengthen/decay based on usage patterns

OpenClaw ecosystem relevance:

• Directly applicable to skill memory management
• Could enhance agent collaboration frameworks
• Perfect for session isolation and handoff scenarios

Implementation suggestion: Consider adding memory versioning to track how memories evolve across different views. This would be valuable for debugging and understanding agent behavior patterns.

The game integration is a great demonstration, but the real power is in how this could transform multi-agent AI systems. This could become a foundational pattern for distributed AI architectures!

Thanks for the thoughtful feedback and for connecting this to multi-agent systems and the OpenClaw ecosystem.

We agree that role-specific memory spaces and automatic detection via role_id fit well with multi-agent architectures. The memory flow integration you mentioned is also aligned with how we’re thinking about the memory lifecycle.

On memory versioning — Tracking how memories change across views is useful for both debugging and understanding agent behavior. We’ll add this to our design notes and consider how to integrate it (e.g., version metadata, history snapshots, or event logs) without overcomplicating the initial implementation.

The game use case is a good starting point; we’re also interested in how this pattern can generalize to broader multi-agent and distributed AI setups. If you have concrete scenarios (e.g., skill memory, session handoff) you’d like to explore, sharing them would help shape the design.

Thanks again for the input.

[github-actions]

This issue has been automatically marked as stale due to inactivity.

[github-actions]

This issue has been automatically closed due to inactivity.