---
name: xiaoyao-search
description: 小遥搜索 MCP 工具 - 本地文件智能搜索（语义/全文/图像/语音/混合搜索）
metadata:
  tags: mcp, search, local-files, semantic-search, vector-search, ai-tools
  version: 1.0.0
  author: xiaoyao-search
title: xiaoyao-search
canonical_url: https://skilld.dev/gh/dtsola/xiaoyaosearch/xiaoyao-search
last_updated: 2026-10-01T22:31:27.000Z
---

> Index of every page on this site: [llms.txt](https://skilld.dev/llms.txt)

> **Skill from skilld.dev.** Follow the user's requested action.
> If the user asked to fork this Skill, follow the fork workflow below. Do not execute the Skill instructions while copying it.
> Otherwise, follow the instructions below for this session. You do not need to install anything.
>
> Supporting files, fetch one when the Skill refers to it: [rules/connection.md](https://skilld.dev/api/skills-raw/dtsola/xiaoyaosearch/xiaoyao-search/rules/connection.md), [rules/examples.md](https://skilld.dev/api/skills-raw/dtsola/xiaoyaosearch/xiaoyao-search/rules/examples.md), [rules/tools.md](https://skilld.dev/api/skills-raw/dtsola/xiaoyaosearch/xiaoyao-search/rules/tools.md).
>
> If the user asked to install this Skill, run `npx skilld install dtsola/xiaoyaosearch/xiaoyao-search`. Install writes the Skill files into the project, so every session loads them.
>
> ## Fork workflow
>
> A fork creates an editable local Skill with its original author and licence. The request authorizes copying and local installation.
> 1. Check `./skills/xiaoyao-search`, the project lockfile, and selected Agent targets together. If the local directory or installed Skill exists, stop. Never overwrite an existing directory or Agent target.
> 2. Read [source metadata](https://skilld.dev/api/v1/skills/dtsola/xiaoyaosearch/xiaoyao-search) once. Use sourceUrl, sourceCommit, skillPath, sourceGone, and license. If the source is gone or its path is missing, stop. If license is null, read licence files at the source commit.
> 3. Fetch only the source commit into a temporary Git repository. Do not clone full history. Derive repository_url from sourceUrl, including repository renames. If sourceCommit is absent, resolve the sourceUrl ref once. Set source_commit to that actual commit. Run these commands in one shell call:
>
> ```sh
> git init --quiet "$temporary_dir"
> git -C "$temporary_dir" fetch --quiet --depth=1 "$repository_url" "$source_commit"
> git -C "$temporary_dir" checkout --quiet --detach FETCH_HEAD
> ```
>
> Read applicable licence declarations and notices at that commit. If copying is not permitted, report the restriction and stop.
> 4. Inspect source entries together, then copy the directory containing skillPath into `./skills/xiaoyao-search`. Use the user's path if selected. Keep the original SKILL.md, relative links, scripts, binary assets, and executable modes. Exclude .git metadata. Reject symlinks and paths outside the Skill directory. After checking entries, use cp -a where available. A regular source directory needs no custom copy script. Do not save this page wrapper as SKILL.md.
> Preserve author credit, notices, and applicable licence files from repository or parent directories. Add PROVENANCE.md with the Skill page, source URL, actual commit, original path, and licence. Retain any existing PROVENANCE.md and record new provenance separately. Batch source inspection, copying, and provenance work where practical.
> 5. In the project root, run `skilld install ./skills/xiaoyao-search --mode copy --plain`. If skilld is unavailable, use `npx skilld install ./skills/xiaoyao-search --mode copy --plain`. This known command needs no help lookup. Install does not support --json. Use detected Agent targets, or add --agent for the targets the user selected. Install the local path, never the upstream selector. If installation fails, preserve the local copy and report the exact failure.
> 6. Confirm the local lockfile source and installed Agent copies once. Report the local path, actual commit, and Agent targets. After edits, reinstall the same local path. Upstream updates must not replace it. Do not publish or push unless the user asks.

## When to use

当需要搜索本地文件时使用此 Skill：

- 通过自然语言描述搜索本地文档
- 查找包含特定关键词的文件
- 用图片搜索相似的本地图片
- 用语音转文字后搜索文件
- 需要综合语义和全文搜索结果

## Capability

小遥搜索 MCP 服务器提供以下 5 个搜索工具：

### 1. semantic_search（语义搜索）
- **用途**：基于 BGE-M3 向量模型的语义搜索
- **适用**：自然语言查询，如"关于机器学习的文档"
- **参数**：
  - `query`: string - 搜索关键词（1-500字符）
  - `limit`: int - 返回结果数（1-100，默认20）
  - `threshold`: float - 相似度阈值（0.0-1.0，默认0.5）
  - `file_types`: string[] - 文件类型过滤（可选）

### 2. fulltext_search（全文搜索）
- **用途**：基于 Whoosh 的精确关键词匹配
- **适用**：查找特定术语、代码片段
- **参数**：
  - `query`: string - 搜索关键词（1-500字符）
  - `limit`: int - 返回结果数（1-100，默认20）
  - `file_types`: string[] - 文件类型过滤（可选）

### 3. hybrid_search（混合搜索）
- **用途**：结合语义和全文搜索，RRF 算法融合结果
- **适用**：需要全面搜索结果的场景
- **参数**：
  - `query`: string - 搜索关键词（1-500字符）
  - `limit`: int - 返回结果数（1-100，默认20）
  - `threshold`: float - 相似度阈值（0.0-1.0，默认0.5）
  - `file_types`: string[] - 文件类型过滤（可选）

### 4. image_search（图像搜索）
- **用途**：基于 CN-CLIP 的图像相似度搜索
- **适用**：用图片查找相似的本地文件
- **参数**：
  - `image_path`: string - 图片绝对路径
  - `limit`: int - 返回结果数（1-100，默认20）
  - `threshold`: float - 相似度阈值（0.0-1.0，默认0.5）

### 5. voice_search（语音搜索）
- **用途**：FasterWhisper 语音识别后搜索
- **适用**：通过语音快速搜索本地文件
- **参数**：
  - `audio_path`: string - 音频绝对路径
  - `search_type`: string - 搜索类型（semantic/fulltext/hybrid，默认semantic）
  - `limit`: int - 返回结果数（1-100，默认20）
  - `threshold`: float - 相似度阈值（0.0-1.0，默认0.5）

## Usage

### 前提条件

1. 小遥搜索后端服务已启动（默认 http://127.0.0.1:8000）
2. MCP 功能已启用（默认启用）
3. 索引已构建完成

### 调用方式

Claude Code 通过 MCP 协议自动发现和调用工具：

```
用户：帮我找一下关于 Python 异步编程的文档
→ Claude Code 调用 semantic_search(query="Python 异步编程")
→ 返回搜索结果
→ Claude 基于结果回答用户
```

### 响应格式

搜索结果返回 JSON 格式：
```json
{
  "total": 10,
  "search_time": 0.523,
  "results": [
    {
      "file_name": "async_guide.md",
      "file_path": "D:/docs/async_guide.md",
      "file_type": "markdown",
      "relevance_score": 0.95,
      "preview_text": "Python异步编程指南..."
    }
  ]
}
```

## Examples

### 示例1：语义搜索
```
用户：查找关于机器学习算法优化的文档

调用：semantic_search(
  query="机器学习算法优化",
  limit=10,
  threshold=0.5
)
```

### 示例2：全文搜索
```
用户：查找包含 async def 的 Python 文件

调用：fulltext_search(
  query="async def",
  limit=20,
  file_types=["python"]
)
```

### 示例3：图像搜索
```
用户：找一下跟这张图片相似的设计稿

调用：image_search(
  image_path="C:/Users/test/reference.png",
  limit=10
)
```

### 示例4：语音搜索
```
用户：（语音输入）帮我找一下上周的会议记录

调用：voice_search(
  audio_path="C:/Users/test/voice.mp3",
  search_type="semantic"
)
```

### 示例5：混合搜索
```
用户：全面搜索关于 Vue3 组件开发的资料

调用：hybrid_search(
  query="Vue3 组件开发",
  limit=20,
  threshold=0.6
)
```

## Configuration

### 环境变量

| 变量名 | 默认值 | 说明 |
|-------|--------|------|
| MCP_SSE_ENABLED | true | 启用 MCP SSE |
| MCP_SERVER_NAME | xiaoyao-search | 服务器名称 |
| MCP_DEFAULT_LIMIT | 20 | 默认结果数 |
| MCP_DEFAULT_THRESHOLD | 0.5 | 默认相似度阈值 |
| MCP_VOICE_ENABLED | true | 启用语音搜索 |

### 连接测试

```bash
# 测试 MCP SSE 端点
curl http://127.0.0.1:8000/mcp

# 检查后端健康状态
curl http://127.0.0.1:8000/api/system/health
```

## Troubleshooting

### 问题1：工具调用失败
- 检查后端服务是否启动
- 确认 MCP_SSE_ENABLED=true
- 查看 /mcp 端点是否可访问

### 问题2：搜索结果为空
- 确认索引已构建
- 检查查询词是否正确
- 降低 threshold 值尝试

### 问题3：图像/语音搜索失败
- 确认文件路径为绝对路径
- 检查文件格式是否支持
- 检查文件大小是否在限制内（10MB）
