本技能用于 AI 助理的本地长期记忆管理。它把用户偏好、固定规则、项目上下文、词汇表、任务状态和历史摘要统一存入本地 SQLite 数据库,并以可直接注入的方式返回给模型,减少模型自行筛选记忆所带来的 token 消耗与解析成本。
This skill provides local long-term memory management for AI assistants. It stores preferences, rules, project context, glossary items, task state, and history summaries in a local SQLite database, then returns them in a direct-injection format so the model can use them immediately without spending extra tokens on memory selection.
python scripts/memory_cli.py direct --query "继续上次项目"
python scripts/memory_cli.py direct --query "继续上次项目" --compact
python scripts/memory_cli.py add --category preference --text "用户偏好中文回答"
共 5 个版本