从 MEMORY.md 迁移到 MemoClaw:逐步指南

📄 中文摘要

OpenClaw 代理的 MEMORY.md 文件可能已经变得冗长,包含了大量信息。在每次会话中,代理需要将整个文件加载到上下文窗口中,消耗大量令牌来读取几个月前的内容,导致重要信息可能被遗忘。该指南提供了将这些积累的知识迁移到 MemoClaw 的步骤,使代理能够更有效地回忆相关信息,而不是重新阅读所有内容。OpenClaw 工作区通常包含两层记忆:一是长期上下文的 MEMORY.md 文件,二是按日期记录的每日会话日志。这两者在会话开始时都会被加载到上下文窗口中。

📄 English Summary

Migrating from MEMORY.md to MemoClaw: a step-by-step guide

The MEMORY.md file of an OpenClaw agent can become lengthy, containing a significant amount of information. During each session, the agent loads the entire file into its context window, consuming tokens to read content from months ago, which may lead to forgetting important details. This guide outlines the steps to migrate this accumulated knowledge to MemoClaw, allowing the agent to recall relevant information more efficiently instead of re-reading everything. A typical OpenClaw workspace consists of two layers of memory: the MEMORY.md file for long-term context and daily session logs organized by date. Both are loaded into the context window at the start of each session.

Powered by Cloudflare Workers + Payload CMS + Claude 3.5

数据源: OpenAI, Google AI, DeepMind, AWS ML Blog, HuggingFace 等