stampede run

Run a scenario in-process and write a report (no server needed)

Synopsis

Run a scenario with an in-process engine. Nothing else needs to be running: no server, no database. Prints a summary and can write HTML, PDF, CSV, JSON, JUnit and Markdown reports.

With –cluster the scenario runs on the workers of the server you signed in to with stampede login instead: the file is saved to the server as a new version, the run is started against a target saved on the server (–target, or the one whose base URL is –base-url, or the project’s only target) and followed live, and the server’s report is written to the same outputs. This is the same as stampede start –file.

Exit codes: 0 when every target passes (or none are set), 3 when a target fails, 1 on any other error.

stampede run <scenario.yaml> [flags]

Examples

  stampede run checkout.yaml
  stampede run checkout.yaml --shape spike --rate 200/s -o report.html
  stampede run smoke.yaml -e TARGET_URL=http://localhost:8090 --junit junit.xml
  stampede run checkout.yaml --cluster --project shop --target staging --duration 10m -o report.html

Options

      --ai-base-url string    provider API base URL for --narrative
      --allow-host strings    extra public hosts requests may reach besides the target
      --base-url string       override target.baseURL
      --cluster               run on the signed-in server's workers instead of in-process
      --csv string            write per-step figures as CSV to this file (- for stdout)
  -d, --detach                with --cluster: print the run id and return without following
      --duration string       override the duration, e.g. 30s or 5m
  -e, --env stringArray       set ${env.KEY} for the scenario (KEY=VALUE, repeatable)
  -h, --help                  help for run
      --iterations int        run a fixed number of iterations instead
      --json string           write a JSON report to this file (- for stdout)
      --junit string          write JUnit XML (one test per target) to this file
      --max-vus int           cap on virtual users in rate mode (default: five times the peak rate); raise it for long journeys
      --md string             write a Markdown summary to this file (- for stdout)
      --model string          model for --narrative (default claude-sonnet-5-5 for anthropic)
      --narrative             add an AI-written summary; every claim cites the report's figures (needs a provider key)
      --note string           with --cluster: note recorded with the run
  -o, --out string            write an HTML report to this file
      --pause string          pause between repeats (default "10s")
      --pdf string            write a PDF report to this file (needs Chrome or Chromium)
      --project string        with --cluster: project name, slug or id (default: the only project)
      --provider string       model provider for --narrative: anthropic, openai, gemini, ollama or openai-compatible (default "anthropic")
  -q, --quiet                 no live progress
      --rate string           override the arrival rate, e.g. 100/s (switches to rate mode)
      --region stringArray    with --cluster: split the load by worker region, REGION=PERCENT (repeatable, adds up to 100%), e.g. --region mumbai=50% --region frankfurt=50%; replaces load.regions
      --repeat int            run the scenario this many times (for stampede compare); report files get -1, -2 ... suffixes (default 1)
      --shape string          apply a traffic shape: smoke, baseline, stress, spike, soak, breakpoint, steps, recovery, wave
      --target string         with --cluster: target name, base URL or id saved on the server
      --timeline-csv string   write the per-second timeline as CSV to this file (- for stdout)
  -v, --verbose               log step errors as they happen
      --vus int               override the number of virtual users
      --workers int           with --cluster: number of workers (0 = all)

SEE ALSO

• stampede - Self-hosted, distributed load testing