← 返回 Skills 市场

Local Vector Store

Name: Local Vector Store
Author: jackfeng0614-prog

作者 jackfeng0614-prog · GitHub ↗ · v1.0.0 · MIT-0

cross-platform ⚠ suspicious

111

总下载

当前安装

版本数

在 OpenClaw 中安装

/install local-vector-store

功能描述

Implements semantic search using local vector embeddings for knowledge base indexing and similarity matching. Use when you need to search documents by meanin...

使用说明 (SKILL.md)

Local Vector Store

A lightweight semantic search engine that indexes documents as vectors and enables similarity-based retrieval without external APIs.

Features

Document indexing with vector embeddings
Semantic similarity search
Local storage (no external dependencies)
Batch indexing support
Configurable embedding dimensions
Cosine similarity matching

Usage

const vectorStore = require('./local-vector-store');

// Initialize store
const store = await vectorStore.create({
  dimension: 384,
  storePath: '/tmp/vector-store'
});

// Index documents
await store.index({
  id: 'doc1',
  content: 'Machine learning is a subset of artificial intelligence',
  metadata: { source: 'wiki' }
});

// Search by semantic similarity
const results = await store.search({
  query: 'AI and deep learning',
  topK: 5,
  threshold: 0.7
});

// Batch operations
await store.indexBatch([
  { id: 'doc2', content: 'Neural networks process data' },
  { id: 'doc3', content: 'Algorithms solve computational problems' }
]);

Configuration

Set environment variables:

VECTOR_DIMENSION: Embedding dimension (default: 384)
STORE_PATH: Local storage directory (default: /tmp/vector-store)
SIMILARITY_THRESHOLD: Minimum similarity score (default: 0.5)

Output Format

{
  "query": "semantic search",
  "results": [
    {
      "id": "doc1",
      "content": "...",
      "similarity": 0.92,
      "metadata": {}
    }
  ],
  "searchTime": 45
}

安全使用建议

This skill appears to implement exactly what it claims (a local vector index) and does not request credentials or network access, but it has dangerous file operations that you should review before installing. Specifically: do not set STORE_PATH to system or important directories; prefer running in a sandbox or container with limited filesystem permissions; review the code paths for delete() and clear() (they delete all files in STORE_PATH) and consider hardening (validate IDs, restrict to a dedicated subdirectory, only remove *.json files, or require confirmation). Also verify the usage example matches your installation path (the example require path may not match index.js). If you cannot review or sandbox the code, treat this as higher risk.

能力评估

✓ Purpose & Capability

Name, description, SKILL.md, and code align: the module implements local vector generation, indexing, search, batch operations, persistence to a local storePath, and configuration via environment variables. There are no network calls or unrelated dependencies requested.

ℹ Instruction Scope

SKILL.md stays within the feature set (index, batch, search, env config). It only references configuration env vars declared in the README (VECTOR_DIMENSION, STORE_PATH, SIMILARITY_THRESHOLD). Minor inconsistency: example shows require('./local-vector-store') while the repository provides index.js at the root — this may be a usage/documentation mismatch but not a direct security issue.

✓ Install Mechanism

No install spec or external downloads; the skill is instruction-only with included source files. Nothing is fetched from third-party URLs or package registries during install.

✓ Credentials

No credentials or sensitive environment variables are requested. The only env vars are configuration (dimension, store path, threshold), which are proportionate to the claimed functionality.

⚠ Persistence & Privilege

The code persists documents to a file path and exposes delete() and clear() that synchronously unlink files in the configured STORE_PATH without validation. If STORE_PATH is misconfigured (e.g., set to / or another important directory), clear() will attempt to delete all files in that directory. There are no safeguards against path traversal in document IDs or checks to limit deletions to a safe subdirectory or filename pattern.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install local-vector-store
安装完成后，直接呼叫该 Skill 的名称或使用 /local-vector-store 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v1.0.0

Local-vector-store 1.0.0 initial release - Implements local semantic search by indexing documents as vector embeddings. - Supports document indexing, batch indexing, and similarity-based retrieval using cosine similarity. - No external dependencies; stores data locally. - Configurable embedding dimensions and storage path. - Environment variables for easy setup of vector dimension, store path, and similarity threshold.

元数据

Slug local-vector-store

版本 1.0.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 1

常见问题