LocalLLaMA

4738 readers

5 users here now

Welcome to LocalLLaMA! Here we discuss running and developing machine learning models at home. Lets explore cutting edge open source neural network technology together.

Get support from the community! Ask questions, share prompts, discuss benchmarks, get hyped at the latest and greatest model releases! Enjoy talking about our awesome hobby.

As ambassadors of the self-hosting machine learning community, we strive to support each other and share our enthusiasm in a positive constructive way.

Rules:

Rule 1 - No harassment or personal character attacks of community members. I.E no namecalling, no generalizing entire groups of people that make up our community, no baseless personal insults.

Rule 2 - No comparing artificial intelligence/machine learning models to cryptocurrency. I.E no comparing the usefulness of models to that of NFTs, no comparing the resource usage required to train a model is anything close to maintaining a blockchain/ mining for crypto, no implying its just a fad/bubble that will leave people with nothing of value when it burst.

Rule 3 - No comparing artificial intelligence/machine learning to simple text prediction algorithms. I.E statements such as "llms are basically just simple text predictions like what your phone keyboard autocorrect uses, and they're still using the same algorithms since <over 10 years ago>.

Rule 4 - No implying that models are devoid of purpose or potential for enriching peoples lives.

founded 3 years ago

MODERATORS

pax@sh.itjust.works

noneabove1182@sh.itjust.works

Smokeydope@lemmy.world

MonsterBug@sh.itjust.works

Is there any good general AI-Agent /workflow platform which isn't vibe-coded? (palaver.p3x.de)

submitted 3 weeks ago* (last edited 3 weeks ago) by hendrik@palaver.p3x.de to c/localllama@sh.itjust.works

7 comments fedilink hide all child comments

I'm looking for something more like an traditional Free and Open Source project with an active community, different use-cases...

I tried googling it. But there's just way to many results these days. And they're mostly(?) cooked up by some AI agent and tend to get abandoned randomly after a few weeks. Or they have broad claims in a shiny README.md and then I install something and in reality it sucks and doesn't even do half of it. Or they're made by lunatics like Peter Steinberger who default to giving their agents root permissions on everything. That's why I try to avoid that category of projects.

I know I can code everything myself with Python, but it'd be great to have some workflows and integrations laid out for me, memory, RAG, a sandboxed Linux shell, cron, webhooks... So I can just go ahead and connect it to my local LLM and use it for various things. React to my my messages, look up information, read new pull-requests from a repository or RSS feed, write something to a homepage, pipe something into TTS or Ace-Step do a radio show or whatever. Make a small group of agents or my own tools...

Idk, something roughly alike n8n just proper open-source? Is there anything out there you other people use?

I'm asking in the LocalLlama community since I try to run it locally. And I need some amount of customizability so I can create some clever workflows. Something like OpenCode also doesn't really help if it wastes a million tokens on some mundane task and it's not really designed to fit with my limited amount of compute resources. Or if it's super hard to customize it to do so.

you are viewing a single comment's thread
view the rest of the comments

[–] MagicShel@lemmy.zip 5 points 3 weeks ago

You can use Cline with a local AI. It doesn't work great for enterprise level stuff because the number of tokens quickly swamps my MacBook, but qwen code can easily handle bash/ Python scripts and the like. Then you can use .clinerules to shape the agents but it's all mostly vibe from there on out

At work, we use a highly structured folder of agent prompts and infrastructure information with Claude. Because it's so highly structured it feels more like intentional code/config but I couldn't quantify any improvement metric at this time. There might be, but it would be premature to suggest you'd see any improvement over very be prompts.