My API Is Up, but Slow. Is It My Code or the Server?
A hypothetical slow API on a small VPS shows how application and host metrics can point to CPU contention from another process.
Engineering Insights
Practical notes, release announcements, and engineering deep-dives from the projects built at PVR Labs.
Monitoring a FastAPI App on a Small VPS
Try practical FastAPI monitoring locally, then run the same small StatLite and SQLite setup alongside an application on a VPS.
Lightweight Express.js Monitoring Without Prometheus and Grafana
A small Express helper exposes /statlite/metrics so StatLite can chart traffic, errors, latency, and Node.js memory without a Prometheus and Grafana stack.
Spring Boot Fat JAR vs Extracted Layout: Does It Actually Matter?
The official extracted layout cut median Spring startup 23.3% and first-request latency 34.1% in six paired runs. Settled RSS and swap did not show a material winner.
Spring Boot vs Quarkus on a 512 MB VPS
The same Star Pulse workload, JDK 25, and 80 MiB heap. Quarkus started larger, then had the smaller resident set, while RSS hid swap and the box later ran out of headroom.
Your Coding Agent Doesn’t Need 1,476 Lines of Test Output
A successful test run produced 1,476 lines. The agent needed roughly one.
Your coding-agent quota is at 99%. Now what?
The agent already explored the repo and understands the task. Hand off that session state instead of reconstructing it somewhere else. Or import a local Codex or Claude Code session directly.
Spring Boot on a 256 MB VPS: What Changed with JDK 25 Compact Object Headers
A JDK 25 Alpine follow-up completed the hour on a nominal 256 MiB VPS, but high swap use and a long scheduling delay kept the result firmly in stretch-test territory.
Spring Boot on a 512 MB VPS: How Lightweight Monitoring Still Fits
A Spring Boot 3.5 app with JPA, Tomcat, and Actuator ran on 512 MB plus a small swapfile. The monitor stayed around 12 MiB.
Reviewing agent-generated code with an independent second model in VS Code
Review complete Git changes in VS Code, then use focused repository context when an AI reviewer needs more evidence.
I Taught Maven to Be an Introvert
How a small Bash wrapper reduced successful Maven output by more than 99.7% while preserving useful diagnostics for coding agents.
Local-First AI Coding: Keeping Proprietary Repositories Local
Use AI coding assistants without granting them repository access: map locally, disclose only reviewed context, and confirm every proposed write.
Lightweight Spring Boot Monitoring Without Prometheus and Grafana
Turn Spring Boot Actuator data into a focused dashboard for a few applications on a VPS, without operating a full Prometheus and Grafana stack.
Reduce Claude Code Context with Local Repository Mapping
Autonomous coding agents consume tokens rapidly during reviews by rescanning projects and pulling broad context. Learn how to map locally, extract precise snippets, and review efficiently in any web chat.