Web scraping y ampliación de datos

AI agents often need up-to-date, specific information that isn't part of their pre-trained knowledge or available via simple API calls.

This is where web scraping comes in! It allows agents to "read" and extract data directly from web pages, just like a human would.

Web scraping is the automated process of extracting data from websites. Instead of manually copying information, a program does it for you.

For AI agents, this means they can dynamically gather facts, prices, news, or any other publicly available information from the internet to inform their decisions or responses.

Python has powerful libraries that make web scraping straightforward:

• requests: Used to make HTTP requests (like visiting a website) and get the raw HTML content.
• BeautifulSoup (often imported as bs4): A library for parsing HTML and XML documents, making it easy to extract data.

Together, they form a common duo for scraping tasks.

First, we use the requests library to fetch the content of a web page. This sends a request to the server and gets the HTML response.

Try running this example to see how we fetch the content from example.com:

Web pages are built with HTML (HyperText Markup Language). HTML uses tags like <h1>, <p>, <a> to structure content.

BeautifulSoup helps us navigate this structure. To extract data, we need to know what tags, classes, or IDs the information is wrapped in.

Once you have the raw HTML, BeautifulSoup turns it into a parse tree, allowing you to search for elements easily.

Here's how to create a BeautifulSoup object and extract a simple title from a string:

BeautifulSoup offers powerful methods like find() (for the first match) and find_all() (for all matches) to locate elements by tag name, attributes, or CSS selectors.

Let's find the text inside an <h1> tag:

Data augmentation, in this context, refers to the process of enhancing an agent's internal knowledge base or current context with newly acquired information.

When an agent scrapes a website, it's not just "reading" for itself; it's gathering data that can be used to improve its understanding, answer questions, or make more informed decisions.

Imagine an agent that needs to provide real-time stock prices or the latest news headlines. Its pre-trained model won't have this info.

• Real-time data: Scrape current stock prices or news.
• Specific facts: Extract product details from an e-commerce site.
• Contextual understanding: Get background info on a topic not covered in its training.

This scraped data can then be passed to the LLM as part of the prompt, augmenting its knowledge.

When scraping, always be mindful of:

• robots.txt: A file on websites that tells bots which parts of the site they shouldn't access. Respect it!
• Terms of Service: Many sites prohibit scraping in their terms.
• Rate Limiting: Don't bombard a server with too many requests too quickly; it can be seen as a DoS attack. Introduce delays.

Scrape responsibly and ethically.

You've learned about fetching web content and parsing HTML. Which Python library is primarily used for navigating and extracting data from HTML documents?

In this lesson, you learned how to use Python's requests and BeautifulSoup libraries for web scraping. We covered fetching raw HTML and then parsing it to extract specific information.

You also understood how this scraped data contributes to data augmentation, enabling AI agents to access current and specific information, thereby enriching their knowledge and capabilities dynamically. Always remember to scrape responsibly!