spiderforce4ai 2.5

Python wrapper for SpiderForce4AI HTML-to-Markdown conversion service with LLM post-processing

SpiderForce4AI Python Wrapper

A Python package for web content crawling and HTML-to-Markdown conversion. Built for seamless integration with SpiderForce4AI service.

Features

• HTML to Markdown conversion
• Parallel and async crawling support
• Sitemap processing
• Custom content selection
• Automatic retry mechanism
• Detailed progress tracking
• Webhook notifications
• Customizable reporting

Installation

pip install spiderforce4ai

Quick Start

from spiderforce4ai import SpiderForce4AI, CrawlConfig
from pathlib import Path

# Initialize crawler
spider = SpiderForce4AI("http://localhost:3004")

# Configure crawling options
config = CrawlConfig(
    target_selector="article",
    remove_selectors=[".ads", ".navigation"],
    max_concurrent_requests=5,
    save_reports=True
)

# Crawl a sitemap
results = spider.crawl_sitemap_server_parallel("https://example.com/sitemap.xml", config)

Key Features

1. Smart Retry Mechanism

• Automatically retries failed URLs
• Monitors failure ratio to prevent server overload
• Detailed retry statistics and progress tracking
• Aborts retries if failure rate exceeds 20%

# Retry behavior is automatic
config = CrawlConfig(
    max_concurrent_requests=5,
    request_delay=1.0  # Delay between retries
)
results = spider.crawl_urls_async(urls, config)

2. Custom Webhook Integration

• Flexible payload formatting
• Custom headers support
• Variable substitution in templates

config = CrawlConfig(
    webhook_url="https://your-webhook.com",
    webhook_headers={
        "Authorization": "Bearer token",
        "X-Custom-Header": "value"
    },
    webhook_payload_template='''{
        "url": "{url}",
        "content": "{markdown}",
        "status": "{status}",
        "custom_field": "value"
    }'''
)

3. Flexible Report Generation

• Optional report saving
• Customizable report location
• Detailed success/failure statistics

config = CrawlConfig(
    save_reports=True,
    report_file=Path("custom_report.json"),
    output_dir=Path("content")
)

Crawling Methods

1. Single URL Processing

# Synchronous
result = spider.crawl_url("https://example.com", config)

# Asynchronous
async def crawl():
    result = await spider.crawl_url_async("https://example.com", config)

2. Multiple URLs

urls = ["https://example.com/page1", "https://example.com/page2"]

# Server-side parallel (recommended)
results = spider.crawl_urls_server_parallel(urls, config)

# Client-side parallel
results = spider.crawl_urls_parallel(urls, config)

# Asynchronous
async def crawl():
    results = await spider.crawl_urls_async(urls, config)

3. Sitemap Processing

# Server-side parallel (recommended)
results = spider.crawl_sitemap_server_parallel("https://example.com/sitemap.xml", config)

# Client-side parallel
results = spider.crawl_sitemap_parallel("https://example.com/sitemap.xml", config)

# Asynchronous
async def crawl():
    results = await spider.crawl_sitemap_async("https://example.com/sitemap.xml", config)

Configuration Options

config = CrawlConfig(
    # Content Selection
    target_selector="article",              # Target element to extract
    remove_selectors=[".ads", "#popup"],    # Elements to remove
    remove_selectors_regex=["modal-\\d+"],  # Regex patterns for removal
    
    # Processing
    max_concurrent_requests=5,              # Parallel processing limit
    request_delay=0.5,                      # Delay between requests
    timeout=30,                             # Request timeout
    
    # Output
    output_dir=Path("content"),             # Output directory
    save_reports=False,                     # Enable/disable report saving
    report_file=Path("report.json"),        # Report location
    
    # Webhook
    webhook_url="https://webhook.com",      # Webhook endpoint
    webhook_timeout=10,                     # Webhook timeout
    webhook_headers={                       # Custom headers
        "Authorization": "Bearer token"
    },
    webhook_payload_template='''            # Custom payload format
    {
        "url": "{url}",
        "content": "{markdown}",
        "status": "{status}",
        "error": "{error}",
        "time": "{timestamp}"
    }'''
)

Progress Tracking

The package provides detailed progress information:

Fetching sitemap from https://example.com/sitemap.xml...
Found 156 URLs in sitemap
[━━━━━━━━━━━━━━━━━━━━━━━━━━━━] 100% • 156/156 URLs

Retrying failed URLs: 18 (11.5% failed)
[━━━━━━━━━━━━━━━━━━━━━━━━━━━━] 100% • 18/18 retries

Crawling Summary:
Total URLs processed: 156
Initial failures: 18 (11.5%)
Final results:
  ✓ Successful: 150
  ✗ Failed: 6
Retry success rate: 12/18 (66.7%)

Output Structure

1. Directory Layout

content/                    # Output directory
├── example-com-page1.md   # Markdown files
├── example-com-page2.md
└── report.json            # Crawl report

2. Report Format

{
  "timestamp": "2025-02-15T10:30:00",
  "config": {
    "target_selector": "article",
    "remove_selectors": [".ads"]
  },
  "results": {
    "successful": [...],
    "failed": [...]
  },
  "summary": {
    "total": 156,
    "successful": 150,
    "failed": 6
  }
}

Performance Optimization

1. Server-side Parallel Processing

   • Recommended for most cases
   • Single HTTP request
   • Reduced network overhead
   • Built-in load balancing

2. Client-side Parallel Processing

   • Better control over processing
   • Customizable concurrency
   • Progress tracking per URL
   • Automatic retry handling

3. Asynchronous Processing

   • Ideal for async applications
   • Non-blocking operation
   • Real-time progress updates
   • Efficient resource usage

Error Handling

The package provides comprehensive error handling:

• Automatic retry for failed URLs
• Failure ratio monitoring
• Detailed error reporting
• Webhook error notifications
• Progress tracking during retries

Requirements

• Python 3.11+
• Running SpiderForce4AI service
• Internet connection

Dependencies

• aiohttp
• asyncio
• rich
• aiofiles
• httpx

License

MIT License

Credits

Created by Peter Tam