Skip to content

scrapegraphaivskatana

MIT 4 17 23,278
59.6 thousand (month) Jan 15 2024 1.76.0(2026-04-09 09:41:03 ago)
16,499 6 18 MIT
Nov 07 2022 v1.5.0(2026-03-10 14:52:47 ago)

ScrapeGraphAI is a Python library that uses large language models (LLMs) to create web scraping pipelines automatically. Instead of writing CSS selectors or XPath expressions, you describe what data you want in natural language and provide a Pydantic schema — the library handles the rest.

Key features include:

  • Natural language extraction Describe what you want to extract in plain English (e.g., "Extract all product names and prices") and the LLM figures out how to find and extract the data.
  • Pydantic schema output Define the expected output structure using Pydantic models for type-safe, validated extraction results.
  • Graph-based pipeline Built on a directed graph architecture where each node performs a specific task (fetching, parsing, extracting, merging). This makes pipelines modular and debuggable.
  • Multiple graph types SmartScraperGraph (single page), SearchGraph (search + scrape), SpeechGraph (audio output), and more specialized pipelines.
  • Multiple LLM providers Works with OpenAI, Anthropic, Google, Groq, local models via Ollama, and more.
  • HTML and JSON support Can extract data from both HTML pages and JSON API responses.

ScrapeGraphAI is particularly useful for rapid prototyping of scrapers and for extracting data from pages with complex or frequently changing layouts where traditional selectors would be brittle.

Katana is a next-generation web crawling and spidering framework written in Go by ProjectDiscovery. It is designed for fast, comprehensive endpoint and asset discovery and is widely used in the security research and bug bounty communities.

Katana offers multiple crawling modes:

  • Standard mode Fast HTTP-based crawling without a browser. Parses HTML, JavaScript files, and other resources to discover endpoints and links.
  • Headless mode Uses a headless Chrome browser for crawling JavaScript-rendered pages and single-page applications (SPAs).
  • Passive mode Discovers URLs from external sources (Wayback Machine, CommonCrawl, etc.) without actively visiting the target.

Key features include:

  • Scope control Configurable crawl scope with regex patterns for including/excluding URLs, domains, and file extensions.
  • JavaScript parsing Extracts endpoints from JavaScript files, inline scripts, and AJAX requests even in standard (non-headless) mode.
  • Customizable output Filter and format output with field selection, JSON output, and custom templates.
  • Rate limiting Built-in rate limiting and concurrency control to avoid overwhelming targets.
  • Proxy support HTTP and SOCKS5 proxy support with rotation.
  • Form filling Can detect and auto-fill forms to discover endpoints behind form submissions.

While Katana was designed for security research and reconnaissance, its fast crawling capabilities and JavaScript parsing make it equally useful for web scraping discovery and sitemap generation.

Highlights


ai-poweredpopular
fastpopularlarge-scale

Example Use


```python from scrapegraphai.graphs import SmartScraperGraph from pydantic import BaseModel, Field from typing import List # Define the output schema class Product(BaseModel): name: str = Field(description="Product name") price: float = Field(description="Price in USD") rating: float = Field(description="Customer rating out of 5") class ProductList(BaseModel): products: List[Product] # Create a scraping graph with natural language instruction graph = SmartScraperGraph( prompt="Extract all products with their names, prices, and ratings", source="https://example.com/products", schema=ProductList, config={ "llm": { "model": "openai/gpt-4o", "api_key": "YOUR_API_KEY", }, }, ) # Run the graph result = graph.run() for product in result["products"]: print(f"{product['name']}: ${product['price']} ({product['rating']}/5)") ```
```go package main import ( "context" "math" "github.com/projectdiscovery/katana/pkg/engine/standard" "github.com/projectdiscovery/katana/pkg/output" "github.com/projectdiscovery/katana/pkg/types" ) func main() { // Configure crawl options options := &types.Options{ MaxDepth: 3, FieldScope: "rdn", // restrict to root domain BodyReadSize: math.MaxInt, Timeout: 10, Concurrency: 10, Parallelism: 10, Delay: 0, RateLimit: 150, Strategy: "depth-first", OnResult: func(result output.Result) { // Process each discovered URL println(result.Request.URL) }, } // Create and run the crawler crawlerOptions, _ := types.NewCrawlerOptions(options) defer crawlerOptions.Close() crawler, _ := standard.New(crawlerOptions) defer crawler.Close() // Start crawling _ = crawler.Crawl("https://example.com") } ```

Alternatives / Similar


Was this page helpful?