Agentic patters from scratch using Groq | Hacker News Reader

Agentic patters from scratch using Groq

github.com

120 pointsmtrofficus40 comments

> No LangChain, no LangGraph, no LlamaIndex, no CrewAI

Bless you. Using these over complicated abstractions (except CrewAI which I haven't yet checked out) never made sense to me. I understand that LLM is no magic wand and there is a need to make it systematic rather than slapping prompts everywhere. But these frameworks are not the solution to it. Next I will be looking at is Microsofts semantic-kernel. Anybody has any good words for it ?

Built a couple of things with Semantic Kernel. Both some private test projects, but also two customer facing applications and one internal.

It's heavily tilted towards OpenAI and it's offerings (either through OpenAI API or through Azure). However, it works decent enough for other alternatives as well, like: huggingface or ollama. Compared to the others (CrewAI etc). I kind of feel like Semantic Kernel hasn't really solved observe ability yet. Sure you can connect what ever logging/metric solution .Net supports, but it's not as seamless like the others. Semantic Kernel is available in .Net, Java and Python. But it's quite obvious .Net is a lot more polished then the others. Python usually gets new features faster, or at least pocs or previews.

Some learnings from it all:

- It's quite easy to get started with

- I like the distinction between native plugins and textbased ones (if a plugin should run code or not)

- There is a feeling of black magic in the background, in the sense of observe ability

- A bit more manual work to get things in order, compared to the alternatives

- Rapid development, it's quite clear the development team from Microsoft is doing a lot of work with this library

All and all, if you feel comfortable with writing C#, then Semantic Kernel is totally a viable option. If you prefer python over anything else, then I would say llamaindex or langchain is probably a better option (for now).

edit: updated some formatting

Thanks. I would have preferred to use Go instead of Python, but somehow the language is not picking up a lot in terms of new LLM frameworks.

As of now, I am using very light weight abstractions over prompts in python and that gets the job done. But, it is way too early and I can see how pipelining multiple LLM calls would need a good library that is not too complex and involved. In the end it is just a API call and you hope for the best result :)

Since you prefer go, you might be interested in one of my pet projects. Where I've glued together some libraries, which lets you basically code all interactions with LLM's through lua. The project is written in go.

Currently it only supports ollama, but I've been thinking about adding support for more providers

At which url can we check it out? Didn't find it on your profile and Github user with same username seems to be someone different.

Sorry about the late response. Thought I already had pushed it, you can find it here: https://github.com/asabla/gogent

As you can see, it's in a very early stage. I'm not a go developer, and I use this repository as a way to explore things both within Ollama and with go.

I'll probably add more things as the time goes by, but it isn't something I hack on every day or for that matter week. Just something to poke around and explore things with.

I'm currently building some stuff with SK using csharp.

Biggest gripes are that APIs have been changing quite a bit over time and the documentation isn't exhaustive in describing what is available.

If you're comfortable digging around in the source, looking at their tests and piecing things together it's pretty solid to build off of and is working well for me thus far.

I really like that there is abstraction enough for me to use other LLMs or implement them if need be.

I echo your apprehension around these abstractions and have built closer towards some of OP's patterns.

Check out dspy. It gets out of the way for the most part, while giving optional prompt optimization tooling that future-proofs your work.

I'm working on a very abstraction light package for LLMs called magentic. The goal is to handle the main boilerplate of building with LLMs, but leave the abstractions up to the user. I'd be interested to hear your opinion on it vs others.

https://github.com/jackmpcollins/magentic

Interesting! How does it compare to something like Mirascope? I feel like I'm constantly on the hunt for the slimmest llm interface layer, and mirascope is my most recent attempt.

I found `ell` pretty appealing, but I still prefer using the raw openai/anthropic packages.

I wanted to share a GitHub repository I've started a couple of months ago where I'm implementing, from scratch, the 4 agentic patterns as defined by Andrew Ng: Reflection pattern, Tool pattern, Planning pattern and Multiagent pattern.

You put more work into the README than you did the entire codebase.

The readme was the most important part for me!

glad you find it useful :)

peace man xD I mean, I didn't want to make a new framework, just try to offer some educational value. But sure, the code needs a refactor! :)

Here's the link to Andrew Ng's letters where he lays these out:

https://www.deeplearning.ai/the-batch/how-agents-can-improve...

This is the first letter, which is an introduction and ends with an index for the letters where he introduces four patterns.

Wow, thanks for sharing, that’s hilarious. Even in the one about “multi agent” systems theres no reference older than 2023.

I know I shouldn’t be shocked by how arrogant the connectionist got with their (arguably unexpected) success, but I can’t help it! They legit act like “AI” is a new phenomenon, which is especially funny for someone like Ng, who’s been an AI celebrity for at least a decade. No hate—his course was my first intro to real ML & AI, like I’m sure it was for many of us. Just a teeny bit of righteous condescension, I guess.

For anyone interested in this kind of stuff, this would be the super-popular first stop: Marvin Minsky’s Society of Mind

https://en.wikipedia.org/wiki/Society_of_Mind

https://courses.media.mit.edu/2016spring/mass63/wp-content/u...

The society of mind is an interesting reference. I remember browsing through it around the late 90's when it came out. It seemed to provide some theory for the basis of some of our cognitive functions in terms of a collection of cooperating agents. But then, I guess, what the agents themselves are made of was not clear/understood? Are today's LLM models capable of taking the form of those agents, and can we take inspiration from SoM to see how they can evolve together towards a more powerful (real/AG?) intelligence?

Is it me, or are the patterns somewhat tautological?

I think the Tool Use is. It would be more accurate and enlightening if it were clear that there were a single agent making the decision on which tool to use, rather than making it look like you're calling one of four tool agents (or if you ARE calling four agents, then the discriminating agent in front of them is being left out).

hey! I thought the diagram was clear, but I can see it's not. Thanks for the advise!! I'll try to change it so that it's clear we are using just one agent :)

I really like what you've done overall. Great job! Thank you!

typo in title: “patters” should be “patterns”.

ups, you're right. Sorry for that.

Why Groq?

edit: I'm dumb. Thought Groq was the Elon thing.

They offer a lot of top LLMs for free. That's the reason why. I mean, I'm using llama3.1 fine tuned for tools completely free. That's the reasoning behind haha But you could choose any other. At first I was thinking about Ollama, but Groq came as a great choice.

Not going to lie, I definitely confused Groq for Grok here.

Yep, Groq™ even called out Grok for this: https://groq.com/hey-elon-its-time-to-cease-de-grok/

Fast (suited for agents) and generous free tier (14k req/day for 70B models) would be my guess.

Much better explained than my own comment xD

Amazing diagrams.

Did you use https://excalidraw.com?

yep!!

That tool is the bees knees.

100% amazing

Reply on news.ycombinator.com