# unknown

> Perform web searches using Kimi 2.5 API and save complete search results (links, titles, full content) locally in both JSON and Markdown formats.

- Author: sethyinn
- Repository: sethyinn/Kimi-research
- Version: 20260209183135
- Stars: 0
- Forks: 0
- Last Updated: 2026-02-09
- Source: https://github.com/sethyinn/Kimi-research
- Web: https://mule.run/skillshub/@@sethyinn/Kimi-research~unknown:20260209183135

---

---
name: kimi-search
description: Kimi 2.5 powered web search with full content extraction. Use when users ask to: (1) Search the web using Kimi/Moonshot AI, (2) Perform multi-source keyword searches and save results locally, (3) Retrieve detailed content (links, titles, full text) from web searches, (4) Research topics using Kimi 2.5's联网搜索 capability, (5) Save search results in both JSON and Markdown formats
---

# Kimi 2.5 Web Search

Perform web searches using Kimi 2.5 API and save complete search results (links, titles, full content) locally in both JSON and Markdown formats.

## Quick Start

### First-Time Setup

Before using this skill, configure your Kimi API key:

1. Get your API key from https://platform.moonshot.cn/
2. Create `scripts/.env` file and add your key:
   ```
   KIMI_API_KEY=your_actual_api_key_here
   ```
3. Install dependencies:
   ```bash
   pip install -r scripts/requirements.txt
   ```

### Basic Usage

Search with 1-3 keywords:

```bash
python scripts/kimi_searcher.py --keywords "人工智能,机器学习" --output results/
```

This generates two files:
- `search_人工智能_机器学习_20240209_143022.json` - Structured data
- `search_人工智能_机器学习_20240209_143022.md` - Human-readable report

## Workflow

When a user requests a web search:

1. **Extract keywords**: Identify 1-3 key search terms from the user's request
2. **Execute search**: Run `kimi_searcher.py` with the keywords
3. **Verify results**: Check that both JSON and Markdown files were created successfully
4. **Report to user**: Provide file locations and result count

Example interaction:
- User: "帮我搜索关于量子计算的最新进展"
- Claude: Extracts keywords ["量子计算", "最新进展"], runs search, reports: "Found 8 results saved to quantum_20240209_143022.json and quantum_20240209_143022.md"

## Advanced Usage

### Process Existing Results

Convert or filter saved results:

```bash
# Convert JSON to HTML
python scripts/process_results.py --input results/search_data.json --format html --output report.html

# Filter results by keywords
python scripts/process_results.py --input results/search_data.json --filter "量子,计算" --format markdown
```

### Multiple Searches

Search multiple keyword combinations:

```bash
python scripts/kimi_searcher.py --keywords "人工智能" --output results/ai/
python scripts/kimi_searcher.py --keywords "机器学习,深度学习" --output results/ai/
python scripts/kimi_searcher.py --keywords "神经网络" --output results/ai/
```

## Output Formats

### JSON Format
Structured data suitable for programmatic processing:
```json
{
  "results": [
    {
      "link": "https://example.com/article",
      "title": "文章标题",
      "content": "完整的文章内容..."
    }
  ]
}
```

### Markdown Format
Human-readable document with:
- Search timestamp
- Result count
- Numbered results with titles, links, and full content

## Important Notes

- **Kimi 2.5 Model**: Uses `moonshot-v1-128k` for enhanced web search capabilities
- **Full Content Only**: Returns complete article content, not summaries
- **Multi-source**: Searches across multiple web sources
- **Local Storage**: All results saved locally, no cloud dependencies
- **API Rate Limits**: Respect Moonshot API rate limits (typically 60 requests/minute)
- **Timeout**: Search operations may take 30-120 seconds depending on query complexity

## Troubleshooting

**Error: KIMI_API_KEY not found**
→ Ensure `scripts/.env` file exists with valid API key

**Error: API request failed**
→ Check internet connection and verify API key is valid

**Empty results**
→ Try different keyword combinations or check if topic has web coverage

**JSON parse error**
→ Kimi may have returned non-JSON response; check API status