Iterate with follow-up prompts
Keep working on the same QuickPizza flow from earlier milestones so each reply is a small, reviewable delta. Prefer the Grafana Cloud k6 script editor when you can so you can use Create and Run after changes; if you still use a local quickpizza-assistant.js file, paste updates back into the same chat thread each round.
To iterate on your script, complete these steps.
Paste the current full script, or only the section you are changing, back into the chat, or keep going in the same thread if your workflow allows.
Apply one focused follow-up at a time. For QuickPizza, the public demo does not require API tokens, so use refinements like load shape, checks, and thresholds. For private APIs (not QuickPizza), you can add bearer auth from
__ENV.API_TOKENonly when an endpoint actually needs it. For example, use prompts like these.Keep the target https://quickpizza.grafana.com. Add tags: { name: "quickpizza-smoke" } to every HTTP request in the default function so metrics use a stable group name.Add a ramp stage from 1 to 5 VUs over 2 minutes, then 30 seconds at 5 VUs, for this QuickPizza script only.Add thresholds so http_req_failed stays below 1% and p95 http_req_duration stays below 800ms for this short test.
After each reply, sync the latest script into the k6 script editor when you are working in Cloud, then run Create and Run for a smoke check. Refer to Open the k6 script editor and Use the script editor. If you use a local file instead, run this command.
k6 run --iterations 1 quickpizza-assistant.jsUse your real filename. Confirm the new behavior appears and nothing obvious regressed.
Keep a one-line comment in the script noting each prompt, for example
// follow-up: added p95 threshold.Stop iterating when you have a realistic load shape for a demo test, meaningful checks, and thresholds you are willing to explain to a reviewer.
You now have a practical loop for refining generated code in Cloud or locally. In the next milestone, you apply a final verification checklist before CI.