web-crawler（网络爬虫）

Rust Web Crawler (rcrawler)

High-performance web crawler built in pure Rust with production-grade features for fast, reliable site crawling.

When to Use This Skill

Use this skill when the user requests:

• Web crawling or site mapping
• Sitemap discovery and analysis
• Link extraction and validation
• Site structure visualization
• robots.txt compliance checking
• Performance-critical web scraping
• Generating interactive web reports with graph visualization

Core Capabilities

🚀 Performance

• 60+ pages/sec throughput with async Tokio runtime
• <50ms startup time - Near-instant initialization
• ~50MB memory usage - Efficient resource consumption
• 5.4 MB binary - Single executable, no dependencies

🤖 Intelligence

• Sitemap discovery: Automatically finds and parses sitemap.xml (3 standard locations)
• robots.txt compliance: Respects crawling rules with per-domain caching
• Smart filtering: Auto-excludes images, CSS, JS, PDFs by default
• Domain auto-detection: Extracts and restricts to base domain automatically

🔒 Safety

• Rate limiting: Token bucket algorithm (default 2 req/s)
• Configurable timeout: 30 second default
• Memory safe: Rust's ownership system prevents crashes
• Graceful shutdown: 2-second grace period for pending requests

📊 Output

• Multiple formats: JSON, Markdown, HTML, CSV, Links, Text
• LLM-ready Markdown: Clean content with YAML frontmatter
• Interactive HTML report: Dashboard with graph visualization
• Stealth mode: User-agent rotation and realistic headers
• Content filtering: Remove nav, ads, scripts for clean data
• Real-time progress: Updates every 5 seconds during crawl

📝 Monitoring

• Structured logging: tracing with timestamps and log levels
• Progress tracking: [Progress] Pages: X/Y | Active jobs: Z | Errors: N
• Detailed statistics: Pages found, crawled, external links, errors, duration

Installation & Setup

Binary Location

~/.claude/skills/web-crawler/bin/rcrawler

Build from Source

# Clone the repository
git clone https://github.com/leobrival/rcrawler.git
cd rcrawler

# Build release binary
cargo build --release

# Copy to skill directory
cp target/release/rcrawler ~/.claude/skills/web-crawler/bin/

Build time: ~2 minutes Binary size: 5.4 MB

Command Line Interface

Basic Syntax

~/.claude/skills/web-crawler/bin/rcrawler <URL> [OPTIONS]

Options

Core Options:

• -w, --workers <N>: Number of concurrent workers (default: 20, range: 1-50)
• -d, --depth <N>: Maximum crawl depth (default: 2)
• -r, --rate <N>: Rate limit in requests/second (default: 2.0)

Configuration:

• -p, --profile <NAME>: Use predefined profile (fast/deep/gentle)
• --domain <DOMAIN>: Restrict to specific domain (auto-detected from URL)
• -o, --output <PATH>: Custom output directory (default:./output)

Features:

• -s, --sitemap: Enable/disable sitemap discovery (default: true)
• --stealth: Enable stealth mode with user-agent rotation
• --markdown: Convert HTML to LLM-ready Markdown with frontmatter
• --filter-content: Enable content filtering (remove nav, ads, scripts)
• --debug: Enable debug logging with detailed trace information
• --resume: Resume from checkpoint if available

Output:

• -f, --formats <LIST>: Output formats (json,markdown,html,csv,links,text)

Profiles

Fast Profile (Quick Mapping)

~/.claude/skills/web-crawler/bin/rcrawler <URL> -p fast

• Workers: 50
• Depth: 3
• Rate: 10 req/s
• Use case: Quick site structure overview

Deep Profile (Comprehensive Crawl)

~/.claude/skills/web-crawler/bin/rcrawler <URL> -p deep

• Workers: 20
• Depth: 10
• Rate: 3 req/s
• Use case: Complete site analysis

Gentle Profile (Server-Friendly)

~/.claude/skills/web-crawler/bin/rcrawler <URL> -p gentle

• Workers: 5
• Depth: 5
• Rate: 1 req/s
• Use case: Respecting server resources

Usage Examples

Example 1: Basic Crawl

~/.claude/skills/web-crawler/bin/rcrawler https://example.com

Output:

[2026-01-10T01:17:27Z] INFO Starting crawl of: https://example.com
[2026-01-10T01:17:27Z] INFO Config: 20 workers, depth 2
Fetching sitemap URLs...
[Progress] Pages: 50/120 | Active jobs: 15 | Errors: 0
[Progress] Pages: 100/180 | Active jobs: 8 | Errors: 0

Crawl complete!
Pages crawled: 150
Duration: 8542ms
Results saved to: ./output/results.json
HTML report: ./output/index.html

Example 2: Stealth Mode with Markdown Export

~/.claude/skills/web-crawler/bin/rcrawler https://docs.example.com \
  --stealth --markdown -f markdown -d 3

Use case: Content extraction for LLM/RAG pipelines Expected: Clean Markdown with frontmatter, anti-detection headers

Example 3: Fast Scan

~/.claude/skills/web-crawler/bin/rcrawler https://blog.example.com -p fast

Use case: Quick blog mapping Expected: 50 workers, depth 3, ~3-5 seconds for 100 pages

Example 4: Multi-Format Export

~/.claude/skills/web-crawler/bin/rcrawler https://example.com \
  -f json,markdown,csv,links -o ./export

Use case: Export data in multiple formats simultaneously Expected: Generates results.json, results.md, results.csv, results.txt

Example 5: Debug Mode

~/.claude/skills/web-crawler/bin/rcrawler https://example.com --debug

Output: Detailed trace logs for troubleshooting

Output Format

Directory Structure

./output/
├── results.json       # Structured crawl data
├── results.md         # LLM-ready Markdown (with --markdown)
├── results.html       # Interactive report
├── results.csv        # Spreadsheet format (with -f csv)
├── results.txt        # URL list (with -f links)
└── checkpoint.json    # Auto-saved state (every 30s)

JSON Structure (results.json)

{
  "stats": {
    "pages_found": 450,
    "pages_crawled": 450,
    "external_links": 23,
    "excluded_links": 89,
    "errors": 0,
    "start_time": "2026-01-10T01:00:00Z",
    "end_time": "2026-01-10T01:00:07Z",
    "duration": 7512
  },
  "results": [
    {
      "url": "https://example.com",
      "title": "Example Domain",
      "status_code": 200,
      "depth": 0,
      "links": ["https://example.com/page1", "..."],
      "crawled_at": "2026-01-10T01:00:01Z",
      "content_type": "text/html"
    }
  ]
}

HTML Report Features

• Interactive dashboard with key statistics
• Graph visualization using force-graph library
• Node sizing based on link count (logarithmic scale)
• Status color coding: Green (success), red (errors)
• Hover tooltips: In-degree and out-degree information
• Click to navigate: Opens page URL in new tab
• Light/dark mode: Auto-detection via CSS
• Collapsible sections: Reduces scroll for large crawls
• Mobile responsive: Works on all devices

Implementation Workflow

When a user requests a crawl, follow these steps:

1. Parse Request

Extract from user message:

• URL (required): Target website
• Workers (optional): Number of concurrent workers
• Depth (optional): Maximum crawl depth
• Rate (optional): Requests per second
• Profile (optional): fast/deep/gentle

2. Validate Input

• Check URL format (add https:// if missing)
• Validate workers range (1-50)
• Validate depth (1-10 recommended)
• Validate rate (0.1-20.0 recommended)

3. Build Command

~/.claude/skills/web-crawler/bin/rcrawler <URL> \
  -w <workers> \
  -d <depth> \
  -r <rate> \
  [--debug] \
  [-o <output>]

4. Execute Crawl

Use Bash tool to run the command:

~/.claude/skills/web-crawler/bin/rcrawler https://example.com -w 20 -d 2

5. Monitor Progress

Watch for progress updates in output:

• [Progress] Pages: X/Y | Active jobs: Z | Errors: N
• Updates appear every 5 seconds
• Shows real-time crawl status

6. Report Results

When crawl completes, inform user:

• Number of pages crawled
• Duration in seconds or minutes
• Path to results: ./output/results.json
• Path to HTML report: ./output/index.html
• Offer to open HTML report: open./output/index.html

Natural Language Parsing

Example User Requests

Request: "Crawl docs.example.com" Parse: URL = https://docs.example.com, use defaults Command: rcrawler https://docs.example.com

Request: "Quick scan of blog.example.com" Parse: URL = blog.example.com, profile = fast Command: rcrawler https://blog.example.com -p fast

Request: "Deep crawl of api-docs.example.com with 40 workers" Parse: URL = api-docs.example.com, workers = 40, depth = deep Command: rcrawler https://api-docs.example.com -w 40 -d 5

Request: "Crawl example.com carefully, don't overload their server" Parse: URL = example.com, profile = gentle Command: rcrawler https://example.com -p gentle

Request: "Map the structure of help.example.com" Parse: URL = help.example.com, depth = moderate Command: rcrawler https://help.example.com -d 3

Error Handling

Binary Not Found

# Check if binary exists
ls ~/.claude/skills/web-crawler/bin/rcrawler

# If missing, build it
cd ~/.claude/skills/web-crawler/scripts && cargo build --release

Crawl Failures

Network errors:

• Verify URL is accessible: curl -I <URL>
• Check if site is down or blocking crawlers
• Try with lower rate: -r 1

robots.txt blocking:

• Crawler respects robots.txt by default
• Check rules: curl <URL>/robots.txt
• Inform user of restrictions

Timeout errors:

• Increase timeout in code (default 30s)
• Reduce workers: -w 10
• Lower rate limit: -r 1

Too many errors:

• Enable debug mode: --debug
• Check specific failing URLs
• May need to exclude certain patterns

Performance Benchmarks

Test: adonisjs.com

• Pages: 450
• Duration: 7.5 seconds
• Throughput: 60 pages/sec
• Workers: 20
• Depth: 2

Test: rust-lang.org

• Pages: 16
• Duration: 3.9 seconds
• Workers: 10
• Depth: 1

Test: example.com

• Pages: 2
• Duration: 2.7 seconds
• Workers: 5
• Depth: 1

Technical Architecture

Core Components

1. CrawlEngine (src/crawler/engine.rs)

- Worker pool management - Job queue coordination - Shutdown signaling - Statistics tracking

1. RobotsChecker (src/crawler/robots.rs)

- Per-domain caching - Rule validation - Fallback on errors

1. RateLimiter (src/crawler/rate_limiter.rs)

- Token bucket algorithm - Configurable rate - Shared across workers

1. UrlFilter (src/utils/filters.rs)

- Regex-based filtering - Include/exclude patterns - Default exclusions

1. HtmlParser (src/parser/html.rs)

- CSS selector queries - Title extraction - Link discovery

1. SitemapParser (src/parser/sitemap.rs)

- XML parsing - Index traversal - URL extraction

Key Dependencies

• tokio: Async runtime (multi-threaded)
• reqwest: HTTP client (connection pooling)
• scraper: HTML parsing (CSS selectors)
• quick-xml: Sitemap parsing
• governor: Rate limiting (token bucket)
• tracing: Structured logging
• dashmap: Concurrent HashMap
• robotstxt: robots.txt compliance
• clap: CLI argument parsing
• serde/serde_json: Serialization

Tips & Best Practices

1. Start with Default Settings

First crawl should use defaults to understand site structure.

2. Use Profiles for Common Scenarios

• fast: Quick overviews
• deep: Comprehensive analysis
• gentle: Respectful crawling

3. Monitor Progress

Watch the [Progress] lines to ensure crawl is progressing.

4. Check HTML Report

Interactive visualization helps understand site structure better than JSON.

5. Respect Rate Limits

Default 2 req/s is safe for most sites. Increase cautiously.

6. Enable Debug for Issues

--debug flag provides detailed logs for troubleshooting.

7. Review robots.txt

Check <URL>/robots.txt to understand crawling restrictions.

8. Use Custom Output for Multiple Crawls

Avoid overwriting results with -o flag.

Future Enhancements (V2.0)

• Checkpoint resume: Full integration of checkpoint system
• Per-domain rate limiting: Different rates for different domains
• JavaScript rendering: chromiumoxide for dynamic sites
• Distributed crawling: Redis-based job queue
• Advanced analytics: SEO analysis, link quality scoring

Support & Resources

• GitHub Repository: leobrival/rcrawler
• Binary: ~/.claude/skills/web-crawler/bin/rcrawler
• Skill Documentation: This file (SKILL.md)
• Quick Start: README.md (this repository)
• Development Guide: DEVELOPMENT.md

Version: 1.0.0 Status: Production Ready