convez.ai
PluginEscrita & ConteúdoIntermediário
josephrmartinez

mktute - Gerador de Tutoriais a partir de Diffs

josephrmartinez/mktute

Converta seus commits em tutoriais Markdown com IA

Ver no GitHub
InicialPluginScore · de 0 a 100

CLI que transforma commits recentes do Git em tutoriais Markdown passo a passo usando IA. Seleciona diffs entre commits e gera documentação automática com Claude, GPT-4 ou modelos locais via Ollama. Ideal para documentar trabalho real e compartilhar conhecimento.

Instalação

$npx convez-ai josephrmartinez/mktute

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
1
Saúde
28%
Último commit
há 2 anos
Versão
—
Issues abertas
0
Contribuidores
2

Casos de uso

  • Auto-documentação de mudanças de código
  • Criar tutoriais internos para desenvolvedores
  • Publicar aprendizado e conhecimento adquirido
  • Documentar implementações diferentes das sugestões de LLMs

Veredito

Ferramenta prática que resolve um problema real: transformar código que você já escreveu em documentação clara. A integração com múltiplos modelos (local, Claude, GPT-4) oferece flexibilidade, mas exige instalação de dependências e API keys para versões cloud.

README

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

CLI App for Generating Markdown Tutorials Based on Recent Diffs

Turn your recent git commits into a step-by-step programming tutorial:

  1. Run mktute, select start and end commits.
  2. Select whether you would like to use a local AI model, Anthropic's Claude 3.5 Sonnet, or OpenAI's GPT-4-turbo to generate a step-by-step tutorial covering the code you recently committed.
  3. The tutorial is automatically saved as a markdown file in the current working directory.

Use cases:

  • Self-documentation
  • Drafting internal tutorials for other developers
  • Learning in public
  • Publish learning residue

But, why?

  • Knowledge Sharing: You figured out something while coding that would be highly useful to share with other developers. mktute helps you quickly document your work in a user-friendly format. Publishing tutorials based on your actual work is one way that you can learn generously.

  • Fill in Knowledge Gaps: You're working with a framework that has changed significantly from the last training date of a popular LLM. Or your specific implementation is different from what the LLMs often suggest. Publishing a totorial based on your actual work can help with providing learning resources where LLMs are falling short.

  • Leverage LLM Strengths: While LLMs may struggle with generating code from scratch, they are generally very good at summarizing and explaining existing code. By feeding in your actual code changes, mktute leverages a real strength of LLMs to produce high-quality, step-by-step tutorials.

Installation

Install mktute globally:

sudo npm install -g mktute

Or install mktute in the current working directory:

npm install mktute

Usage

simply type mktute in the command line to run the interactive command line user interface. You will have control over which diffs to send and which model you use for inference.

Usage with Local Model

  1. Download the Ollama application: https://ollama.com/download

  2. Once installation is complete, open a terminal and run ollama run llama3

The first time your type this command, you will pull the llama3 7b model. The model is about 4.7 GB and will take a few minutes to download. mktute currently only uses the llama3 7b model.

  1. Close the terminal you used to run ollama run llama3 but keep the Ollama application running.

  2. Run mktute and select "LOCAL - Ollama Llama 3" when prompted to select an AI provider. When you load the model for the first time, it may take a few seconds to start up. Please keep in mind that local interference performance is highly dependent on your machine's capabilities. You can run llama3 on a consumer grade laptop, but inference may be slow.

When using a local model, the output will print to stdout and be saved in a markdown file in the current working directory.

You do not need to be connected to the internet, but you must be running the Ollama application in the background in order to use the local model option with mktute.

Example output and cost

The following tutorial was generated with GPT-4-turbo at a cost of $0.0235: How to Center a <div> Using TailwindCSS

When you run mktute, you will be presented with an option to choose which model you would like to use to generate the tutorial. A price estimation is included with these model selection choices.

You can also choose to run mktute with a local model to avoid any costs.

A Note on "Slop"

Please give consideration before publishing tutorials generated with mktute. While mktute lowers the barrier to writing a coding tutorial, you should absolutely spend time thoroughly editing the generated content before publishing.

Ver o original no GitHub

Bom para

Relacionados