Forem: Daniel (San) Ávila

Complete Guide to Claude Code Templates

Daniel (San) Ávila — Wed, 27 Aug 2025 20:56:31 +0000

What is Claude Code Templates

Claude Code Templates is an open-source project that simplifies Claude Code adoption through ready-to-use configurations.

Visit aitmpl.com to browse over 400 components including agents, commands, settings, hooks, MCPs, and templates.

Claude Code Templates - Supercharge Your AI Development with Anthropic Claude

*Professional templates and configurations for Anthropic's Claude Code. Get Claude Opus 4.1 working at terminal velocity…*aitmpl.com

The core insight behind this project: how you configure your AI development agents is often more critical to project success than the actual code they generate.

Currently hosting over 400 components with thousands of downloads, Claude Code Templates has become the de facto package manager for the Claude Code ecosystem.

Component

Claude Code Templates organizes components into six categories:

Agents

Specialized AI personas with specific expertise. Examples include frontend developers, security auditors, and legal reviewers. Each agent comes with optimized system prompts and tool configurations.

Commands

Reusable command configurations for common development tasks like code review, testing, and deployment checks.

Settings

Pre-configured permission sets, tool allowlists, and environment configurations that ensure secure and efficient AI interactions.

Hooks

Integration points for CI/CD pipelines, git workflows, and automated development processes.

MCPs

Model Context Protocol integrations that connect Claude Code to external services like databases, APIs, and development tools.

Templates

Complete project setups combining multiple components for specific use cases or technology stacks.

Using the Search Function

The search interface allows you to filter through hundreds of components efficiently. Enter keywords related to your technology stack, development role, or specific needs.

For example:

Search “react” to find frontend development components
Search “security” for auditing and security review tools
Search “game” for game development resources
Search “api” for backend and API development components

You can also browse by category using the filter tabs (Agents, Commands, Settings, Hooks, MCPs, Templates) or explore predefined categories like “Development Team,” “AI Specialists,” or “Business Marketing.”

Installing Individual Components

Each component can be installed independently using the provided command. Click the “Install” button next to any component to copy its installation command.

Individual installation commands follow this pattern:

npx claude-code-templates@latest --agent=development-team/frontend-developer
npx claude-code-templates@latest --command=code-optimization/bundle-analyzer  
npx claude-code-templates@latest --setting=security/enterprise-permissions

The --yes flag can be added to skip confirmation prompts:

npx claude-code-templates@latest --agent=development-team/frontend-developer --yes

Building Complete Stacks with Cart Functionality

Rather than installing components one by one, you can build complete development stacks using the cart system. This approach is more efficient for setting up comprehensive development environments.

To build a stack:

Browse components and click “Add to Cart” for each item you need

The cart accumulates your selections across all component types

When ready, click “Generated Command” to create a single command that installs everything

Copy and run the generated command in your terminal

You also can share this command on X or Threads

A typical full-stack development setup might include:

Frontend developer agent
Code reviewer agent
Security auditor agent
Performance optimization commands
Git integration hooks
Development environment settings
Relevant MCP integrations

This results in a single command that configures your entire Claude Code environment at once.

Discovering Popular Components with Trending

The trending section shows which components the community uses most, updated weekly. This data helps identify battle-tested components and emerging patterns in how developers configure their AI workflows.

Conclusion

Claude Code Templates transforms Claude Code from a powerful but complex tool into an accessible development platform. By providing curated, tested configurations, it removes the barrier between developers and productive AI-assisted coding.

AI agent configuration has become as important as the code itself. A thoughtfully configured development environment with the right agents, permissions, and integrations can dramatically improve productivity and code quality. Claude Code Templates makes this sophistication accessible to every developer.

For advanced usage including Global Agents that run directly from your terminal, check out this step-by-step tutorial.

Contribute to the project at github.com/davila7/claude-code-templates.

You can follow me on my social media:

LinkedIn (Spanish): https://www.linkedin.com/in/daniel-avila-arias/
Twitter (English): https://x.com/dani_avila7
Youtube (Spanish): https://www.youtube.com/@daniiielsan

Claude Code: From Zero to Hero

Daniel (San) Ávila — Tue, 26 Aug 2025 18:10:16 +0000

What is Claude Code?

Claude Code is an AI-powered coding assistant that lives in your terminal, understanding your codebase and accelerating development through natural language. It integrates directly into your workflow, providing a flexible and safe way to leverage AI for coding tasks.

Why Use Claude Code?

Claude Code offers several key benefits:

Codebase Understanding: Quickly understand project architecture and logic.
Code Editing & Bug Fixing: Edit files and fix bugs with natural language.
Testing & Linting: Execute and fix tests and linting errors.
Git Integration: Simplify Git operations like commits, PRs, and conflict resolution.
Web Search: Access documentation and online resources.
MCP Integration
Secure & Private: Direct API connection to Anthropic, operating within your terminal.
Codebase Understanding: Quickly understand project architecture and logic.
Code Editing & Bug Fixing: Edit files and fix bugs with natural language.
Testing & Linting: Execute and fix tests and linting errors.
Git Integration: Simplify Git operations like commits, PRs, and conflict resolution.
Web Search: Access documentation and online resources.
MCP Integration
Secure & Private: Direct API connection to Anthropic, operating within your terminal.

Installation and Authentication

Install Claude Code: Ensure you have Node.js 18+ installed, then run:

npm install -g @anthropic-ai/claude-code

Authenticate: Run claude in your terminal and follow the authentication prompts.

Pro tips: You can authenticate via the Anthropic Console, Claude App (Pro/Max plan), or enterprise platforms like Amazon Bedrock or Google Vertex AI.

Pro tips: You can authenticate via the Anthropic Console, Claude App (Pro/Max plan), or enterprise platforms like Amazon Bedrock or Google Vertex AI.

If you want to make any changes to Claude Code’s initial configuration, just run the /config command

1. Initializing Your Project

Initialize: Use the /init command to generate a CLAUDE.md project guide.

Running this command, Claude Code will create the CLAUDE.md file with all the necessary information to work properly with this project.

Click Yes and you’ll see the file created in your project.

You can ask Claude Code to commit the CLAUDE.md file it just generated: “commit the CLAUDE.md file”.

Claude Code will execute git and add this file to staging.

2. Basic Usage

Asking Questions

Start by understanding your codebase:

> what does this project do?

You can try different questions like these:

> what technologies does this project use? > explain the folder structure

Make Code Changes

Instruct Claude to make edits:

> Create a GitHub Action that, on every npm publish, automatically creates a GitHub Release and publishes the package to GitHub Packages.

Claude Code will show you how it’s performing the tasks one by one

When finished, it will give you a summary of the changes it just made

Claude will show you the proposed changes and ask for your approval before modifying any files.

3. Essential Commands and Workflows

Navigating the Command Line Interface (CLI)

Claude Code offers a simple yet powerful command-line interface with tab completion for files and commands. Use /help to see all available commands and /clear to reset the conversation context.

Using Claude Code as a Unix-Style Utility

Command Breakdown: Security Scanning

cat package.json | claude -p "review this file for security vulnerabilities and dependency issues" > security_report.txt

This command demonstrates Claude Code’s versatility as a Unix-style utility that can be integrated into your existing shell scripts and workflows. Let’s break it down:

Input Piping: cat package.json | reads the contents of your package.json file and pipes it directly to Claude Code.
**Headless Mode: **The -p flag runs Claude in headless (non-interactive) mode with the specified prompt.
Specialized Analysis: The prompt instructs Claude to perform a security-focused code review, specifically looking for vulnerabilities and dependency issues.
Output Redirection: > security_report.txt captures Claude’s analysis in a text file for documentation or further processing.

Input Piping: cat package.json | reads the contents of your package.json file and pipes it directly to Claude Code.
**Headless Mode: **The -p flag runs Claude in headless (non-interactive) mode with the specified prompt.
Specialized Analysis: The prompt instructs Claude to perform a security-focused code review, specifically looking for vulnerabilities and dependency issues.
Output Redirection: > security_report.txt captures Claude’s analysis in a text file for documentation or further processing.

Working with your IDE

The /ide command connects Claude Code to your IDE (VS Code, Cursor, Windsorf or JetBrains), enabling powerful integrations:

# Connect Claude to your IDE> /ide

Automatic Context Sharing When you select files or code in your IDE, Claude automatically receives this context.

Or you can add a file as context using @

Creating Custom Slash Commands

The custom slash commands feature in Claude Code lets you create reusable prompts for common tasks:

# Create a project-specific commandmkdir -p .claude/commandsecho "Analyze this code for security vulnerabilities and suggest fixes:" > .claude/commands/security-review.md

Project Commands vs. Personal Commands

Project Commands (shared with your team):

> /project:security-review

Stored in .claude/commands/ directory
Available to everyone who clones the repo
Great for standardizing team workflows
Stored in .claude/commands/ directory
Available to everyone who clones the repo
Great for standardizing team workflows
Personal Commands (just for you):

> /user:optimize

Stored in ~/.claude/commands/directory
Available across all your projects
Perfect for your individual preferences
Stored in ~/.claude/commands/directory
Available across all your projects
Perfect for your individual preferences

Adding Command Arguments

Make commands flexible with the $ARGUMENTS placeholder:

# Create a command with argumentsecho "Find and fix issue #$ARGUMENTS. Follow these steps:1. Understand the issue described in the ticket2. Locate the relevant code3. Implement a solution4. Add appropriate tests" > .claude/commands/fix-issue.md

Then use it with:

> /project:fix-issue 123

Organizing Commands

You can create subdirectories for better organization:

.claude/commands/frontend/component.md → /project:frontend:component

Custom commands transform Claude Code into a powerful, personalized coding assistant that adapts to your specific workflows and team standards.

Creating Custom Slash Commands for npm Packages

Here’s a practical example of creating a custom slash command for npm package development:

Create the file .claude/commands/npm-contributing-docs.md

With the following content:

Create a CONTRIBUTING.md file with:   - Development setup instructions   - Testing guidelines   - Pull request process

Now you can use this command in your npm package project (you must restart claude to see the command):

The command will be executed and in this case will create the CONTRIBUTING.md file as we requested in the created command file.

Working with Model Context Protocol (MCP)

MCP allows Claude Code to connect with external tools and data sources, extending its capabilities beyond your local environment.

# Basic MCP server additionclaude mcp add postgres-db -- /path/to/postgres-mcp-server --connection-string "postgresql://user:pass@localhost:5432/mydb"

Managing MCP Servers

# List all configured serversclaude mcp list# Get details for a specific serverclaude mcp get postgres-db# Remove a serverclaude mcp remove postgres-db

Scopes for MCP Servers

Local (-s local): Available only to you in the current project
Project (-s project): Shared with everyone via .mcp.json file
Local (-s local): Available only to you in the current project
Project (-s project): Shared with everyone via .mcp.json file
User (-s user): Available to you across all projects
User (-s user): Available to you across all projects

Example: Using CodeGPT’s Deep Graph MCP

Let’s see how to implement a powerful code analysis MCP in practice:

Let’s integrate the Deep Graph MCP with Claude Code. This integration transforms Claude Code into a powerful code comprehension tool that understands your entire codebase at a semantic level, making it ideal for large, complex projects.

Here you can see all the documentation for this CodeGPT MCP and all the information to use it in Claude Code: https://github.com/JudiniLabs/mcp-code-graph

Add the MCP with the following command and don’t forget to add your CodeGPT API Key

# Add the Deep Graph MCP serverclaude mcp add "Deep Graph MCP" npx -- -y mcp-code-graph@latest YOUR_CODEGPT_API_KEY

Interacting with Deep Graph MCP

Once configured, you can use it in your Claude Code sessions:

Check if you have Deep Graph MCP installed by running “claude mcp list”

To call the MCP and use its tools, you can mention it and then request something directly. Claude Code will detect the tools it needs to execute and show you the confirmation message:

Deep Graph MCP: list graphs

With the Deep Graph MCP you can work with all the repositories you have converted into knowledge graphs.

Once your repositories are listed, you could ask Claude Code to work directly with a graph.

Deep Graph MCP: show all the endpoints from danielavila.me@main

Claude Code will execute the corresponding tools with the selected graph and finally you’ll have a direct response in the console.

Amazing! Now you can work with any repository within your project using Claude Code, special commands, Deep Graph MCP and all the tools that Claude has available!

You can follow me on my social media:

LinkedIn (Spanish): https://www.linkedin.com/in/daniel-avila-arias/
Twitter (English): https://x.com/dani_avila7
LinkedIn (Spanish): https://www.linkedin.com/in/daniel-avila-arias/
Twitter (English): https://x.com/dani_avila7

Llama 3.2 Running Locally in VSCode: How to Set It Up with CodeGPT and Ollama

Daniel (San) Ávila — Mon, 30 Sep 2024 12:09:19 +0000

Llama 3.2 models are now available to run locally in VSCode, providing a lightweight and secure way to access powerful AI tools directly from your development environment.

With the integration of Ollama and CodeGPT, you can download and install Llama models (1B and 3B) on your machine, making them ready to use for any coding task.

In this guide, I’ll walk you through the installation process, so you can get up and running with Llama 3.2 in VSCode quickly.

Step-by-Step Installation Guide: Llama 3.2 in VSCode

Step 1: Install Visual Studio Code (VSCode)

To start, make sure you have Visual Studio Code installed. If you don’t have it yet, download it from here and follow the instructions for your operating system.

Step 2: Install CodeGPT Extension

The CodeGPT extension is necessary to integrate AI models like Llama 3.2 into your VSCode environment. Here’s how to get it:

Open VSCode.
Click on the Extensions icon on the left sidebar.
Search for “CodeGPT” in the marketplace.

Step 3: Install Ollama

Ollama enables local deployment of language models. To install it:

Visit the Ollama website.
Download the appropriate installer for your operating system.
Follow the installation instructions provided on the site.
Once installed, verify it by typing the following in your terminal:

ollama --version

output: ollama version is 0.3.12

Step 4: Download Llama 3.2 Models

With CodeGPT and Ollama installed, you’re ready to download the Llama 3.2 models to your machine:

Open CodeGPT in VSCode
In the CodeGPT panel, navigate to the Model Selection section.
Select Ollama as the provider and choose the Llama 3.2 models (1B or 3B).

Click “Download Model” to save the models locally.

Step 5: Verify Your Setup

Once the model is downloaded, you can verify it’s ready to use:

Open a code file or project in VSCode.
In the CodeGPT panel, make sure Llama 3.2 is selected as your active model.
Begin interacting with the model for code completions, suggestions, or any coding assistance you need.

Ready to Use Llama 3.2 in VSCode

That’s it! With Llama 3.2 running locally through CodeGPT, you’re set up to enjoy a secure, private, and fast AI assistant for your coding tasks — all without relying on external servers or internet connections.

If you found this guide helpful, let us know in the comments, and feel free to reach out if you encounter any issues during the setup!

Lightning-Fast Code Assistant with Groq in VSCode

Daniel (San) Ávila — Sun, 21 Jul 2024 20:29:57 +0000

In this tutorial, I’ll show you how to integrate Groq’s cloud-based language models as a code assistant in Visual Studio Code (and Jetrbains) using the CodeGPT extension.

You’ll learn to set up an AI-powered coding assistant to enhance your productivity.

First, What is Groq?

Groq is the AI infrastructure company that delivers fast AI inference.

The LPU™ Inference Engine by Groq is a hardware and software platform that delivers exceptional compute speed, quality, and energy efficiency.

Let’s begin!

To use Groq’s services, you must first create an account on Groq Cloud by accessing the following link: https://console.groq.com/

Once you’re logged in, navigate to the API Keys section in the main menu.

Create a new API Key and copy it.

Next, you need to open VSCode and install the CodeGPT extension.

Open CodeGPT in VSCode and in the main selector, select Groq as the provider.

Now select “Connect” and you’ll see the window to paste the API Key you just created. Paste it and then click Connect.

Excellent! We’re now connected with Groq.

Next, we’ll select a model from the ones Groq has available. With CodeGPT, you’ll always have an updated list of models, so you’ll be able to see each new release directly in CodeGPT.

In this case, we’ll be working with the llama3–70b-8192 model.

Once the model is selected, we can start using the features offered by CodeGPT.

Explain a file:

In the CodeGPT text input, we select the /Explain option, then use the @ symbol to select a file from our project, and when we press Enter, Llama3 will provide a complete explanation of the ENTIRE file.

Document a file:

It seems like there’s no content to translate. Could you please provide the text you’d like me to translate into native English? I’ll be happy to assist you.

Experience the speed of Groq in software development with CodeGPT as your code assistant!

YoutubeGPT, start a chat with a video 🤖

Daniel (San) Ávila — Thu, 09 Feb 2023 03:15:23 +0000

With Youtube GPT you will be able to extract all the information from a video on YouTube just by pasting the video link.
You will obtain the transcription, the embedding of each segment and also ask questions to the video through a chat.

All code was written with the help of Code GPT

Features

Video transcription with OpenAI Whisper
Embedding Transcript Segments with the OpenAI API (text-embedding-ada-002)
Chat with the video using streamlit-chat and OpenAI API (text-davinci-003)

Example

For this example we are going to use this video from The PyCoach
https://youtu.be/lKO3qDLCAnk

Add the video URL and then click Start Analysis

Pytube and OpenAI Whisper

The video will be downloaded with pytube and then OpenAI Whisper will take care of transcribing and segmenting the video.

# Get the video 
youtube_video = YouTube(youtube_link)
streams = youtube_video.streams.filter(only_audio=True)
mp4_video = stream.download(filename='youtube_video.mp4')
audio_file = open(mp4_video, 'rb')

# whisper load base model
model = whisper.load_model('base')

# Whisper transcription
output = model.transcribe("youtube_video.mp4")

Embedding with "text-embedding-ada-002"

We obtain the vectors with text-embedding-ada-002 of each segment delivered by whisper

# Embeddings
segments = output['segments']
for segment in segments:
    openai.api_key = user_secret
    response = openai.Embedding.create(
        input= segment["text"].strip(),
        model="text-embedding-ada-002"
    )
    embeddings = response['data'][0]['embedding']
    meta = {
        "text": segment["text"].strip(),
        "start": segment['start'],
        "end": segment['end'],
        "embedding": embeddings
    }
    data.append(meta)
pd.DataFrame(data).to_csv('word_embeddings.csv')

OpenAI GPT-3

We make a question to the vectorized text, we do the search of the context and then we send the prompt with the context to the model "text-davinci-003"

We can even ask direct questions about what happened in the video. For example, here we ask about how long the exercise with Numpy that Pycoach did in the video took.

Running Locally

Github: https://github.com/davila7/youtube-gpt

Clone the repository

git clone https://github.com/davila7/youtube-gpt
cd youtube-gpt

Install dependencies

These dependencies are required to install with the requirements.txt file:

streamlit
streamlit_chat
matplotlib
plotly
scipy
sklearn
pandas
numpy
git+https://github.com/openai/whisper.git
pytube
openai-whisper

pip install -r requirements.txt

Run the Streamlit server

streamlit run app.py

Upcoming Features 🚀

Semantic search with embedding
Chart with emotional analysis
Connect with Pinecone

Compare human and AI responses on Stackoverflow using CodeGPT

Daniel (San) Ávila — Thu, 26 Jan 2023 23:35:39 +0000

In this article I will explain how to use Ask StackOverflow functionality of the Code GPT extension.

Code GPT extension, available for VSCode, allows you to search for questions on StackOverflow and compare the best human response with the response from artificial intelligence such as OpenAI's GPT-3.

Official Code GPT documentation: https://codegpt.co

Install CodeGPT in VSCode

Follow the instructions below to install the extension in VSCode

Install CodeGPT

Ask StackOverflow

To activate this option you just have to press the cmd + shift + p keys and look for the Ask StackOverflow option

Now search for a question or even it can be just with a keyword, in this case, we just enter python and we can get all the queries with this concept that exist on StackOverflow

Select a question and Code GPT will do the magic 😍

It will show you the question, then the best answer and for the last time you will be able to see the answer of the artificial intelligence that you have selected.

In this case, we perform the test with OpenAI and the text-davinci-003 model

Example:

Human vs AI

Try this tool on your own so that you have both human responses and AI responses within Visual Studio Code to improve your projects.

GPT-3 inside VSCode with official OpenAI API

Daniel (San) Ávila — Thu, 29 Dec 2022 19:06:35 +0000

Code GPT extension for VSCode

Using the official OpenAI API inside the IDE with Code GPT.

New feature! Get code from comments

Write a comment asking for a specific code
Press cmd + shift + i
Use the code 😎

Examples

Create a README.md with "Ask CodeGPT"

Select the code and then click on `Explain CodeGPT`

Result:

Features

Get Code: Create a comment asking for a specific code and CodeGPT will open a new Editor with the code (You don't need to write the code language. CodeGPT will detect it automatically).
Ask CodeGPT: CodeGPT will open a new Editor and respond the question.
Explain CodeGPT: CodeGPT will open a new Editor and explain the code.
Refactor CodeGPT: CodeGPT will open a new Editor and refactor the code.
Document CodeGPT: CodeGPT will open a new Editor and Document the code.
Find Problems CodeGPT: CodeGPT will open a new Editor and find problems in the code.
Set API KEY: To save your API KEY securely.
Remove API KEY: To remove your API KEY.

Installation

Go to Settings > Extensions > CodeGPT

Complete the following information:

To enter your API Key press cmd+shift+p and search for CodeGPT: Set API KEY. Your API KEY will be safely stored.

CI/CD con Synthetic Monitoring de Datadog y Bitbucket pipelines

Daniel (San) Ávila — Sat, 17 Dec 2022 04:31:17 +0000

En este artículo te mostraré como enlazar la herramienta Synthetic Monitoring de
Datadog HQ
con la herramienta Pipelines de
Atlassian Bitbucket
y tener un panel de control completo de pruebas CI/CD.

DataDog Synthetic Monitoring
Esta herramienta permite automatizar pruebas en tus procesos de despliegue de aplicaciones ya sea en ambiente staging o producción.

En este ejemplo vamos a realizar una prueba básica a la página inicial de BoxMagic (https://go.boxmagic.app) mediante diferentes navegadores y dispositivos para que sean ejecutadas cada vez que se realiza un push en el repositorio del proyecto en Bitbucket.

Browser Test
Para comenzar ingresamos al panel de Datadog y seleccionamos el menú UX Monitoring / Synthetic Tests

Creamos un nuevo test de tipo Browser Test

Agregamos la URL donde se ejecutará el Test, en este caso será https://go.boxmagic.app (nuestra url inicial), agregamos el nombre y el ambiente. Luego seleccionamos que navegadores y dispositivos serán los que realizarán las pruebas.

Se pueden agregar las ubicaciones desde donde se realizarán las pruebas.

Por último configuramos la frecuencia en que se ejecutará esta prueba y además podemos configurar el monitoreo, alertas y permisos.

Pruebas
En este caso solo crearemos una prueba de Assertion / Test an element’s content que permite verificar el contenido de un elemento dentro de la página:

Solo debes hacer click en el elemento y completar los datos del formulario, en este caso se está comprobando si el div contenga la frase “Ingresa aquí”.

Seleccionamos Apply para finalizar y comenzar a ejecutar la prueba.

Api Keys y test public ID
Para que esta prueba pueda correr cada vez que se ejecute el pipeline de bitbucket debemos obtener 3 variables Public Test ID, APP Key y API Key.

El Public Test ID se obtiene seleccionando la prueba que acabamos de realizar

El APP Key y el API Key se obtienen dentro del menú Organization Settings donde debes crear un nuevo APP Key y un API Key.

NPM datadog-ci
Puedes ejecutar estas pruebas de forma local con npm para probar que estén correctamente configuradas.

Primero ejecutamos la instalación con npm

npm install -g @datadog/datadog-ci

Luego ejecutamos synthetics run-test con los parámetros que obtuvimos del paso anterior.

datadog-ci synthetics run-tests --public-id $DD_TEST_PUBLIC_ID --apiKey $DD_API_KEY --appKey $DD_APP_KEY

El resultado debería verse de esta forma si todas las pruebas se realizaron correctamente:

Bitbucket Pipeline
Con todas las pruebas configuradas solo nos hace falta agregarlas al pipeline de bitbucket para que sean ejecutadas cada vez que se realice un push al repositorio

Agregamos las variables de entorno a nuestro repositorio en bitbucket en Repository Setting / Repository variables

En el archivo bitbucket-pipeline.yml agregamos los comandos para ejecutar las pruebas.

pipelines:
     default:
           - step:
               name: datadog-synthetics testing
               image: node:15-alpine3.10
               script:
                 - npm install -g @datadog/datadog-ci
                 - datadog-ci synthetics run-tests --public-id        $DD_TEST_PUBLIC_ID  --apiKey $DD_API_KEY --appKey $DD_APP_KEY

Una vez que realizas un push al repositorio debes ingresar al del menú Pipelines de bitbucket podrás ver la ejecución de las pruebas.

Puedes ver la realización de cada prueba en el panel de Datadog

También puedes ver la ejecución de las pruebas y en la columna RUN TYPE deberías poder observar que fueron ejecutadas por el pipeline de CI/CD

Eso es todo, ya puedes comenzar a crear todo tipo de pruebas dentro de datadog, se ejecutarán automáticamente en tu CI/CD y de esta forma te aseguras que cada push que se realice al repositorio deba pasar todas las pruebas para que se pueda desplegar en ambientes de staging o producción.

Prompt y Completion en OpenAI

Daniel (San) Ávila — Fri, 16 Dec 2022 05:07:04 +0000

En este artículo revisaremos el concepto más importante de GPT-3, el prompt.

El prompt es una instrucción que se le da a GPT-3 para que genere texto de salida.

Por ejemplo, un usuario podría usar GPT-3 para generar generar una función en Javascript usando un prompt como: “You: Cómo obtener el largo de un string”. GPT-3 leerá el texto y generará una respuesta con la solución basado en ese texto.

¿Pero cómo entiende lo que estoy escribiendo?
Gracias a el PLN o Procesamiento del Lenguaje Natural, que es un campo de la Inteligencia Artificial que se centra en permitir que las computadoras entiendan el lenguaje humano y se comuniquen con él. Esta tecnología se usa para procesar texto, voz o incluso imágenes para obtener información significativa.

Esto significa que la computadora es capaz de comprender el contexto y la intención del usuario para generar una respuesta apropiada.

GPT-3: Generative Pre-trained Transformer 3
GPT-3 fue entrenado con cientos de miles de millones de palabras (incluso código) para poder comprender el texto y generar una salida acorde a él prompt de entrada.

Si quieres saber más sobre GPT, te dejo el paper oficial 👇

Ahora volvamos al prompt…

El prompt es una frase que se utiliza como entrada para generar un texto. Esta frase puede ser de 3 formas:

zero-shot prompts
Es el más simple de los tipos de prompts. Solo entrega la descripción de un texto corto para que GPT-3 pueda comenzar con la predicción del texto siguiente.

Ejemplo de un prompt zero-shot:

En este caso con un zero-shot “Write a function in Javascript” basta para obtener el resultado esperado, ya que es una solicitud corto y directo.

one-shot prompt
En este caso se entrega un ejemplo para que GPT-3 pueda usarlo para entender cual debe ser la mejor forma de completar lo solicitado.

En el caso le entregaremos una prompt emulando una conversación entre un bot y una persona:

GPT-3 completa el texto desde “Bot:” que fue nuestra última palabra en el prompt, siguió la conversación como si fuera el Bot que le entraga la respuesta a la persona que preguntó.

few-shot prompt
Este es el más poderoso de los prompt. En este caso debes proveer un texto extenso de entre 10 a 100 tokens, entregar multiples ejemplos

Con few-shot prompt, GPT-3 incrementará la calidad de la completion ya que el prompt entrega mucha más información para que aprender.

Si ya jugaste en el playground o con ChatGPT te darás cuenta que mientras más escribas mejor serán las respuestas que entregue GPT-3 ya que tendrá más contexto.

Ya conoces el prompt… Ahora elige el modelo 🧐

Como existen diferentes tipos de prompts para que GPT-3 entienda el contexto de la conversación, también existen diferentes tipos de modelos con los que se procesará el texto de salida.

Puedes revisar los modelos en el siguiente artículo 👇

Modelos de GPT-3 y Codex
En este artículo revisaremos que es un modelo y cuales son los distintos tipos de modelos que OpenAI ofrece para ocupar…
medium.com

Puedes encontrar todos mis artículos ordenados por categorías aquí: rebrand.ly/davila7

Si tienes dudas, te dejo mi https://www.linkedin.com/in/daniel-avila-arias/

Crea una función Lambda con Serverless Framework y Datadog

Daniel (San) Ávila — Wed, 14 Dec 2022 07:03:47 +0000

En este artículo les mostraré como ocupar las dos herramientas increíbles para crear y monitorear funciones lambda… Serverless Framework y Datadog 🙌

¿Qué es Serverless Framework?

Es una herramienta Open Source diseñada para construir funciones lambda de AWS. Además esta herramienta se puede ejecutar para otros proveedores cloud como Azure o Google Cloud.

Puedes crear tu cuenta gratuita acá 👉serverless.com

¿Qué es Datadog?

Si tienes una web tienes que ocupar Datadog. Es una herramienta increíblemente completa para el monitoreo de todo lo que se te ocurra dentro de una web. Controla el CI/CD, front, back, cloud y además puedes crear paneles personalizados con la información que necesites.

Puedes crear tu cuenta gratuita acá 👉 datadoghq.com

Ahora comencemos…

Comenzamos con Serverles Framework

Primero debes ingresar en tu panel web de serverless y selecciona el menú org

Ve a la pestaña provaiders y selecciona add. Te entregará 3
formas de integrar tu cuenta de AWS. Yo prefiero la primera ya que es bastante simple y segura.

Al hacer click en Connect AWS provider se abrirá tu cuenta de AWS con un template listo con los permisos para ejecutar serverless. Selecciona crear deberas esperar a que el estado cambie a UPDATE_COMPLETE.

Vuelve a serverless y tu cuenta ya estará conectada.

Para instalar serverless en tu computador ejecuta el siguiente comando

npm install -g serverless

Ya instalado el primer comando que debes correr para crear un proyecto es serverless y te mostrará un listado de varios proyectos base que puedes ocupar. En este caso vamos a seleccionar AWS — Python — Starter

Te solicitará el nombre del proyecto y luego te pedirá hacer login en la cuenta de Serverless Framework.

Se abrirá el navegador para confirmar que es tu cuenta 👍

Luego te preguntará si quieres hacer deploy… no lo hagas aún, primero vamos a instalar el plugin de datadog.

Ingresamos en la carpeta y ejecutamos el siguiente comando:

serverless plugin install --name serverless-plugin-datadog

Debería mostrarse el plugin instalado correctamente

Abre el proyecto creado con VSCode

Ingresamos al archivo serverless.yml y agregamos el siguiente código

custom:
  datadog:
    site: DATADOG_SITE
    apiKey: DATADOG_API_KEY

Completa los datos DATADOG_SITE y DATADOG_API_KEY, acá te doy una ayuda de donde puedes encontrar cada uno 👇

DATADOG_SITE : datadoghq.com

o puedes verificar en el siguiente enlace:

Para generar una API Key debes ingresar en el panel de Datadog y selecciona la última opción del menú, luego ingresa en Organization Settings

Luego ve a la opción API Keys

New Key, ponle el nombre, luego cópiala y pegala en DATADOG_API_KEY

Una vez configurado tu archivo serverless.yml solo debes ejecutar el siguiente comando (sls = serverless)

sls deploy

Así de fácil Serverless Framework se encargará de TODO!!! Permisos de aws, función, conectar datadog, etc…

No estás seguro de lo que pasó? Vamos a confirmar

1- Revisamos si la función Lambda se creó correctamente ✅
Ingresa en la consola de AWS y ve al servicio Lambda, la función ya está ahí!

Ingresamos y podemos ver todo el código de nuestro proyecto.

Al realizar un test deberíamos obtener el mensaje por defecto de la función hello dentro de handler.py, puedes cambiar el código en tu proyecto local y volver a ejecutar sls deploy para ver lo fácil que es subir el código.

2- El deploy al panel de Serverless se realizó correctamente ✅
Ingresamos a serverless.com y podemos ver que en el menú apps ya se encuentra la app HolaServerlessDatadog. Además nos entrega unas métricas de ejecución.

*3- Estamos recibiendo la información en Datadog correctamente * ✅
Para esta confirmación primero debemos armar un Dashboard. Para eso solo debes hacer click en el siguiente enlace:

Lambda Dashboard Template

Selecciona Clone Dashboard y dale un nombre

Ya tienes un dashboard completo con todo lo que necesitas saber sobre la función lambda que acabas de implementar 😱

Conclusión

Serverless Framework es una herramienta increíble para comenzar un proyecto con Lambda, tienes cientos de templates que puedes ocupar con tan agregar el comando template, por ejemplo con este comando ya tenemos lista una función que ejecuta un cron job en Nodejs.

serverless --template-url=https://github.com/serverless/examples/tree/v3/aws-node-scheduled-cron

Datadog es una herramienta increíble para monitoreo de absolutamente cualquier proyecto, la instalación es rápida y la creación de paneles es muy fácil. Puedes personalizarlos como quieras, de esta sabrás en todo momento lo que ocurre con tus funciones lambda.

Forem: Daniel (San) Ávila

Complete Guide to Claude Code Templates

What is Claude Code Templates

Component

Agents

Commands

Settings

Hooks

MCPs

Templates

Using the Search Function

Installing Individual Components

Building Complete Stacks with Cart Functionality

Discovering Popular Components with Trending

Conclusion

Claude Code: From Zero to Hero

What is Claude Code?

Why Use Claude Code?

Installation and Authentication

1. Initializing Your Project

2. Basic Usage

Asking Questions

Make Code Changes

3. Essential Commands and Workflows

Navigating the Command Line Interface (CLI)

Using Claude Code as a Unix-Style Utility

Working with your IDE

Creating Custom Slash Commands

Project Commands vs. Personal Commands

Adding Command Arguments

Organizing Commands

Creating Custom Slash Commands for npm Packages

Working with Model Context Protocol (MCP)

Managing MCP Servers

Scopes for MCP Servers

Example: Using CodeGPT’s Deep Graph MCP

Interacting with Deep Graph MCP

Llama 3.2 Running Locally in VSCode: How to Set It Up with CodeGPT and Ollama

Step-by-Step Installation Guide: Llama 3.2 in VSCode

Lightning-Fast Code Assistant with Groq in VSCode

First, What is Groq?

Let’s begin!

Explain a file:

Document a file:

YoutubeGPT, start a chat with a video 🤖

Features

Example

Pytube and OpenAI Whisper

Embedding with "text-embedding-ada-002"

OpenAI GPT-3

Running Locally

Upcoming Features 🚀

Compare human and AI responses on Stackoverflow using CodeGPT

Install CodeGPT in VSCode

Ask StackOverflow

GPT-3 inside VSCode with official OpenAI API

Code GPT extension for VSCode

New feature! Get code from comments

Examples

Create a README.md with "Ask CodeGPT"

Select the code and then click on Explain CodeGPT

Result:

Features

Installation

CI/CD con Synthetic Monitoring de Datadog y Bitbucket pipelines

Prompt y Completion en OpenAI

Crea una función Lambda con Serverless Framework y Datadog

¿Qué es Serverless Framework?

¿Qué es Datadog?

Comenzamos con Serverles Framework

Conclusión

Select the code and then click on `Explain CodeGPT`