I wanted to read a couple feeds from their beginnings, something RSS doesn't particularly do and nobody particularly uses it for. But I've done the keep-a-tab-open-for-8-months thing more than once to work through an archive, and I don't fancy doing it again. As I worked through the steps and showed it to friends, we accumulated some other quality-of-life use cases that fit well enough into "just filter some XML" that they glommed on.

So, Sponder can:

- Run basic filtering on RSS feeds, either by keywords or regular expressions.

- Parse any webpage into an RSS feed, including autodetection of title/image/link/etc elements, following page links back through history, and coming back for new items later.

- Control the pace of that full historical feed to serve you an article per week, or 12 per day, whatever pace you like.

- Automatically detect and filter rerun episodes from a podcast feed.

- Be configured either by UI or typing some YAML.

It does not:

- Replace your RSS or podcast client, it's middleware that publishes a modified feed for you.

- Replace every one of your feeds, just the ones you wish were different. Though you can important and export OPMLs if you wish a lot were different.

- Run content through LLMs, though I'm considering it for rerun detection since metadata similarity only gets so far.

I'd love to hear from you fine folks:

- What bugs you about your feeds

- How configuring a flow goes

Comments URL: https://news.ycombinator.com/item?id=47387480

Points: 4

# Comments: 0

The SpotifyPlus HA integration [2] was near at hand and does a reasonably good job clustering with a version of ColorThief [3] under the hood. It has the same two problems you started with though: muddying when there's lots of gradation, even within a cluster; and no semantic understanding when the cover has something resembling a frame. A bit swapped from okmain's goal, but I can invert with the best of them and will give it a shot next time I fiddle. Thanks for posting!

[1] https://gist.github.com/kristjan/b305b83b0eb4455ee8455be108a... [2] https://github.com/thlucas1/homeassistantcomponent_spotifypl... [3] https://github.com/thlucas1/SpotifyWebApiPython/blob/master/...

The filtering was easy, but RSS doesn't do "from the beginning" (RFC 5005 exists, but is mostly unused), so scope crept into a webpage-to-RSS tool that lets me convert favorite.site/s/archive - autodetection of the article structure was a fun side quest.

The whole thing is a little function engine (Yahoo Pipes called), so the final goal is `merge(archive, live_feed) | drip(N items per D days)` to have the archive transition seamlessly into current content. I expect I can push that live tomorrow or so.

And of course Podcasts are just RSS, so hey, let's skip reruns. That's doable with filters on the episode description, but with history in place I'll add title similarity checking. I'm trying to think how to recognize cross-promoted episodes too, without having to crawl every podcast.

Importantly, Sponder's _not_ a client. There are enough clients, any many are great. Each implements some subset of features, so Sponder's an intermediary that consumes and publishes pure RSS for us to use anywhere we want.

Project two started over the weekend and is the NYTimes' Pips, but colors. You're building a stained glass window with regional constraints, and the big difference from using dominos is colors can mix. Also, triangles! The engine works, and I'm designing the tutorial and first handful of puzzles now.

Different RSS clients provide different filtering options, and lots of them limit you to a few keywords and/or put them behind a $7-12/mo subscription. I'm building Sponder so you can curate what you see, and it just presents another RSS feed, so you can keep using your favorite client but fill in the feature gaps.

Right now it can merge and filter by string or regex, and next I'm building (because it's what I want) history replay and smarter podcast rerun detection. it's new and I'm very open to feedback and feature requests.

https://en.wikipedia.org/wiki/Jon_Bentley_(computer_scientis...

    b=`bundle exec`, though this is becoming less often used and I may remove it
    d=dotenv to easily wrap a command in environment variables 
        (https://github.com/bkeepers/dotenv)
    g=git
    h=hub (GitHub's extensions to Git)
    j=autojump to move to common directories using just a few characters
        from their name (https://github.com/wting/autojump)

    alias g=git

On the other hand, once a month or so I hide from the office for an afternoon/night and replace some infrastructure or build an internal tool that's never going to get prioritized. Everyone's happy about it, and you get to feel like a bit of a hero :)

Comments URL: https://news.ycombinator.com/item?id=9257271

Points: 1

# Comments: 0

Disclosure: I work at Apartment List. Since someone already mentioned LiveLovely, I suppose I'll recognize them too :-D

I always liked the term not-just-for-profit.

Congrats on the merge, guys!

[Disclosure: I work at Singly]

Comments URL: https://news.ycombinator.com/item?id=4951742

Points: 2

# Comments: 0