tools/read_rss_feed.py

from smolagents.tools import Tool
from markdownify import MarkdownConverter

class CustomMarkdownConverter(MarkdownConverter):
    """
    Create a custom Markdown converter that ignores images and
    ignores links, keeping only their text.
    """
    def convert_img(self, el, text, parent_tags):
        return ""
    def convert_a(self, el, text, parent_tags):
        return text

def markdown_summary(html, article_limit):
    rendered = CustomMarkdownConverter().convert(html).strip()
    if len(rendered) <= article_limit:
        return rendered
    else:
        return rendered[:article_limit] + '...'

class ReadRssFeedTool(Tool):
    name = "read_rss_feed"
    description = "Read the articles from an RSS feed and return them in a single text with markdown format."
    inputs = {
        'url': {'type': 'string', 'description': 'The url of the RSS feed (example: https://www.eldiario.es/rss).'},
        'article_limit': {'type': 'integer', 'nullable': True, 'default': 512, 'description': 'Number of characters that will be considered for each article'},
        'feed_limit': {'type': 'integer', 'nullable': True, 'default': 4096, 'description': 'Number of characters that will be considered for the whole article'},
    }
    output_type = "string"

    def forward(self, url: str, article_limit: int=512, feed_limit: int=4096) -> str:
        """Read all the articles from a given RSS feed and return them as a single markdown formatted text."""
        import re
        from requests import get
        from rss_parser import RSSParser
        from markdownify import markdownify
        response = get(url)
        rss = RSSParser.parse(response.text)
        articles = "# Articles\n\n"
        for item in rss.channel.items:
            articles += "## " + item.title.content + "\n"
            articles += item.pub_date.content + "\n\n"
            content = markdown_summary(item.description.content, article_limit)
            articles += content + "\n\n"
        return articles[:feed_limit]