Design patterns for adding persistent memory to AI agents. Covers conversation memory, entity extraction, knowledge graphs, tiered memory, and memory management strategies.
这里应该同时让用户和 Agent 知道第一步该复制什么、安装什么、落到哪里。
从对话缓冲开始，逐步升级到实体记忆和分层记忆。
定义 AI Agent 如何跨对话和会话记住信息的设计模式。
一句话总结：AI Agent 记忆设计模式，对话缓冲/摘要记忆/实体提取/知识图谱/分层记忆（Letta 模式），Mem0/Letta/Zep 等工具实现，生产 Agent 必备。
短对话 → 缓冲；长对话 → 摘要；多会话 → 实体+向量；复杂领域 → 知识图谱；全自主 → 分层。
Q: Claude Code 有记忆吗？ A: 有，auto-memory 系统在 ~/.claude/projects/*/memory/ 存储用户/项目/反馈记忆。
Mem0 | Letta | Zep | LangChain Memory
Comprehensive comparison of all AI coding agents in 2026. Covers Claude Code, Cursor, Codex, Gemini CLI, Cline, Roo Code, Windsurf, and Aider with feature matrices and recommendations.
Framework for building stateful, multi-step AI agent workflows as graphs. LangGraph enables cycles, branching, human-in-the-loop, and persistent state for complex agent systems.
Ultra-fast LLM inference at 2000+ tokens/second. Cerebras provides the fastest cloud inference for Llama and Qwen models with OpenAI-compatible API for instant AI responses.