OpenClaw 为什么总失忆？一套从单会话到永久记忆的 Memory 方案

1
2
3
4
5

# 1. 通过 OpenClaw 插件安装
openclaw plugins install @martian-engineering/lossless-claw

# 如果你是从源码运行 OpenClaw
pnpm openclaw plugins install @martian-engineering/lossless-claw

1
2
3
4
5
6
7

{
  "plugins": {
    "slots": {
      "contextEngine": "lossless-claw"
    }
  }
}

1
2
3
4

# 添加到你的 ~/.profile 或 ~/.zshrc
export LCM_FRESH_TAIL_COUNT=32
export LCM_INCREMENTAL_MAX_DEPTH=-1
export LCM_CONTEXT_THRESHOLD=0.75

1
2

openclaw plugins list
# 应该看到 lossless-claw 已启用

1
2
3
4
5

# 1. 安装支持扩展的 SQLite（macOS）
brew install sqlite

# 2. 安装 Bun（如果还没有）
curl -fsSL https://bun.sh/install | bash

1
2
3
4
5

# 全局安装 qmd（注意：使用 GitHub 仓库地址，不是 npm 包）
bun install -g https://github.com/tobi/qmd

# 或者直接下载 release
# https://github.com/tobi/qmd/releases

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

{
  "agents": {
    "defaults": {
      "memory": {
        "backend": "qmd",
        "citations": "auto",
        "qmd": {
          "includeDefaultMemory": true,
          "update": {
            "interval": "5m",
            "debounceMs": 15000
          },
          "limits": {
            "maxResults": 6,
            "timeoutMs": 4000
          },
          "scope": {
            "default": "deny",
            "rules": [
              { "action": "allow", "match": { "chatType": "direct" } }
            ]
          },
          "paths": [
            { "name": "notes", "path": "~/notes", "pattern": "**/*.md" },
            { "name": "docs", "path": "~/project/docs", "pattern": "**/*.md" }
          ]
        }
      }
    }
  }
}

1
2
3
4
5
6

# 重启 OpenClaw
openclaw restart

# 检查 qmd 是否正常运行
openclaw status
# 应该显示 memory backend: qmd

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

# 使用与 OpenClaw 相同的 XDG 配置
STATE_DIR="${OPENCLAW_STATE_DIR:-$HOME/.openclaw}"
export XDG_CONFIG_HOME="$STATE_DIR/agents/main/qmd/xdg-config"
export XDG_CACHE_HOME="$STATE_DIR/agents/main/qmd/xdg-cache"

# 强制刷新索引 + 生成嵌入
qmd update
qmd embed

# 预热模型下载
qmd query "test" -c memory-root --json >/dev/null 2>&1

1
2

# 在 ~/.profile 中添加，使用 Qwen3-Embedding 支持 119 种语言
export QMD_EMBED_MODEL="hf:Qwen/Qwen3-Embedding-0.6B-GGUF/Qwen3-Embedding-0.6B-Q8_0.gguf"

1
2

# 使用上述 XDG 配置后执行
qmd embed --force

1
2

nmem status
# 应该显示 Nowledge Mem 正在运行

1

openclaw plugins install @nowledge/openclaw-nowledge-mem

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17

{
  "plugins": {
    "slots": {
      "memory": "openclaw-nowledge-mem"
    },
    "entries": {
      "openclaw-nowledge-mem": {
        "enabled": true,
        "config": {
          "autoRecall": true,
          "autoCapture": false,
          "maxRecallResults": 5
        }
      }
    }
  }
}

 1
 2
 3
 4
 5
 6
 7
 8
 9
10

# 验证配置
openclaw nowledge-mem status
# 应该显示 "Nowledge Mem is accessible"

# 测试流程：
# 1. /remember 我们选择 PostgreSQL 做任务事件的数据库
# 2. /recall PostgreSQL
# 3. /new 开启新会话
# 4. 问：任务事件我们选的什么数据库？
# 5. 如果能回答出来，就配置成功了

1

# 按照 SKILL.md 说明一步安装，mem9 提供自动安装脚本

1
2
3
4
5

# 方案 A：克隆到 workspace/plugins/ 目录
cd /path/to/your/openclaw/workspace
git clone https://github.com/CortexReach/memory-lancedb-pro.git plugins/memory-lancedb-pro
cd plugins/memory-lancedb-pro
npm install

1

npm install memory-lancedb-pro@beta

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12

{
  "autoCapture": true,
  "autoRecall": true,
  "autoRecallMinLength": 8,
  "autoRecallSelectionMode": "mmr",
  "autoRecallExcludeReflection": true,
  "retrieval": {
    "candidatePoolSize": 20,
    "minScore": 0.45,
    "hardMinScore": 0.55
  }
}

1

openclaw gateway restart

1
2

cd /path/to/OpenViking
npx ./examples/openclaw-memory-plugin/setup-helper

1
2
3
4
5
6
7

# 1. 安装 OpenViking
pip install openviking --upgrade --force-reinstall

# 2. 安装 OpenClaw 插件
mkdir -p ~/.openclaw/extensions/memory-openviking
cp $OPENViking_REPO/examples/openclaw-memory-plugin/* ~/.openclaw/extensions/memory-openviking/
cd ~/.openclaw/extensions/memory-openviking && npm install

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20

{
  "plugins": {
    "enabled": true,
    "slots": {
      "memory": "memory-openviking"
    },
    "entries": {
      "memory-openviking": {
        "enabled": true,
        "config": {
          "mode": "local",
          "configPath": "~/.openviking/ov.conf",
          "targetUri": "viking://user/memories",
          "autoRecall": true,
          "autoCapture": true
        }
      }
    }
  }
}

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17

{
  "vlm": {
    "provider": "volcengine",
    "api_key": "<YOUR_API_KEY>",
    "model": "doubao-seed-2-0-pro-260215",
    "api_base": "https://ark.cn-beijing.volces.com/api/v3"
  },
  "embedding": {
    "dense": {
      "provider": "volcengine",
      "api_key": "<YOUR_API_KEY>",
      "model": "doubao-embedding-vision-251215",
      "api_base": "https://ark.cn-beijing.volces.com/api/v3",
      "dimension": 1024
    }
  }
}

1

source ~/.openclaw/openviking.env && openclaw gateway

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21

{
  "plugins": {
    "enabled": true,
    "slots": {
      "memory": "memory-openviking"
    },
    "entries": {
      "memory-openviking": {
        "enabled": true,
        "config": {
          "mode": "remote",
          "baseUrl": "http://your-server:1933",
          "apiKey": "optional-api-key",
          "targetUri": "viking://user/memories",
          "autoRecall": true,
          "autoCapture": true
        }
      }
    }
  }
}

1
2
3

lossless-claw  +  qmd / LanceDB
│           └── OpenClaw 原生混合搜索记忆后端（qmd GGUF本地 / LanceDB 嵌入式）
└─────────────── 单会话内无损压缩，长对话原始信息不丢

1
2
3

lossless-claw  +  OpenViking
│           └── 文件系统范式上下文数据库，任务完成率最高，Token成本最低
└────────────── 单会话内无损压缩

1
2
3
4

lossless-claw  +  Nowledge Mem  +  qmd / LanceDB
│           │           └── OpenClaw 原生混合搜索记忆后端（qmd GGUF本地 / LanceDB 嵌入式）
│           └────────────── 跨会话记住决定偏好，本地优先，结构化
└────────────────────────── 单会话内无损压缩

1
2
3
4

lossless-claw  +  OpenViking (独立部署) / mem9.ai  +  qmd / LanceDB
│           │                       └── 云端/独立部署服务，跨设备同步记忆
│           └───────────────────────── 零运维开箱即用，OpenViking 效果更好但需要自己部署服务
└────────────────────────────────────── 单会话压缩

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19

{
  "memory": {
    "backend": "postgres",
    "postgres": {
      "connectionString": "postgresql://user:pass@host:5432/dbname",
      "tablePrefix": "openclaw_memory",
      "embedding": {
        "provider": "voyage",
        "model": "voyage-3-lite",
        "dimensions": 512
      },
      "hybrid": {
        "enabled": true,
        "vectorWeight": 0.7,
        "textWeight": 0.3
      }
    }
  }
}