HackerTrans
TopNewTrendsCommentsPastAskShowJobs

Koffiepoeder

no profile record

Submissions

Open source DAG for managing multi-agent pipelines

github.com
1 points·by Koffiepoeder·3 miesiące temu·1 comments

comments

Koffiepoeder
·30 dni temu·discuss
I once came across a similar "solution". The signing algorithm was directly executed from the update package. How would we otherwise be able to update the signature algorithm? Worst part was that it was correct at some point. It was an introduced regression because of a signature change due to " post-quantum safe" signatures now being required by the security team.
Koffiepoeder
·2 miesiące temu·discuss
Hey dipshit,

Maybe I develop games. Maybe I develop IoT devices. I might even be working in a high-stakes environment where formal verification is needed, who knows.

Whatever the case may be, we all have our reasons for choosing certain technologies. Not everyone is building run-of-the-mill 'backends' after all.

So please, let's stuff that neckbeardy arrogance away. It serves no purpose and distracts from the discussion.

Thanks.

P. s. I develop my backends in go.
Koffiepoeder
·2 miesiące temu·discuss
Thanks for the recommendation. This is very close to what I am looking for, at least with regards to the CLI.

The networking part I can fix with a second docker container and network_mode I think.

The centralised key and permission management and agent dashboarding is severly lacking though. But that's for now my least worry I think.
Koffiepoeder
·2 miesiące temu·discuss
Slightly related: I am looking for:

- Easy single command CLI agent spawning with templates

- Automatic context transfer (i. e. a bit like git worktrees)

- Fully containerised, but remote (a bit like pods)

- Central, mitm-proxy zero trust authn/authz management (no keys or credentials inside the agents), rather enrichment in the hypervisor/encapsulation

- Multi agent follow-up functionalities

- Fully self hosted/FOSS

Basically a very dev-friendly, secure, "kubernetes"-like solution for running remote agents.

Anyone has an idea of how to achieve this or potential technologies?
Koffiepoeder
·2 miesiące temu·discuss
Also this would not account for cooling shrinkage, a very annoying problem when making high quality parts to spec.
Koffiepoeder
·3 miesiące temu·discuss
I would be more concerned for remotely triggered inverter spikes tbh. These could sabotage the whole grid if I'm not mistaken.
Koffiepoeder
·3 miesiące temu·discuss
Reminds me of this (by now completely outdated) middle east friendship chart I once came across.

[0]: https://www.slate.com/blogs/the_world_/2014/07/17/the_middle...
Koffiepoeder
·3 miesiące temu·discuss
Working with agents and current-generation coding tools to me still feels relatively cumbersome. You have to do a lot of /manual /skill invocations and go through many review cycles before I consider code even close to workable, let alone commitable. I believe orchestrating this whole ordeal can probably be automated/improved through making good use of coding pipelines, thereby reducing manual intervention in the coding phase itself.

So, to minimise reviewing and AI oversight efforts, I was looking for a good way to orchestrate my local pi agent fleet. I found 'synapse' via this Reddit thread [0], in which the author (not me!) also asks for feedback and collaborators. The tool seems very interesting/promising to me!

Final note: I do _strongly_ believe that supervision of the coding process is still a _strict_ necessity. It's just that I think it can probably be more streamlined.

[0]: https://www.reddit.com/r/artificial/comments/1slu8m1/i_built...
Koffiepoeder
·3 miesiące temu·discuss
Rohansi was basically asking 'why', you keep on reiterating that DDR uses more power than LPDDR, but fail to answer why this is the case. Is it clock speed? Is it voltage? Is it a protocol/specification difference? 'various reasons' is not an answer.
Koffiepoeder
·3 miesiące temu·discuss
You did not answer the question.
Koffiepoeder
·3 miesiące temu·discuss
I understand trimming input fields is typically a useful default, but in this case this prevents me from searching for a space. So maybe it'd be worthwhile to add a `if (trim(str)=="") return str` exception or something similar?
Koffiepoeder
·3 miesiące temu·discuss
Imagine my face when I had to take periodic backups of stateless, immutable read-only filesystem, non-root containers for "compliance".
Koffiepoeder
·3 miesiące temu·discuss
To be honest, I would even go further: if you think certification equals security, you are even more lost.

So many controls are dubious, sometimes even actively harmful for some set-ups/situations.

And even moreso, it's also perfectly feasible to pass the gates with a burning pile of trash.
Koffiepoeder
·4 miesiące temu·discuss
Can confirm. Have encountered many on-prem and lift-and-shift solutions with no automated means of updating certs. The worst contenders are usually 1) executables on windows server (version 2012, of course), 2) old, obscure or very outdated database servers and 3) custom hardware firewalls. They are the worst.

To make things easy they usually all use different cert formats as well, requiring you to have an arsenal of conversion scripts ready.
Koffiepoeder
·4 miesiące temu·discuss
(Note that if OP considers to do this, they probably want to do this in a private tab, as to not leak potential sensitive cross-site cookies)
Koffiepoeder
·4 miesiące temu·discuss
We have an OCR job running with a lot of domain specific knowledge. After testing different models we have clear results that some prompts are more effective with some models, and also some general observations (eg, some prompts performed badly across all models).

Sample size was 1000 jobs per prompt/model. We run them once per month to detect regression as well.
Koffiepoeder
·5 miesięcy temu·discuss
Do you mean attacking a second country in the top 10 oil/gas reserve ranking in mere weeks, while threatening to invade a third?
Koffiepoeder
·5 miesięcy temu·discuss
The A3B part in the name stands for `Active 3B`, so for the inference jobs a core 3B is used in conjunction with another subpart of the model, based on the task (MoE, mixture of experts). If you use these models mostly for related/similar tasks, that means you can make do with a lot less than the 35B params in active RAM. These models are therefore also sometimes called sparse models.
Koffiepoeder
·6 miesięcy temu·discuss
Exactly there's so much stuff you simply cannot configure otherwise. For example disallowing applications to take sole ownership of a mic, in-detail power plans, etc. If they remove the old control panel, your machine basically becomes unconfigurable.
Koffiepoeder
·6 miesięcy temu·discuss
Had to search a bit, but here's a demo page: https://maplibre.org/maplibre-gl-js/docs/examples/display-a-... Can be compared with: https://maplibre.org/maplibre-gl-js/docs/examples/display-a-...

In that example I saw this in the console:

    before - 2.41+26.29+24.87+71.28+59.2+77.57 - 261.62kb
    after  - 2.45+22.4 +22.66+60.6+51.99+77.57 - 237.67kb
So roughly a ~10% compression improvement, neat!