- Who is this made for? I think this is crux of my issues so far here. I have been targeting developers who maybe don't see as much value in this as you mentioned.

That said I think that there is a decent cost to developers for maintaining code each time you want more determinism, or don't want to spend tokens on repeatable steps. My hope was that having a lower cost solution closer to markdown would be a nice middle ground.

I built Margarita to allow for markdown and logical operators to exist together, which means you can bring in determinism through code structures when it makes sense, and fall back to llm dynamic code when that makes sense. As an added bonus allows for composable prompts ala React which solve my other gripe with skills libraries being a mash of text everywhere.

Overall I've been getting pretty luke warm responses from Reddit, so I'll probably just shelve it, but it was a blast to make. Got to build code agents for pretty much every llm provider and built my own harness. I would recommend doing that it's a great learning experience.

https://www.margarita.run https://github.com/Banyango/margarita

Comments URL: https://news.ycombinator.com/item?id=48756596

Points: 7

# Comments: 4

Let's say you wanted to use Gemma4B to run your workflow, however that model doesn't always do what you want it to. You could write the Margarita code to do the variable tracking, loops, etc and use the llm call where Gemma is able to work better. It's about finding the sweet spot of code vs llm agentic/tool calling.

Can you write that in js/go/etc? sure, but this lets you write it in a much simpler syntax that's much closer to markdown.

There's a few example on the website: https://www.margarita.run/custom-ralph.html https://www.margarita.run/components.html

Why did I build it? At work I've seen two major problems with our ai workflows/ skills libraries. There is a lack of determinism when your whole workflow is a markdown file of 100 steps, and markdown skill libraries lack composability. Meaning we violate things like DRY in the all the md files in the skills library.

I built Margarita to allow for markdown and logical operators to exist together, which means you can bring in determinism through code structures when it makes sense, and fall back to llm dynamic code when that makes sense. As an added bonus allows for composable prompts ala React which solve my other gripe with skills libraries being a mash of text everywhere.

Overall I've been getting pretty luke warm responses from Reddit, so I'll probably just shelve it, but it was a blast to make. Got to build code agents for pretty much every llm provider and built my own harness. I would recommend doing that it's a great learning experience.

https://www.margarita.run https://github.com/Banyango/margarita

Github: https://github.com/Banyango/margarita

I've always wanted to write a programming language, and AI made that a bit easier. So I tackled a kind of quasi-extension to markdown that lets you control coding agents easier.

Margarita provides two file formats:

.mg — A templating language that renders dynamic prompts to Markdown. Extends Markdown with variables, conditionals, loops, and includes.

.mgx — An agent scripting language that extends .mg with agentic execution: state, memory, tool calls, user input, and more.

That said I'm not sure how useful this thing is since everyone just uses claude code now, but it was a really great learning experience.