memory-manager
用于为 PromptWar̊e ØS 矢量内存子系统准备和管理数据的工具。
📖 中文导读:用于为 PromptWar̊e ØS 矢量内存子系统准备和管理数据的工具。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。
name: Memory Manager description: Tools for preparing and managing data for the PromptWar̊e ØS Vector Memory Subsystem. category: memory tags: [vector, chunking, embedding, rag] tools: - ./chunker.ts
Memory Manager
A suite of tools designed to bridge the gap between raw files and the PromptWar̊e ØS Vector Database.
Purpose
The Kernel provides the mechanism (Vector.Store), but not the policy for how to slice data. This skill provides the policy: intelligent chunking, metadata extraction, and formatting.
Tools
1. Chunker (chunker.ts)
Splits text or files into semantic chunks suitable for embedding.
Usage:
# Chunk a file
deno run -A chunker.ts --file README.md
# Chunk raw text
deno run -A chunker.ts "Some long text..."
# Custom chunk size (default: 1000 chars)
deno run -A chunker.ts --file README.md --size 500
# JSON Output (for programmatic use)
deno run -A chunker.ts --file README.md --json
Output Format (JSON):
[
{
"text": "Chunk 1 content...",
"metadata": {
"source": "README.md",
"index": 0,
"total": 5
}
},
...
]
Workflow
- Hydrate: Agent loads this skill.
- Chunk: Agent runs
chunker.tson a target file. - Store: Agent iterates over the output and calls
Vector.Storefor each chunk.