Docker-native optimization

Optimize any Docker workload with CLI flags and stdout metrics.

Build an image, let HyperOptimizer inject --hpo-* values, and print hpo.metrics.* lines. We run the trial containers, collect results, and rank the best parameter sets.

Start the quickstart Metric contract

The contract

Three requirements. Nothing else.

1

Docker image

image:tag

Runs one trial from a normal container command.

2

CLI inputs

--hpo-*

HyperOptimizer injects parameter values at runtime.

3

Metric output

hpo.metrics.*

Your process prints structured results to stdout.

How it works

Package once. Run many trials.

HyperOptimizer schedules parameterized containers, reads stdout, and uses completed results to choose better trials.

Illustration showing HyperOptimizer packaging a Docker image, running many parameterized trials, collecting metrics, and ranking the best result. — One Docker image becomes many trials. Each run gets a different parameter set, prints metrics, and feeds the dashboard ranking.

Start here

Pick one path.

The sidebar has the full docs tree. This page only points you to the first useful next step.

QuickstartBuild the smallest working trialCreate a Docker image, parse parameters, emit metrics, and run it.Open Bring your ownWrap an existing container or scriptUse Python, Rust, Julia, shell scripts, or any binary that runs in Docker.Open TradingConnect a strategy workloadStart with Freqtrade, NautilusTrader, or another backtesting engine.Open

Reference

Look up the exact rules when you need them.

Metric formatObjective metrics, guardrails, and value syntax.CLI argumentsFlag naming, injection rules, and constraints.Container requirementsEntrypoints, image expectations, logs, and reproducibility.TroubleshootingMissing metrics, failed trials, and unexpected results.