# Example Queries for Sample Dataset
This document provides example queries you can use to test the `sample-docs` dataset.
## Basic Queries
### Getting Started
```json
{
"tool": "knowledge.search",
"arguments": {
"dataset": "sample-docs",
"query": "How do I get started with MCP Power?",
"topK": 3
}
}
```
**Expected**: Returns documents about getting started and installation.
### Configuration
```json
{
"tool": "knowledge.search",
"arguments": {
"dataset": "sample-docs",
"query": "How to configure a dataset",
"topK": 5
}
}
```
**Expected**: Returns documents about dataset configuration and manifest files.
### Tool Usage
```json
{
"tool": "knowledge.search",
"arguments": {
"dataset": "sample-docs",
"query": "What tools are available?",
"topK": 3
}
}
```
**Expected**: Returns documents about the search and listDatasets tools.
### Troubleshooting
```json
{
"tool": "knowledge.search",
"arguments": {
"dataset": "sample-docs",
"query": "Search is not working",
"topK": 3
}
}
```
**Expected**: Returns troubleshooting documentation.
## Advanced Queries
### Semantic Understanding
```json
{
"tool": "knowledge.search",
"arguments": {
"dataset": "sample-docs",
"query": "I'm having issues with performance",
"topK": 3
}
}
```
**Expected**: Should find troubleshooting content about performance issues.
### Different Phrasings
```json
{
"tool": "knowledge.search",
"arguments": {
"dataset": "sample-docs",
"query": "setup instructions",
"topK": 2
}
}
```
**Expected**: Should match "Getting Started" content despite different wording.
## List Datasets Example
```json
{
"tool": "knowledge.listDatasets",
"arguments": {}
}
```
**Expected Response**:
```json
{
"datasets": [
{
"id": "sample-docs",
"name": "Sample Knowledge Base",
"description": "A sample knowledge base demonstrating the MCP Power server capabilities...",
"defaultTopK": 5
}
],
"metadata": {
"total": 1,
"ready": 1,
"errors": 0
}
}
```
## Testing from Command Line
You can test these queries using the MCP inspector or by writing a simple client script.
### Using Node.js
```javascript
import { Client } from '@modelcontextprotocol/sdk/client/index.js';
import { StdioClientTransport } from '@modelcontextprotocol/sdk/client/stdio.js';
const transport = new StdioClientTransport({
command: 'node',
args: ['dist/cli.js', '--datasets', './datasets']
});
const client = new Client({
name: 'test-client',
version: '1.0.0'
}, {
capabilities: {}
});
await client.connect(transport);
// Search example
const result = await client.callTool({
name: 'knowledge.search',
arguments: {
dataset: 'sample-docs',
query: 'How do I get started?',
topK: 3
}
});
console.log(JSON.stringify(result, null, 2));
```
## Tips for Writing Effective Queries
1. **Be Natural**: Write queries as you would ask a person
- Good: "How do I configure a dataset?"
- Also good: "dataset configuration"
2. **Use Keywords**: Include important terms
- "troubleshooting search errors"
- "FAISS index validation"
3. **Adjust topK**:
- Use 3-5 for focused results
- Use 10+ for comprehensive coverage
4. **Experiment**: Try different phrasings to see semantic matching in action
