Jay Taylor's notes

back to listing index

Show HN: Beehive – An open source IFTTT powered by Go's templating engine | Hacker News

[web search]
Original source (news.ycombinator.com)
Tags: golang go beehive ifttt news.ycombinator.com
Clipped on: 2017-04-22

Image (Asset 1/2) alt=
Image (Asset 2/2) alt=
"Shortly" we had that topic here already: https://news.ycombinator.com/item?id=4441913 But most of the stuff is not alive anymore...

But there are some interesting new ones:



not sure if this is alive: https://sourceforge.net/projects/semanticwebpipe

Do you know some other (open source) event and agent system tools? (there are tons of commercial hosted IFTTT clones, they are not interesting...)

Node-RED seems similar and looks like it has a nicer interface (I've not tried it though).

Node-red also has a lot of connectors, from messaging eco systems to databases and is very flexible. If you hit a wall to the available node then you can always hand code Javascript in a function node.

The Raspbian Distro comes with Node-red pre-installed. Fastest and cheapest way to give it a try.

Project Flogo: http://flogo.io

There is/was Factor.io, but I'm not sure its still active.

I recently tried to use this kind of system to pipe news from facebook pages into an RSS feed. Impossible. Facebook hand-picks which developers get API access, and only allows ultra-restricted usage, so for example there's no way they would allow API access for a generic Beehive user, even just for reading public posts.

Want to scrape data from the HTML? They aggressively detect anything that may or may not be automated, and give you a CAPTCHA.

I see it as yet another abuse of their monopoly position. So many artists post public updates exclusively to this platform, and this is exploited to make sure everyone stays locked in and is forced to use facebook directly, ads included. Even the slightest competition would make it obvious how ridiculous this is, but the current monopoly is self-perpetuating.

Since you would self-host your own Beehive instance, there's luckily a way around that: create your own Facebook app for development (you will never publish it) and use its tokens to connect your Beehive with it.

That being said, there's no Facebook Hive yet (any takers?), but that's exactly how it works with the Twitter Hive.

Facebook doesnt hand pick anyone to its api.

just create a facebook app (no approval needed), have users sign in to that app to get their access token. then use the api posts endpoint to get all posts from any page. its that simple

OK, facebook hand-picks the authorization for an app (and its associated API access) to become public. They also hand-pick private access to more advanced API features. It's still essentially impossible to create an easy-to use program that reads public facebook posts, since each individual user has to create a developer account. I guess my first post was poorly phrased, but the main point remains that their policy forces the general public to use their front-end, and none other, to view any posts on public pages.

1. your users do not need to have a developer account. they just sign in via your app and you get their access token as long as they are a fb user.

2. i have over 10 fb apps and fb did not have to authorize any of them to be made public so i am not sure how you got that info. you just click a button to make the app public. can you clarify?

https://developers.facebook.com/docs/facebook-login/review/w... lists only "public_profile, user_friends and email" as scopes available without a review, accessing posts seems to require a different one.

do you want users to pipe news from public fb pages like cnn and nytimes? or feeds from private individuals/friends?

that permission only applies to the latter.

This is mostly wrong.

It's true they probably won't allow a generic Beehive user. But they will allow individual users to setup their own app accounts which will get access to public posts.

This seems to meet the above usecase exactly.

could you maybe run an android fb app in a vm then sniff the traffic?

Give our service at https://feedity.com a try for Facebook pages (or any other public webpage) into feeds. It's stable and scalable.

Quite interesting that this just been posted.

I have to produce an events application with a UI similar to this: http://i.imgur.com/e0R1Kp5.png

The UI is already done, so that's not the reason for the post.

My question is, on the backend, is there a pattern I can use to easily execute the steps?

I know there is a combination of logic, cron, a MQ and a db to hold the flow. But I am unsure on how best to approach the development of this framework.

Probably if someone is passing by and you've built something like this. I'd love your input as I'm a bit unsure how to proceed.

I'd really like best practice and implement something that is robust!


I think that you're probably looking for a finite state machine. There are a number of libraries for this, so have a look around and pick one that looks good. I haven't used it myself but would love to hear about other people's experiences with them.

Thank you for the "finite state machine" mention. I haven't seen this in my research so far. Will be adding this to the list to look up.

Worth perusing this for thoughts about doing an FSM in go: https://talks.golang.org/2011/lex.slide#1

Maybe you could get some inspiration from http://nodered.org. It's pretty generic despite the IoT mention.

would love to use it. the only people i know who has something similar are

- treeline.io which is backed by sailsjs.

- aws step functions. Used to coordinate lambda functions. check https://states-language.net/spec.html

> aws step functions...

By the way, AWS step functions allow the creation of state machines just like @rashkov mentioned in the adjacent comment. If you don't mind being locked in to AWS, using step machines with lambda functions would be a great way to do this.

that interface looks really useful, is there anyway I could get my hands on something like that. It'll be of even greater benefit if you can describe the process of building that interface. edit: nevermind I read the comments down the thread

Conversio.com offers abandoned cart recovery

which library did u use to make this flowcharty UI?

I needed a diagram to showcase my question. This isn't my UI.

However, I used https://jsplumbtoolkit.com to create my own UI.

What program did you use to create the diagram?

Sorry, you are taking me too literally there. This is just a screenshot.

So the imgur image you posted is just a random flow chart you found online? Is that correct?

When they said

>I have to produce an events application with a UI similar to this: http://i.imgur.com/e0R1Kp5.png

I read that as "I have to produce an events application with a UI similar to OP, here's what mine looks like". Probably this is what confused you as well.

Google tells me it's a screenshot of a marketing tool: https://www.getresponse.com/features/marketing-automation.ht...

I have been using huginn for more than one year and I am very satisfied. What's the difference between huginn and beehive?

huginn is a great project, by all means keep using it! The need for Beehive originally stemmed from our local hacker space, where we wanted to run a little daemon for 'home-automation' tasks, with the ability to control it via IRC. huginn was just too much of a beast for our Raspberry Pi and so we started hacking away on Beehive.

It's certainly a lot lighter, younger and less feature-complete than huginn - but it's also a lot easier to deploy and maintain in comparison. It's written in Golang, if that's your cup of tea, and really easy to extend.

If you ask nice, I bet you could get Ari Lerner to give you getbeehive.com.


I guess it can't harm to ask nicely. Thanks for the tip!

Documentation doesn't mention it, but this appears to not work on Windows - at least not out of the box. Annoying to go to the trouble of downloading it to try, to discover it was never going to work.

This actually seems to be a problem with certain test fixtures in one of the dependencies: (github.com/ChimeraCoder/anaconda) .

The offending files are:



This occurs because Windows can't have question marks in the file names.

I would suggest creating a pull request, or an issue on that repo asking that those files be renamed and the test logic be modified to not require the files to have the question marks (a simple character replacement in the required tests should probably do the job).

EDIT: There are some more issues! It uses github.com/guelfey/go.dbus which as far as I can tell will never really work on Windows.

The dbus-bee obviously won't run on Windows, but it's not a hard dependency of Beehive. You can simple comment out the import of any Bee in beehive.go.

There's a discussion to be had about how to enable/disable particular bees at build time.

Update: I've committed a fix that prevents building this bee (freedesktop.org notification support) on Windows/OSX.

Nice work :-) Hopefully I'll have some time over the next few days to do some testing. If I discover any more Windows based issues I'll send you a PR or create an issue.

Neat! I've been looking at a few things similar to this for various purposes recently, one thing I'd like is some sort of mix of this plus some content analysis system for managing an email inbox and then performing actions based on and using the data contained within, for example an email comes in with some sort of semi-structured data / information - it's contents is analysed, it's put in some sort of group like a 'this is an SNMP alert' then the data is formatted / munged into another form of structured / passable days and then routed off to another email address, I.e. Create a markdown / whatever table, Look for IP addresses and FQDNs, pair them if they're on the same line and create an extra column with a human-usable hyperlink, send it on to an email that logs a ticket in whatever ticket management system is in use at the receiving end. That's just one simple example I could make up off the top of my head, but I'd like to easily create some neat logic rules based on the content and then use some of that content to decide where it should be routed onto, again to use a ticketing / helpdesk situation as an example just because it's what I thought of just then - an email comes in and if it contains certain keywords or phrases a response will be sent back with a table / list of information to fill out in the form of questions like 'please include the following information in your reply...' and it might have some basic information gathering steps for the end user like getting their IP address, OS etc... then when they reply the system could pickup that data, munge / transform it into some sort of standard / structured format, do some more troubleshooting like pinging the IP provided, doing a whois, perhaps relating the data to a list of clients or something and then passing it on to the right team or ticketing system via email again. I think the key to this would be to use something like the OSS product mentioned here that's more modern and lightweight coupled with a nice query / rule builder interface that's fast and responsive and easy to test. Business rules / logic apps I've used in big corps in the past have been a combination, mix of, or all of the following: slow and resource hungry, proprietary with all forms of vendor lock-in, hard to build rules and apply them, require up to a whole team to support and maintain them, and often not technically focused / directed.

This looks amazing. Thank you for making it available as open source -- I can see how a vibrant community could ensure quality and up to date modules.

When I started learning Elixir/Phoenix this type of system was the first thing that came out of my mind when I was covering the Erlang/BEAM/OTP parts because of their nature.

If anyone else is interested in starting something similar in Elixir just ping me :)

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact