scrapeAI 0.3.0

A Python library to scrape web data using LLMs and Selenium

Scrape-AI

Scrape-AI is a Python library designed to intelligently scrape data from websites using a combination of LLMs (Large Language Models) and Selenium for dynamic web interactions. It allows you to configure the library to use a specific LLM (such as OpenAI, Anthropic, Azure OpenAI, etc.) and fetch data based on a user query from websites in real-time. The core functionality enables the agent-like scraping capabilities by leveraging natural language queries.

Key Features

• LLM Integration: Supports multiple LLM models (OpenAI, Anthropic, Azure, Google, etc.) through a flexible factory pattern.
• Dynamic Web Scraping: Utilizes Selenium WebDriver to interact with dynamic content on websites.
• Agent-Like Functionality: Acts as an intelligent agent that can process user queries and fetch relevant data from the web.
• Configurable: Customizable LLM model settings, verbosity, headless browsing, and target URLs.
• Modular Design: Structured in a modular way to extend scraping strategies and LLM integrations.

Installation

1. Clone the repository or install via pip (if available in PyPi):

pip install scrapeAI

2. Selenium WebDriver Dependencies: Selenium requires a browser driver to interact with the chosen browser. For example, if you're using Chrome, you need to install ChromeDriver. Make sure the driver is placed in your system's PATH (/usr/bin or /usr/local/bin). If this step is skipped, you'll encounter the following error:

selenium.common.exceptions.WebDriverException: Message: 'chromedriver' executable needs to be in PATH.

Here are links to some of the popular browser drivers:

• Chrome: Download ChromeDriver
• Edge: Download EdgeDriver
• Firefox: Download GeckoDriver
• Safari: WebDriver support in Safari

3. Set up your preferred LLM API keys: Ensure you have API keys ready for the LLM model you intend to use (e.g., OpenAI, Azure, Google, Anthropic).

Usage

Here is a basic usage example of how to use scrapeAI to scrape data from a website based on a user query:

from scrapeAI import WebScraper

config = {
    "llm": {
        "api_key": '<Azure OpenAI API KEY>',
        "model": "<Azure OpenAI Deplyement Name>",
        "api_version": "<Azure Open AI API Version>",
        "endpoint": '<Azure OpenAI Endpoint Name>'
    },
    "verbose": False,
    "headless": False,
    "url" : "https://pypi.org/search/?q=genai",
    "prompt" : "Provide all the libraries and their installation commands"
}

scraper = WebScraper(config)

# Invoke the scraping process
result = scraper.invoke()

# Output the result
print(result)

The output will be a json as the following:

[
  {
    'library': 'genai',
    'installation_command': 'pip install genai'
  },
  {
    'library': 'bookworm_genai',
    'installation_command': 'pip install bookworm_genai'
  },
  {
    'library': 'ada-genai',
    'installation_command': 'pip install ada-genai'
  },
  ...{
    'library': 'semantix-genai-serve',
    'installation_command': 'pip install semantix-genai-serve'
  },
  {
    'library': 'platform-gen-ai',
    'installation_command': 'pip install platform-gen-ai'
  },
  
]

Configuration Options

• llm: Defines the configuration for the LLM model, including the API key, model version, and endpoint.
• verbose: If set to True, enables detailed logging of operations.
• headless: If set to True, runs the web scraping in headless mode (without opening a browser window).
• url: The target URL for scraping.
• prompt: The natural language query to ask the LLM and fetch relevant content from the page.

Project Structure

---

The project is organized as follows: markdown Copy code

├── README.md
├── scrapeAI/
│ ├── **init**.py
│ ├── core/
│ │ ├── **init**.py
│ │ ├── base_scraper.py
│ │ ├── direct_scraper.py
│ │ ├── scraper_factory.py
│ │ └── search_scraper.py
│ ├── llms/
│ │ ├── **init**.py
│ │ ├── anthropic_llm.py
│ │ ├── azure_openai_llm.py
│ │ ├── base.py
│ │ ├── google_llm.py
│ │ ├── llm_factory.py
│ │ └── openai_llm.py
│ ├── utils/
│ │ ├── **init**.py
│ │ ├── html_utils.py
│ │ └── logging.py
│ └── web_scraper.py
├── setup.py
├── tests/
│ └── tests_operations.py

Core Components

• core/: Contains the base scraper classes and factory design patterns for scraping strategies.
• llms/: Includes different LLM integration classes such as OpenAI, Anthropic, and Google.
• utils/: Utility functions for HTML parsing and logging.

Contributing

---

We welcome contributions! If you'd like to improve the project, feel free to fork the repository and submit a pull request. Please follow the existing code structure and ensure that your contribution includes proper tests.

License

---

This project is licensed under the MIT License - see the LICENSE file for details.

Selenium Dependencies:

• Selenium requires browser drivers to interact with different browsers.
• The required driver for Chrome is ChromeDriver, and similarly, each browser has its respective driver.
• Make sure the driver is installed and placed in the system's PATH.

Popular Browser Drivers:

• Chrome: ChromeDriver
• Edge: EdgeDriver
• Firefox: GeckoDriver
• Safari: WebDriver for Safari