Past related threads: Huginn: Create agents that monitor and act on your behalf - https://news.ycombinator.com/item?id=21772610 - Dec 2019 (143 comments - beaten only by crows that day: https://news.ycombinator.com/front?day=2019-12-12 ) Huginn can now run for free on Heroku - https://news.ycombinator.com/item?id=8090772 - July 2014 (2 comments) Tell HN: There's two separate products on the front page, huginn and hug…
> beaten only by crows that day Very funny, since "Huginn" and "Mininn" were a mythological pair of ravens [0]. [0]: https://en.wikipedia.org/wiki/Huginn_and_Muninn
Huginn: System for building agents that perform automated tasks online
41–50 of 51 posts
Re: Huginn: System for building agents that perform automated tasks online
#42I ran a startup that I shut down a few years ago that was designed to do something very similar to this. It is dead now but the site is up at https://refinery.io (docs have screenshots) What I figured out is that... it's really hard to get this type of thing right. The more generic something is, the less useful it is. Zapier is a pretty amazing tool but it's also incredibly difficult to write all of their integration…
I thought of creating a similar tool just for marketing automation. I built a landing page [0], but haven't had the time to talk to potential customers.
Re: Huginn: System for building agents that perform automated tasks online
#43Earlier quoted context omitted.
I'm CEO of robusta.dev (an automations tool for Kubernetes) and this has been our experience too. People don't want the ability to write their own automations - they want something that just works out of the box. Very early on we realized this based on feedback and pivoted to focusing on out of the box alert insights (powered by automations we wrote) and not write your own automations. I don't know what the equivalen…
> People don't want the ability to write their own automations I don't think that is true. People do want their own automations, but only when necessary. And as always, it should be as effortless as possible. That's the main fail where most of those tools struggle. > they want something that just works out of the box. Why not both? Personally, I want something gives me everything for the common tasks and services, bu…
I should have elaborated more. We give you the useful content out of the box but you can customize to your heart's content and write your own automations.
All the APIs used to create the out of the box content are available for you too.
Re: Huginn: System for building agents that perform automated tasks online
#44I ran a startup that I shut down a few years ago that was designed to do something very similar to this. It is dead now but the site is up at https://refinery.io (docs have screenshots) What I figured out is that... it's really hard to get this type of thing right. The more generic something is, the less useful it is. Zapier is a pretty amazing tool but it's also incredibly difficult to write all of their integration…
Re: Huginn: System for building agents that perform automated tasks online
#45Earlier quoted context omitted.
> beaten only by crows that day Very funny, since "Huginn" and "Mininn" were a mythological pair of ravens [0]. [0]: https://en.wikipedia.org/wiki/Huginn_and_Muninn
There's also a ‘Munin’ software, of course: a server monitoring program. It was pretty neat around ten years ago, dunno how it compares now.
Re: Huginn: System for building agents that perform automated tasks online
#46An alternative to this (more similar to node red, actually) is n8n [1], which is pretty cool, but as others have said here, I never found a good application for this. [1]: https://n8n.io/
The time I've spent trying to filter seen data from a source is just mind-boggling. It would've been a lot faster to do the whole thing myself.
And I'm not alone in this. Pretty much every task that updates on a timer needs to exclude things it has already seen - and this is a herculean cryptic task in N8N.
Re: Huginn: System for building agents that perform automated tasks online
#47Earlier quoted context omitted.
In my experience, the biggest thing programmers do that normal people don't is imagine {all possible scenarios} and then design a system that's tolerant of them. If you can funnel things into a well-defined state before turning it over to non-programmers (e.g. data warehouse), you get better results.
> In my experience, the biggest thing programmers do that normal people don't is imagine {all possible scenarios} and then design a system that's tolerant of them. In my experience programmers don't do this any more than other people called on to design systems, and in fact frequently leverage the fact that software is updatable to avoid imagining failure scenarios until they occur, and only then making software resi…
Simple example: Go find a business analyst. Someone who's performing a process daily, over a large number of items. And has done so for 2 years+. Ask them to list all the different types of items (i.e. require a different or modified process).
In my experience, ~10% of them will be able to give you a comprehensive list. And only ~5% will bring up anything approaching exception handling for unknown cases.
Re: Huginn: System for building agents that perform automated tasks online
#48I ran a startup that I shut down a few years ago that was designed to do something very similar to this. It is dead now but the site is up at https://refinery.io (docs have screenshots) What I figured out is that... it's really hard to get this type of thing right. The more generic something is, the less useful it is. Zapier is a pretty amazing tool but it's also incredibly difficult to write all of their integration…
> The more generic something is, the less useful it is. Funny how different people reach different conclusions :) In my experience, tools and services that don't generalize enough (or allowing it rather), are less useful in the long-term and just solves something short-term. But it's easier to get traction if you solve something short-term, as it's often easier to market something that is more focused.
The angle of it being "generic" was that it required that you knew how to code. We didn't really give you many pre-defined "building blocks" to get started with. We tried to find a "niche" within that space (like a web scraper) but we just never got it quite right without needing to re-think everything we'd built.
Since then, we've ended up doing Open Source security tooling instead[0]. We wrote up the first technical blog post about the recent log4j vuln[1] and that's been valuable for helping us identify "niche" problems that companies struggle with.
Building a startup is hard work!
Re: Huginn: System for building agents that perform automated tasks online
#49I ran a startup that I shut down a few years ago that was designed to do something very similar to this. It is dead now but the site is up at https://refinery.io (docs have screenshots) What I figured out is that... it's really hard to get this type of thing right. The more generic something is, the less useful it is. Zapier is a pretty amazing tool but it's also incredibly difficult to write all of their integration…
Wow, Temporal is an interesting find. I thought of creating a similar tool just for marketing automation. I built a landing page [0], but haven't had the time to talk to potential customers. [0] https://h11g.com/
I threw you over an email. I'd be happy to chat and help share my lessons about this. :)
Re: Huginn: System for building agents that perform automated tasks online
#50This service runs as an interim before we can migrate to full containers and rely on their orchestration services. Sometimes I imagine SysOps as layers, this is a Layer 7, where as K8s is all of Layers 3-5.