convez.ai
AgenteDesenvolvimento WebIniciante
convergence-human-technology

Executive AI Core

convergence-human-technology/executive-ai-core

Agent IA educacional: entenda IA agentica em poucas linhas de código

Ver no GitHub
InicialPluginScore · de 0 a 100

Um agent IA minimalista em Python totalmente comentado, projetado para ensinar os fundamentos de IA agentica. Permite que iniciantes entendam como um agent autônomo planeja e executa tarefas, com menos de 90 linhas de código bem documentadas.

Instalação

$npx convez-ai convergence-human-technology/executive-ai-core

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
3
Forks
1
Acompanhando
0
Saúde
42%
Último commit
há 7 meses
Versão
—
Issues abertas
0
Contribuidores
2

Casos de uso

  • Aprender fundamentos de IA agentica de forma prática
  • Ensinar conceitos de autonomia e planejamento em IA
  • Base para desenvolver agents mais complexos

Veredito

Excelente recurso educacional que torna IA agentica acessível. Código limpo e bem comentado ideal para quem quer entender como agents funcionam, mas com escopo limitado para aplicações práticas reais.

README

A tradução para o português deste README ainda está na fila.

Convergence Cover

Executive Ai Core

A minimal, fully commented Python executive AI agent. For students, teachers and junior devs who want to understand agentic AI from the ground up.

Mini Executive AI – Understanding AI in a Few Lines of Code

Description :

This project consists of creating a simple but functional artificial intelligence, using the minimum number of lines of code possible. The goal is to make AI accessible to everyone: students, teachers, juniors, or anyone curious, even without technical expertise.

What you will find in this repository :

Code written in a language suited for pedagogy (Python).

Each line and each step are thoroughly commented to explain the role and functioning of each element.

A deliberately minimalist approach to focus on fundamental AI concepts (data, training, prediction, etc.).

Educational Goals :

Allow beginners to understand step by step how an AI works.

Demystify AI by showing that it can fit in very few lines of code.

Provide a simple and clear foundation, ideal for learning, teaching, or as a starting point for more advanced projects.

Target Audience :

Young people aged 18 and above, students, teachers, junior developers, anyone who wants to learn the basics of AI in a concrete and accessible way.

Useful Links :

The project is open source and available on GitHub for everyone to use, modify, and freely enhance.

This project in 20 sentences, for a non-technical 16-year-old. You know ChatGPT. You ask it a question, it answers. That's it. It does nothing more without you asking it again. This project does something different: you give it a goal, and it breaks it down into steps by itself, then it executes each step one by one, without you intervening. This is what is called an executive AI, or an autonomous agent. Imagine you tell someone "prepare me a presentation on global warming". A classic AI just writes text for you. An agent, on the other hand, will first write the list of things to do, then do them in order, and deliver the complete result to you. Concretely, this project is written in Python, a very readable programming language, almost like simple English. It uses the Claude API, meaning it connects to Anthropic's artificial intelligence via the internet, like an application that calls a server. We use Claude because it is the most advanced AI available via API right now, and because its behavior changes entirely depending on the instructions given to it at the start, which this project directly exploits. The code does exactly three things: plan, execute, display. To test, you install Python on your PC, you put your API key in an environment variable, you run the executive_ai.py file and you type any goal. In less than 30 seconds, you see the AI write its own plan then execute it step by step in your terminal. All the code fits in a single file of less than 90 lines, with a comment on each line to explain what is happening. There are also automated tests: these are small programs that verify that the code works correctly, without even needing a real API key. The project is public on GitHub, which means that anyone in the world can download it, read it, improve it, and propose their modifications. It is under the MIT license, which means that everyone can use it freely, even in commercial projects, without asking permission. The documentation is designed for beginners: each concept (token, API, environment variable, system prompt) is defined simply before appearing in the code. This project does not compete with major tools like LangChain or AutoGen. It is made to understand how these tools work, before using them. It is a bit like learning to make bread by hand before buying a bread machine. You understand what is really happening. It is deposited on the organization page of Convergence Human and Technology to concretely show what the company knows how to do and teach. Even a curious high school student with no programming experience can read this code from start to finish and understand how an AI makes decisions. And that is exactly the goal.

Ver o original no GitHub

Bom para

Relacionados