Skip to main content
Create an account →
Create an account →

Claude Skills

The Quanton Operator ships two Claude Code skills (slash commands) that automate the full local setup and benchmarking workflow from your terminal.

Getting started​

Install Claude Code, then start it in the root of the quanton-operator repo:

git clone https://github.com/onehouseinc/quanton-operator
cd quanton-operator
claude

You'll need onehouse-values.yaml from the Onehouse console to install the operator.

Available skills​

/setup-and-run-example​

Sets up a complete local Quanton environment from scratch and runs a sample SparkPi job end-to-end.

What it does:

  1. Checks for prerequisites (minikube, Helm, kubectl, Docker)
  2. Starts minikube if not running
  3. Installs the Kubeflow Spark Operator
  4. Installs the Quanton Operator using your onehouse-values.yaml
  5. Submits the example QuantonSparkApplication
  6. Streams logs until the job completes and prints the result

Usage:

/setup-and-run-example

Claude walks you through each step interactively, handles errors, and gives live progress updates as jobs run.

/run-tpcds-benchmark​

Runs the full TPC-DS benchmark comparing OSS Apache Spark vs Quanton on your local minikube cluster.

What it does:

  1. Asks for scale factor (default: SF=1) and configuration options
  2. Builds the datagen Docker image in minikube's daemon
  3. Generates TPC-DS Parquet data
  4. Runs 99 TPC-DS queries on OSS Spark (baseline)
  5. Runs 99 TPC-DS queries on Quanton
  6. Produces a per-query comparison table with speedup multipliers

Usage:

/run-tpcds-benchmark

Claude gives live progress updates as each phase runs and summarizes the results when complete.

How skills work​

Claude Code skills are slash commands defined in the repository's .claude/ directory. When you run /setup-and-run-example or /run-tpcds-benchmark, Claude Code expands the command into a full task prompt and executes it using the tools available in your terminal (kubectl, helm, docker, etc.).

Skills can:

  • Run shell commands in your local environment
  • Read and write files
  • Stream output from long-running processes
  • Ask clarifying questions and respond to errors interactively

Prerequisites for skills​

SkillRequirements
/setup-and-run-exampleminikube, Helm, kubectl, Docker, onehouse-values.yaml
/run-tpcds-benchmarkminikube (4+ CPU, 8+ GB, 50+ GB disk), Spark Operator, Quanton Operator already installed
Apple Silicon

The default ARM Quanton build is optimized for AWS Graviton and may fail on older Apple Silicon Macs that don't support the latest ARM instruction sets. If you hit instruction set errors, use a cloud-based Spark cluster instead.