{"skill":{"slug":"memory-core","displayName":"Memory Core","summary":"OpenClaw 长期记忆核心：基于 LanceDB 的向量化长期记忆存储与检索，内置意图/场景隔离以防记忆污染。","description":"---\nname: memory-core\ndescription: \"OpenClaw 长期记忆核心：基于 LanceDB 的向量化长期记忆存储与检索，内置意图/场景隔离以防记忆污染。\"\n---\n\n# Memory Core（长期记忆核心）\n\nMemory Core 为 OpenClaw Agent 提供跨会话的长期记忆能力：自动识别输入的意图与场景，并在检索阶段进行严格场景隔离，避免无关记忆污染当前会话。\n\n## 功能特性\n\n- 记忆摄入：将用户关键事实写入长期记忆库\n- 记忆检索：按 Agent + 场景过滤后进行向量相似度搜索\n- 记忆遗忘：按记忆 ID 删除\n- 本地优先：LanceDB 文件本地持久化，默认不出机器\n\n## 命令\n\n在 OpenClaw Workspace 内运行（`{baseDir}` 为技能目录）：\n\n```bash\npython3 {baseDir}/scripts/main.py ingest --agent \"main\" --text \"我是 Python 后端工程师，喜欢用 FastAPI。\"\npython3 {baseDir}/scripts/main.py retrieve --agent \"main\" --query \"我擅长什么框架？\"\npython3 {baseDir}/scripts/main.py forget --id \"<memory_id>\"\n```\n\n## 配置\n\n你可以创建 `{baseDir}/config.json` 来选择本地或云端向量化模型：\n\n### 预算自适应（推荐开启）\n\n默认会根据 `agent_id` 在 `~/.openclaw/openclaw.json` 里对应的模型名做启发式分档，并自动选择检索预算：\n\n- small：400/1200\n- medium：600/1800\n- large：900/2700\n\n如需固定预算，可在 config.json 里设置 `auto_budget: false` 并手工指定 `max_chars_per_memory/max_total_chars`。\n\n### 1) 使用本地 Ollama（推荐本地优先）\n\n```json\n{\n  \"embedding_provider\": \"ollama\",\n  \"embedding_model\": \"nomic-embed-text\",\n  \"embedding_base_url\": \"http://localhost:11434\",\n  \"auto_budget\": true,\n  \"default_tier\": \"medium\",\n  \"embedding_timeout_sec\": 20,\n  \"embedding_max_input_chars\": 2000,\n  \"max_results\": 5,\n  \"max_chars_per_memory\": 600,\n  \"max_total_chars\": 1800,\n  \"min_score\": 0.2\n}\n```\n\n### 2) 使用云端 SiliconFlow（默认）\n\n`embedding_api_key` 可留空，系统会尝试从 `~/.openclaw/openclaw.json` 自动读取。\n\n```json\n{\n  \"embedding_provider\": \"siliconflow\",\n  \"embedding_model\": \"BAAI/bge-m3\",\n  \"embedding_base_url\": \"https://api.siliconflow.cn/v1\",\n  \"auto_budget\": true,\n  \"default_tier\": \"medium\",\n  \"embedding_timeout_sec\": 15,\n  \"embedding_max_input_chars\": 2000,\n  \"max_results\": 5,\n  \"max_chars_per_memory\": 600,\n  \"max_total_chars\": 1800,\n  \"min_score\": 0.2\n}\n```\n","tags":{"latest":"0.1.3"},"stats":{"comments":0,"downloads":2358,"installsAllTime":22,"installsCurrent":22,"stars":1,"versions":3},"createdAt":1772701979766,"updatedAt":1778995509911},"latestVersion":{"version":"0.1.3","createdAt":1772714155807,"changelog":"Auto budget by agent model heuristic (small/medium/large); add auto_budget & default_tier config; keep hard caps to protect small-context models.","license":null},"metadata":null,"owner":{"handle":"lilei0311","userId":"s171k5jax63a1zk2r5by9a2zes885j88","displayName":"MaxStormSpace","image":"https://avatars.githubusercontent.com/u/67162380?v=4"},"moderation":{"isSuspicious":false,"isMalwareBlocked":false,"verdict":"clean","reasonCodes":["review.llm_review"],"summary":"Review: review.llm_review","engineVersion":"v2.4.24","updatedAt":1780089762790}}