Author here. Many thanks for the positive feedback!
Apparently the most requested feature is to lower the minimum score threshold. This restriction made more sense when the keywords filter was not available: A subscription with a score threshold of less than, say, 200 points (this number was fairly arbitrary), ended up in the user being literally spammed by the service -- While testing the service, I roughly estimated that a score of 100 points is equivalent to ~1.5 emails every 15 minutes. The usability, cost, and scalability of this approach is probably not ideal.
However, now that custom keywords are involved in the criteria, I do feel like the minimum score can be lowered. Perhaps even removed, if the user has selected 1+ keywords (they are optional). I'll follow up on this feature in the GitHub issue, https://github.com/ichinaski/hnnotifications/issues/2
I quite like the way this server is designed in a fashion that I can just run it myself - not requiring that I give my email address to any unknown third-party. I just took a look at the go code and its very easy to understand, very easy to build, and quite easy to deploy. Kudo's to the authors for making it open source - it means that I'm now running this locally as a service for my personal needs in finding new HN articles without much fuss! Great stuff!
Agreed. If I want the stories that score 200+ I'll just use the front page.
But when someone posted my company on hn it was on the front page for a few hours before I found out about it. Likewise, I want to know anytime someone posts something about me, my company, and a couple competitors. Those posts rarely (if ever) hit 200. I would probably use a threshold of about 10.
Not sure if that high of a threshold makes your tech possible, but it's a deal breaker for me.
Point counts vary by time. Weekends such as today see less traffic and upvotes.
The threshold you might really want is upvotes as a percentage of readership, but it's not clear whether that data is available or whether the functionality is trying to be in scope for this.
One small thing: once you know an email address subscribed you can mailbomb it with unsubscribe links (you might want to throttle those / don't send them more than once).
I actually wrote something like this a while back for myself. I find it a pain to keep track of sites like this and reddit where it's not a linear feed, items move up and down, and you need to scan the entire homepage to check for new content. In the end I made this http://www.serializer.io as a simple tool to make it a linear feed that's easier for me to follow.
Keywords support is currently quite naive. This filter definitely needs to be more sophisticated. While writing the feature I just wanted to make it as simple as possible, covering the vast majority of cases with a very straightforward policy (no punctuation marks, no symbols, spaces, etc). I agree this limitation needs to be improved.
Thanks! The keywords are searched in the story title. The current (somewhat naive) approach simply splits the title by non letter/number characters, essentially getting rid of punctuation chars, symbols, etc
I see. Trying to detect keywords in an article could be nice improvement.
Another question, why did you limit the lower bound for the score w/ 200 ?
I assume you start indexing the HN submissions. In the long run you can create a context aware version of HN, e.g. 'Bring HN submissions with keyword X submitted during the last week.' This could be a brand new way to browse HN.
Apparently the most requested feature is to lower the minimum score threshold. This restriction made more sense when the keywords filter was not available: A subscription with a score threshold of less than, say, 200 points (this number was fairly arbitrary), ended up in the user being literally spammed by the service -- While testing the service, I roughly estimated that a score of 100 points is equivalent to ~1.5 emails every 15 minutes. The usability, cost, and scalability of this approach is probably not ideal.
However, now that custom keywords are involved in the criteria, I do feel like the minimum score can be lowered. Perhaps even removed, if the user has selected 1+ keywords (they are optional). I'll follow up on this feature in the GitHub issue, https://github.com/ichinaski/hnnotifications/issues/2