Posts on paval.io

Pull Request Code Review in GitHub Actions

Wed, 20 Aug 2025 00:00:00 +0000

As a followup to OpenAI Codex CLI in GitHub Actions this new workflow automates pull request code reviews:

Check out this demo repo that uses OpenAI Codex CLI + MCPs in GitHub Actions.

Markdown to Confluence Sync Workflow

Sat, 29 Nov 2025 00:00:00 +0000

I wanted to jot down a few notes about the Sync to Confluence workflow that keeps the markdown-to-confluence docs mirrored in Atlassian.

Why it’s handy

Keeping authoring in Markdown means the docs stay right beside the codebase, so assistants like GitHub Copilot (and other IDE-aware agents) can ingest real project guidance without leaving the editor. The same repo search powers autocomplete, inline explanations, and context-aware chat, which is way faster than switching to a browser tab and hunting through Confluence. Non-developers still get the polished Confluence pages produced by the workflow, so stakeholders have a familiar place to read without digging into Git.

Embedding Apache Superset Dashboards

Tue, 07 Oct 2025 00:00:00 +0000

A demonstration project showing how to embed Apache Superset dashboards into external applications using the Superset SDK.

Overview

Apache Superset is a powerful open-source data exploration and visualization platform. This demo project showcases how to embed Superset dashboards into a Spring Boot application, enabling seamless integration of analytics dashboards into your own applications.

Repository: danielpaval/superset-embed-demo

Project Stack

The project combines several technologies:

Apache Superset (Python) - The analytics platform
Spring Boot (Java) - Backend application framework
Docker - Containerized deployment
Gradle - Build automation

Key Configuration

To enable embedding Superset dashboards, several configuration changes are required:

Spring Boot API Demo

Tue, 07 Oct 2025 00:00:00 +0000

This WIP spring-boot-api-demo repo gathers reusable components and common practices in building Spring Boot applications all the way from exposing a REST API to the ultimate SQL Server-based JPA persistence.

OpenAI Codex in GitHub Actions

Fri, 08 Aug 2025 00:00:00 +0000

Check out this demo repo that uses OpenAI Codex + MCPs in GitHub Actions:

Structurizr vs LikeC4

Sun, 11 May 2025 00:00:00 +0000

This WIP post compares Structurizr and LikeC4 with a focus on their export capabilities.

2. Export

2.1 Export to Static Site

Structurizr

Prerequisites:

Graphviz dot

Command:

structurizr export -workspace workspace.dsl -format static

Conlusions:

⚠️ The static export of Structurizr is a very trimmed version of the Lite version, without the top menu, just the navigation feature starting with the top context diagram

LikeC4

likec4 build -o ./dist

2.2 Export to SVG/PNG

Structurizr

Structurizr CLI can export to SVG or PNG format via a Node.js and Puppeteer workaround:

Mermaid CLI

Mon, 14 Apr 2025 00:00:00 +0000

Let’s see how Mermaid’s CLI can generate images from diagrams:

npm install -g @mermaid-js/mermaid-cli
mmdc -i .\sequence_diagram.mmd -o sequence_diagram.png

Output:

Hello, Mermaid!

Sun, 13 Apr 2025 00:00:00 +0000

Let’s try Hugo’s support of mermaid diagrams:

---
config:
look: neo
theme: neo
---
sequenceDiagram
Alice->>John: Hello John, how are you?
John-->>Alice: Great!
Alice-)John: See you later!

Hugo does not include out-of-the-box support for Mermaid diagrams, but it can be easily customized:

See https://github.com/danielpaval/danielpaval.github.io/tree/master/layouts external to defined theme

layouts/_default/_markup/render-codeblock-mermaid.html

<pre class="mermaid">
 {{ .Inner | htmlEscape | safeHTML }}
</pre>
{{ .Page.Store.Set "hasMermaid" true }}

layouts/_default/baseof.html before </body>

{{ if .Store.Get "hasMermaid" }}
<script type="module">
 import mermaid from 'https://cdn.jsdelivr.net/npm/mermaid/dist/mermaid.esm.min.mjs';
 mermaid.initialize({ startOnLoad: true });
</script>
{{ end }}

JSON Event Data Audit Demo

Sun, 16 Mar 2025 00:00:00 +0000

Context

The opportunity for this analysis originated with the need to ingest highly-dynamic unstructured data produced by various events and be able to efficiently report on it both programmatically and to (business) users. In an ecommerce scenario, an event could be a customer registration, an order, a shipment, a return, etc.:

{
 "type": "REGISTRATION",
 "data": {
 "name": "Jane Smith",
 "email": "jane.smith@example.com",
 "phone": "987-654-3210",
 "addresses": [
 {
 "default": true,
 "street": "1742 Waverley Drive",
 "city": "Palo Alto",
 "state": "CA",
 "zip": 94301,
 "country": {
 "name": "United States",
 "code": "US"
 }
 }
 ]
 }
}

TLDR

Chain of Thought & Prototyping

The following steps roughly show the evolution of my ideas and solution components along the way. In order to proceed with an objective evaluation of improvements I tried to rely on a core set of non-functional metrics:

To Envoy or not to Envoy

Wed, 05 Mar 2025 00:00:00 +0000

I recently came across the need for a reusable microservice in the form of an OCI / Docker image to meet different authentication and authorization requirements depending on the end project and environment. While the current limited set of possible security configurations is handled by the microservice itself, I analyzed the possibility of delegating this aspect to a specialized sidecar container living in the same Kubernetes pod.

After a bit of prompt engineering and wanting to consider something else than NGINX I decided to first try Envoy Proxy. Long story short, here’s a Docker Compose setup:

Claude Code - Hello world!

Sun, 02 Mar 2025 00:00:00 +0000

As I heard about Claude Code at the end of this week¹, I wanted to try it out myself. It took a bit of time to get the CLI going as Claude Code only works on Linux, and my WSL Node.js setup was both old and misconfigured.

The task I put it up to was to help me show the hidden posts of this Hugo site, as I had no clues. Honestly, I was surprised to hit a speed bump already, having to buy some token credits first ($5) without any trial. But then it correctly suggested that I forgot about the draft: true front matter field I had copy-pasted in multiple locations.

Hugo vs Jekyll

Sat, 01 Mar 2025 00:00:00 +0000

I just got to make a bit of progress on building a technical blog which is something I started to think about a few years ago. Back at the beginning, the focus was on getting familiar with static site generators, mainly Jekyll and Hugo.

Now, I’m going with Hugo for, mostly, simplicity:

Feature	Hugo 🏎️	Jekyll 📝
Performance	⚡ Extremely fast (built with Go)	🚶 Slower (built with Ruby)
Installation	✅ Single binary, no dependencies	❌ Requires Ruby and Bundler
Templates	⚠️ Uses Go templates (harder to learn)	✅ Uses Liquid templates (easier)
Content Format	✅ Markdown, TOML/YAML/JSON	✅ Markdown, YAML
Plugins	❌ No plugin system (built-in features only)	✅ Supports plugins for more functionality
GitHub Pages	⚠️ Requires extra setup	✅ Natively supported
Ease of Use	❌ Steeper learning curve	✅ Easier for beginners
Best For	🚀 Speed, large sites, developers	✍️ Blogs, simple websites, GitHub Pages

Anthropic Model Context Protocol (MCP)

Fri, 28 Feb 2025 00:00:00 +0000

Elasticsearch

Install Python
Install uv: pip install uv
Clone https://github.com/cr7258/elasticsearch-mcp-server
Setup Claude

uvx Setup

{
 "mcpServers": {
 "elasticsearch": {
 "command": "uvx",
 "args": [
 "elasticsearch-mcp-server"
 ],
 "env": {
 "ELASTIC_HOST": "http://localhost:9200",
 "ELASTIC_USERNAME": "elastic",
 "ELASTIC_PASSWORD": "changeme"
 }
 }
 }
}

uv Setup

{
 "mcpServers": {
 "elasticsearch": {
 "command": "uv",
 "args": [
 "--directory",
 "C:\\Repositories\\mcp\\elasticsearch-mcp-server",
 "run",
 "elasticsearch-mcp-server"
 ],
 "env": {
 "ELASTIC_HOST": "http://localhost:9200",
 "ELASTIC_USERNAME": "elastic",
 "ELASTIC_PASSWORD": "changeme"
 }
 }
 }
 }

File System

Setup

{
 "mcpServers": {
 "filesystem": {
 "command": "npx",
 "args": [
 "-y",
 "@modelcontextprotocol/server-filesystem",
 "C:\\Users\\dani\\Desktop"
 ]
 }
 }
}

Atlassian

Win32 Epoch (1/1/1601)

Sat, 18 Jun 2022 00:00:00 +0000

ℹ️ This is a quick blog post to mainly play with Jekyll.

Not long ago Windows on my laptop reported that Skype was last used on 1/1/1601.. While I rarely use Skype now-a-days, it surely wasn’t that long! January 1st does hint at some kind of a date baseline, just like the Unix epoch, but why 1601?

Of course, this was a quick answer to look up and it turns out to be that “1601 was the first year of the 400-year Gregorian calendar cycle at the time Windows NT was made”¹. More specifically, “The FILETIME structure records time in the form of 100-nanosecond intervals since January 1, 1601”².