skills.rest
dtsola
dtsolaCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

xiaoyao-search

Provides an MCP server for semantic, full-text, image, and voice search over local files.

1.0k|65|Updated Nov 11, 2025
Search Tags:#mcp#semantic-search#vector-search#local-search#image-search#ai-tools#fulltext-search
One-click install
npx skills add https://github.com/dtsola/xiaoyaosearch --skill xiaoyao-search
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: xiaoyao-search
Source: https://github.com/dtsola/xiaoyaosearch/tree/main/skills/xiaoyao-search
Command: npx skills add https://github.com/dtsola/xiaoyaosearch --skill xiaoyao-search

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

本技能提供在本地环境中对文件进行高效智能检索的能力，整合了语义、全文、本地图像和语音检索，帮助用户快速定位所需资料。

Core Features & Use Cases

  • 语义搜索（semantic_search）：基于 BGE-M3 向量模型实现自然语言理解的语义匹配，适用于“关于某主题的文档”这类自然语言查询。
  • 全文检索（fulltext_search）：基于 Whoosh 的关键词级匹配，支持精确短语和代码片段的检索。
  • 混合检索（hybrid_search）：将语义与全文结果融合，提升覆盖率和排序准确性。
  • 图像搜索（image_search）：基于 CN-CLIP 的图像相似度检索，用图片查找本地相关内容。
  • 语音检索（voice_search）：通过 FasterWhisper 将语音转文本后执行检索，支持多种搜索类型。

另外，MCP 服务器可与 Claude Desktop/Code 等 AI 助手对接，通过 SSE 实现工具自动发现与调用，帮助实现端到端的本地知识检索工作流。

Quick Start

Start the MCP server and call semantic_search with a query to retrieve results.

Frequently Asked Questions about xiaoyao-search

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I set up local semantic search for my documents using an MCP server?▼

Local semantic search requires starting an MCP server that indexes your files and exposes search tools via SSE, allowing AI assistants like Claude Desktop to invoke semantic_search for natural language queries.

Can I search local images by similarity using a vector model?▼

Yes, local image search uses the CN-CLIP vector model to perform image similarity retrieval, enabling you to find related local content by providing a query image.

What is hybrid search and when do I need it for local file retrieval?▼

Hybrid search combines semantic search using the BGE-M3 model with full-text keyword matching from Whoosh, improving coverage and ranking accuracy when you need both natural language and exact phrase matching.

Does full-text search support exact code snippet matching?▼

Yes, full-text search is powered by Whoosh and supports keyword-level matching, making it suitable for retrieving exact phrases and code snippets from indexed local files.

How do I use voice search to find local files with spoken queries?▼

Voice search converts spoken queries to text using FasterWhisper, then executes the search across indexed local files to retrieve relevant documents based on the transcribed input.

Do I need an active MCP server and indexed data before AI assistants can search my files?▼

Yes, an active MCP server with indexed local data is required for AI assistants to discover and call search tools via SSE, functioning as the backend for all semantic, full-text, image, and voice queries.

Related Skills

Explore More Software Engineering Skills→
lusya123lusya123

MCP Integration

Expose external services as MCP tools across stdio, SSE, HTTP, and WebSocket servers.

Community
Advanced
jpjednorskijpjednorski

qmd-obsidian

Index and search Obsidian vault notes with qmd keyword and semantic queries.

Community
Advanced
fcakyonfcakyon
1.0k

mcp-integration

Configure MCP servers and expose tools within Claude Code plugins.

Community
Intermediate
as4584as4584
2

hybrid-search-implementation

Combine vector and keyword search results using RRF, linear combination, or cross-encoder reranking.

Community
Advanced