Release v0.2.0

We are thrilled to officially release MemOS v0.2.0 (Stellar, 星河), an early preview version of MemOS v1.0.

MemOS is an operating system for Large Language Models (LLMs) designed to enhance their long-term memory capabilities. This framework enables LLMs to store, retrieve, and manage information, facilitating more context-aware, consistent, and personalized interactions.

The current release includes the following features:

• Support for CRUD operations across three memory layers: Memory Manager, MemCube, and MemOS
  • The Memory Manager currently supports: General Memory Management, Tree Memory Management, and KV Cache Memory Management
  • MemCube integrates three memory types (Textual Memory, Activation Memory, and Parametric Memory), representing similar memory contents through different storage formats
  • MemOS serves as the top-level abstraction for managing multiple MemCubes
• MemOS Capabilities
  • Basic chat functionality with LLMs through MemOS
  • Offline batch document loading and parsing into MemCubes
• FastAPI interface for the MemOS layer
• Preliminary MemScheduler functionality
• Various infrastructure components including LLM, Embedder, Neo4j graph DB, Qdrant vector DB, Parser, and Chunker

Note: This is currently a preview version with known limitations and bugs. We welcome contributions from the community!

What's Changed

• refactor: update feature request template by @Ki-Seki in #50
• docs: update documentation links and images by @Ki-Seki in #55
• feat: add structure reorganizer and conflict resolver by @CaralHsi in #14
• feat(eval): add openai memory on locomo - eval guide by @Duguce in #54
• feature: mem reader update and some bug fixs by @Nyakult in #43
• chore: bump version number to v0.2.0 by @fridayL in #66

New Contributors

• @CaralHsi made their first contribution in #14
• @Nyakult made their first contribution in #43

Full Changelog: v0.1.13...v0.2.0

我们很高兴正式发布 MemOS v0.2.0 (Stellar, 星河)，这是 MemOS v1.0 的早期预览版本。

MemOS 是一个专为大型语言模型（LLMs）设计的操作系统，旨在增强其长期记忆能力。该框架使 LLMs 能够存储、检索和管理信息，促进更具上下文感知、一致性和个性化的交互。

当前版本包含以下功能：

• 支持三个内存层的 CRUD 操作：Memory Manager、MemCube 和 MemOS
  • Memory Manager 目前支持：通用内存管理、树形内存管理和 KV 缓存内存管理
  • MemCube 集成了三种内存类型（文本内存、激活内存和参数内存），通过不同的存储格式表示相似的内存内容
  • MemOS 作为管理多个 MemCubes 的顶层抽象
• MemOS 功能
  • 通过 MemOS 与 LLMs 的基本聊天功能
  • 离线批量文档加载和解析到 MemCubes
• MemOS 层的 FastAPI 接口
• 初步的 MemScheduler 功能
• 各种基础设施组件，包括 LLM、嵌入器、Neo4j 图数据库、Qdrant 向量数据库、解析器和分块器

注意：这是目前的预览版本，存在已知的限制和错误。我们欢迎社区的贡献！