This is pretty cool, I drop the following shell script on all my servers:
#!/bin/bash
if [[ -z $1 && -z $2 ]]; then
echo "No Message passed"
else
if [[ -z $2 ]]; then
curl -s --form-string "token=MYAPPTOKEN" --form-string "user=MYUSERTOKEN" --form-string "message=$1" https://api.pushover.net/1/messages.json
else
curl -s --form-string "token=MYAPPTOKEN" --form-string "user=MYUSERTOKEN" --form-string "title=$1" --form-string "message=$2" https://api.pushover.net/1/messages.json
fi
fi
It's SUPER basic and probably shitty but for me it's perfect. I can add " && push 'Command is Done'" to the end of any command and get a notification on my watch/phone/(and desktop? But I don't have pushover on my desktop installed). Great for things you threw into a screen/tmux session and want to know when they finished.
No reason for this other then simply not having a request for it (until now) :).
After some quick Googling, it sort of appears everyone uses this protocol differently with different payloads (I could be really wrong here and would love if someone were to weigh in here). For example Google Talk servers require you to define additional XML which includes a token, where as other services just require you to send the sub/pub standard XML. These differences in payloads would make it difficult to build a simple URL with (while following the current convention).
So rather than support XMPP, maybe it would be better to create a notification per service you want to attach too? Did you have one in mind?
Currently, I use a simple Go client [1] and a self-hosted XMPP-server (ejabberd). In fact, I don't know what the Google servers require, but due to the federated nature of XMPP (similar to e-mail), I think it would be best to create a generic version rather than one for a specific service.
Awesome work, this is great! It seems like I have to write notification code for every little project I build.
One that'd be interesting would be to have an Apprise "server" where all config lives, and a client that receives commands. This way, a single Apprise installation could service multiple hosts without having to duplicate configuration files.
Hmm... I don't know if I'd want the headache of having everyone's passwords, tokens and API keys on a server I managed and deal with the privacy that goes with it...
But I was thinking of still supporting YAML or JSON formatted files to which users could protect with permissions and store their personal data there instead of being visible on the command line. the url would be something like file://path/to/config.yaml or http://localserver/path/to/config.yaml. So in a way, this kinda/sorta fits with your idea too :).
This is awesome! I work at a company where we have to abstract away from multiple services like this. We face some interesting challenges and are constantly learning from our experiences. I think you will have some similar challenges and would love to hear your/HN community's thoughts on how some of the issues would be tackled.
We are basically providing an API on top of multiple services which may work their own way and have some idiosyncrasies.
- I do not believe there is a specification/standard that is being followed by all libraries (i.e: Do not think something like OpenID Connect exists for these). Will we always be able to provide this API? What if we want to add a service that requires slight adjustments?
- How will we monitor and update our implementations when a single service is updated?
What you're describing you're building might be more complicated then the problems I faced.
For apprise, I focused on what was common across all notification services: all of them had a body,
most have a title, and some allow you to attach images. Anything outside of these 3 fields then became very specific to that one (notification) service. The sacrifice made is that if you want in-depth functionality of a specific service with rich features, you won't achieve it with apprise. It can however accomodate some additional attributes via arguments... For example with email you can define the smtp server, from address, etc., but that wouldn't be applicable to say Twitter.
I know I'm drifting from your question, but with respect to what was said:
- enforce your own specification as you're right, you'll find everyone solves great problems differently.
- use inheritance. Base classes make life soo much easier.
- decide what is common or what you want to make common across your API endpoints and stick with that. Like the body, title, image issue I resorted too. Maybe you want to assume you'll always need a user and password too. Prepare to support tokens and open ids out of the gate in your base class
- don't try to extend on one endpoint too much regardless of how much feature rich functionality it offers. Focus on getting them all endpoints working using the standards you come up with. This will help you achieve your goal. Build onto it after.
- unit tests! They're a pain to learn, but you won't regret knowing your code remains stable!
As per monitoring, I presume you're running a web service at the end of the day. So there are many great monitoring solutions out there that can check your servers availability constantly. Nagios (or forks of it) for example can email you when your service gets disrupted allowing you to take action right away. Heck, Nagios+apprise and you can get a text message. There are lots of more modern monitoring tools as well that can do all this for you.
Sorry for the wordy reply, I hope I answered your questions!
No, unfortunately it only simplifies the use of existing notification services.
I don't recall supporting a service that does that though. Perhaps I need a wrapper for Twilio? That service may at least allow text messages to the specific China phones you speak of... Not sure though.
I'll look into this. I actually added Facebook support, but it was around the time Zuckerberg got in trouble with data leaking. The API at the time became utterly useless (for Facebook).
But I never looked into the WhatsApp side of things. Thanks for the idea!
