Find rss, atom, xml, and rdf feeds on webpages
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


Feed Seeker

It slant rhymes with "heat seeker"

Build Status Coverage

A library for finding atom, rss, rdf, and xml feeds from web pages. Produced at the mediacloud project. An incremental improvement over feedfinder2, which was itself based on feedfinder, written by Mark Pilgrim, and maintained by Aaron Swartz until his untimely death.


The library is available on PyPI:

pip install feed_seeker

The library requires Python 3.5+.


By default, the library uses requests to grab html and inspect it and find the most likely feed url:

from feed_seeker import find_feed_url

>>> find_feed_url('')

To do a more thorough search, use generate_feed_urls, which returns more likely candidates first.

from feed_seeker import generate_feed_urls

>>> for url in generate_feed_urls(''):
...     print(url)

For the most thorough search, add a spider argument to do depth-first spidering of urls on the same hostname. Note the below call takes nearly four minutes, compared to 0.5 seconds for find_feed_url.

>>> for url in generate_feed_urls('', spider=1):
...     print(url)

In a hurry?

If you have a long list of urls, you might want to set a timeout with max_time:

>>> for url in ('', ''):
   ...     try:
   ...         print('found feed:\t{}'.format(find_feed_url(url, max_time=3)))
   ...     except TimeoutError:
   ...         print('skipping {}'.format(url))
   found feed:

Differences with feedfinder2

The biggest difference is that all functions are implemented as generators, and are evaluated lazily. Candidate feed links are actually accessed and inspected to determine whether or not they are a feed, which can be quite time consuming. We expose a function to find the most likely feed link, and another to lazily generate links in rough order from most prominent to least.

There are also a few more heuristics based on our experience at mediacloud.