mtrofficus

Hey!

I'm excited to share an open-source course I've created that walks you through building an Agentic Call Center from the ground up.

The course is divided into five lessons covering core architecture, FastRTC for WebRTC and WebSocket connections, plus integrations with Twilio, Superlinked, Qdrant, and more.

Hope you find it useful!

Tired of tutorials that just walk you through connecting an existing MCP server to Claude Desktop?

Yeah, us too.

That's why we built Kubrick AI, an MCP Multimodal Agent for video processing tasks.

Yes! You read that right:

Agents + Video Processing ... and MCP!

Hey everyone!

Excited to share one of my latest projects.I'm recreating famous philosophers and historical thinkers inside a video game simulation. It's been wild bringing these iconic minds to life in a whole new way.

Hope you like it!

Hey guys!

I wanted to share the open-source course we've built, where you'll learn how to build an AI Companion using these technologies:

1. LangGraph (defining all the agentic workflow) 2. Qdrant (long-term memory) 3. Groq (as LLM provider, so free :)) 4. ElevenLabs (for the TTS module) 5. Whisper (for the STT module) 6. FLUX models (for image generation)

Let me know what you think!!

Hi there!

Just wanted to share the open source course we are launching in GitHub, YouTube and Substack. It's about building a multimodal agent connected to WhatsApp taking inspiration from the movie Ex Machina (that's why it's called Ava xD).

The course is structured in 6 lessons, that we are launching every Wednesday (of course, everything is 100% free haha).

Let me know what you think!!

This project shows you how to make Llama 3.1 8B speak like Rick Sanchez by:

1. Creating a custom dataset from Rick and Morty transcripts in ShareGPT format

2. Finetuning the model using Unsloth's optimizations on Lambda Labs GPUs

3. Converting and deploying the model to Ollama for local use

It's a fun way to learn LLM finetuning while creating your own Rick-speaking AI assistant.

Wubba lubba dub dub!

A YouTube playlist containing an implementation of the 4 agentic patterns using pure Python and Groq LLMs. You have the Reflection Pattern, the Tool Pattern and the Planning Pattern already available ... working on the Multiagent Pattern right now :)

The code for all the videos is available in this repo: https://github.com/neural-maze/agentic_patterns

Hope you like it!!

thanks!! glad you liked it :)

“on a site” referring to Medium xD. Welcome to 2024.

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 :)

In the Medium article I'll show you how I programmed some simple Sketch scripts and uploaded to my Arduino UNO using crewAI Agents.

This is a funny project I made one evening: applying Image Search in Qdrant to play Mario Kart 64 using mupen64plus.

ups, you're right. Sorry for that.

Much better explained than my own comment xD

100% amazing

That's from Groq documentation:

"In fact, our name comes from the word “grok” which means the ability “to understand profoundly and intuitively.”"

hahahha wow, that's hard: "pseudointellectual names". I mean, didn't reflect about the naming. Just the fact that it offered a lot of LLMs for free xD

I'm using Groq, not Grok xD

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.

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! :)

yep!!

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.

hey! I've started a repository here: https://github.com/neural-maze/crewai_llama3_arduino

I'll be refactoring it during this week and then I'll create an article in my blog and a video in my Youtube channel talking about it, in case you are interested :)

https://www.youtube.com/channel/UCaixkLsW_TWWe_0yZNmvPaw

https://neural-maze.github.io/blog/

hahahah, you can try it locally if you want. Check the code:

https://github.com/neural-maze/talking_with_hn

But I really think you'd prefer something like an AI girlfriend. You need love bro.

Yes, what I mean by that comment is that this interface let's you access all of this information in a chat-based fashion, You don't need to change the tab (from top to new for example), or create some manual filtering. You just ask the chatbot to perform this operations from you (and you can ask for any number of stories you want, sorted by score, etc.).

Nest steps I have in mind (would like to know your opinion about them):

1. Access comments informations (e.g. asking the chatbot to give you a summary of the sentiment analysis on the comments for a given story)

2. Access content information (e.g. suppose one story links to a news article in Xataka for example, it would be really cool for the chatbot to access the content of the webpage and being to answer questions about it)

Any thoughts?

Next improvements for the app:

1. Being able to access the comments and ask the bot about them (e.g. give me a summary about the comments, some kind of sentiment analysis, etc.)

2. Being able to summarise the content in the URLs (this is only applicable blog posts, news articles etc, but I think it could be cool for the chatbot to parse the blog post using bs2 and being able to reason and answer questions about it)

good advice

Hey! I followed some of the examples in the Streamlit gallery. I agree it looks like some phising page hahaha when I was trying some of the examples I used a openai key that I deleted after I was finished xD

A Chatbot that let's you chat with HackerNews using Langchain and Streamlig.

Repo: https://github.com/neural-maze/talking_with_hn