Memary

The Open Source Memory Layer For Autonomous Agents

No commits in 23 months — may not be actively maintained. See maintained alternatives →

open-sourcememory-knowledge

Visit Website View on GitHub

2.7k

Stars

+12

Stars/month

Commits (90d)

Releases (6m)

Package downloads, last 30 days: PyPI memary 40 · counts from npm and pypistats, updated weekly · most-downloaded agent tools

Star Growth

+75 (2.9%)

Overview

Memary是一个开源的内存层框架，专为自主AI智能体设计，模拟人类记忆机制来提升智能体的推理能力。该工具允许AI智能体在多次会话中保持持久记忆，就像人类处理任务时依赖记忆一样。Memary支持本地模型（通过Ollama运行Llama 3等）和云端模型（如GPT-3.5），提供灵活的模型切换能力。它集成了多种数据库（Neo4J、FalkorDB）来存储记忆数据，并支持视觉模型（LLaVA、GPT-4 Vision）处理多模态输入。这个框架对于构建需要长期上下文理解和人类级别推理能力的AI应用至关重要，特别是在需要跨会话保持状态的复杂任务场景中。

Deep Analysis

Key Differentiator

vs LangChain Memory / Mem0: graph-database-backed memory system emulating human memory (breadth + depth tracking) — agents automatically build and query knowledge graphs rather than flat conversation history

⚡ Capabilities

• Human-like memory system for AI agents (auto-generated memory)
• Memory Stream tracking entity exposure breadth over time
• Entity Knowledge Store monitoring reference frequency and recency
• Graph database integration for contextual knowledge storage
• Recursive retrieval with multi-hop reasoning (max depth 2)
• ReAct agent framework with tool ecosystem (search, vision, location, stocks)
• Automatic response injection back into knowledge graphs

🔗 Integrations

FalkorDBNeo4jOllama (Llama 3)OpenAI (GPT-3.5/GPT-4)Perplexity AIGoogle MapsAlpha VantageLlamaIndex

✓ Best For

✓ Building persistent, context-aware AI agents with evolving memory
✓ User preference tracking and personalization across sessions
✓ Multi-user agent management with separate knowledge contexts

✗ Not Ideal For

✗ Stateless single-turn chatbot applications
✗ Teams without graph database infrastructure
✗ Production agents requiring mature, battle-tested memory

Languages

Python

Deployment

pip installDocker/Ollama for local modelsStreamlit web interface

⚠ Known Limitations

⚠ Python version constraint (<= 3.11.9)
⚠ ReAct agent is demo-only — future versions will support external agents
⚠ Memory rewinding feature not yet implemented
⚠ Requires graph database setup (FalkorDB or Neo4j)

Pros

+ 开源透明的记忆管理系统，允许完全自定义和扩展记忆机制
+ 同时支持本地模型（Ollama）和云端模型（OpenAI），提供灵活的部署选择
+ 内置模型切换功能，可以无缝在不同AI提供商之间切换而无需重写代码

Cons

- 严格的Python版本限制（<=3.11.9），可能与较新的开发环境不兼容
- 复杂的初始配置，需要设置多个API密钥和数据库连接
- 依赖特定的模型框架和外部服务，增加了系统的复杂性和维护成本

Use Cases

• 构建需要跨会话保持记忆的AI客服或助手系统，提供个性化的用户体验
• 开发具有长期学习能力的自主AI智能体，用于复杂的决策和规划任务
• 创建多轮对话AI应用，如教育助手或咨询系统，需要记住历史交互内容

Getting Started

1. 使用Python 3.11.9或更低版本创建虚拟环境，然后运行pip install memary安装包。2. 创建.env文件配置所需的API密钥（OpenAI、Perplexity、Google Maps等）和数据库连接（Neo4J或FalkorDB）。3. 选择并配置AI模型（本地Ollama模型或云端模型），然后开始构建具有记忆功能的智能体应用。