菜单
Aiden的个人记忆API。支持向量语义搜索、混合权重模型、多种返回格式。省token模式用format=compact。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
AI Agent Skill for meeting management. Audio transcription and structured storage with entity linking.
Automatically archive old documents from active/daily/ to archive/ based on file age. Archives files older than specified days into archive/YYYY-MM/ directories.
Security guidance and best practices for code development. Use when writing authentication code, handling sensitive data, validating user input, designing APIs, or reviewing code for security issues.
Generate codemods for large-scale code changes. Use when refactoring patterns across many files.
Automatically archive old documents from active/ to archive/ based on age and relevance rules.
Use Brave Search API to search the web and return structured results (title/url/snippet). Trigger when the user asks to enable Brave search, configure Brave API key, or perform web search using Brave.
| name | vivian-memory |
| description | Aiden的个人记忆API。支持向量语义搜索、混合权重模型、多种返回格式。省token模式用format=compact。 |
Aiden 的个人记忆基础设施。任何 AI 都可以调用，但必须遵守协议。
API代码位置: skills/vivian-memory/api/
https://memory-api.vivian-memory.workers.dev
Authorization: Bearer my-memory-secret-123
curl -s ".../memory/search?q=<query>&limit=5&format=compact" -H "Authorization: Bearer ..."
| 参数 | 默认 | 说明 |
|---|---|---|
| q | 必填 | 自然语言查询 |
| limit | 5 | 返回数量，最大20 |
| type | all | all/personal/assistant/project |
| format | full | text/compact/full（见下方） |
format 参数（省token）：
| 值 | 说明 |
|---|---|
| text | 纯文本，---分隔，最省token |
| compact | 只有text+score |
| full | 全部字段+_debug |
推荐日常用 format=compact
curl -s -X POST ".../memory" -H "Auth..." -d '{"items": [{"text": "内容", "weight": 0.8}]}'
| 字段 | 默认 | 说明 |
|---|---|---|
| text | 必填 | 原子化（一条一个事实） |
| weight | 1.0 | 0.3-1.0 |
| tags | [] | bio/pref/tech/work/goal |
| type | personal | personal/assistant/project |
批量最多20条/请求。
curl -s -X DELETE ".../memory/{id}" -H "Auth..."
curl -s ".../memory/stats" -H "Auth..."
https://memory-api.vivian-memory.workers.dev/viz
final_score = similarity × (0.5×base + 0.3×recency + 0.2×frequency)
| 因子 | 说明 |
|---|---|
| base | 写入时的weight |
| recency | 1/(1+days/30)，越近越高 |
| frequency | log(count+1)/log(10)，召回越多越重要 |
用 format=full 看 _debug 字段调试。
位置: scripts/utils/memory-batch.sh
./memory-batch.sh add "内容" 0.8 "tag" # 暂存
./memory-batch.sh list # 查看
./memory-batch.sh flush # 批量上传
./memory-batch.sh import file.txt # 导入
✅ 写入：新事实、状态变化、明确说"记住"的内容 ❌ 不写：废话、重复、临时细节
原子化：一条 = 一个事实 权重：1.0核心 / 0.8长期 / 0.5临时 / 0.3琐碎
冲突处理：search找旧的 → delete → 写新的