Agentic platform engineering

Talks, demos, and workshop material for building platforms that agents can use safely.

Status: teaching assets and fixture paths are under development. Live API and isolated-cluster paths have not been validated for event delivery. No sponsorship, model credits, deployment, recording, or publication beyond this repository is implied.

Contents

Material	Location
Cloud Native AI Summit Melbourne (28 Oct 2026), 20 minutes + 5 minutes Q&A	Talk A deck, the 3am drift review demo
AI Engineer Sydney (8 Dec 2026), 18 minutes + 2 minutes transition; full-length dry run at AI Engineering Night (29 Oct)	Talk B slides, rehearsal plan, separate release-gate demo
Infracoders, Agentic Platform Engineering with StackQL (15–20 minute slot)	CFP synopsis, slides, StackQL MCP + stackql-deploy demo scaffold
Half-day workshop, 180 teaching/lab minutes + 30-minute break	Participant workbook, facilitator runbook
Shared workshop scenario, fixtures, Python and TypeScript safety paths	Workshop scenarios
Public and live-delivery gates	Release checklist
Current repository rules	AGENTS.md

Safe local path

The default smoke path is offline and uses no API key, network, cluster, or background process:

Set-Location '10-26 Cloud Native AI Summit APAC\demo\agent'
npm ci
npm run smoke
npm run typecheck

The AI Engineer demo is an independent dependency-free Node.js fixture:

Set-Location '11-26 AI Engineer\demo'
npm test
npm run demo

The Python fixture path uses only the standard library:

Set-Location '11-26 AI Engineer\workshop\scenarios\python'
python -m unittest -v

Live model execution is a separate, explicit opt-in and requires an approved OpenAI API account, API key, GPT-6-family model access, and spend controls. The workshop's hosted isolated infrastructure and participant model access remain organizer dependencies; fixture support does not satisfy that hosted commitment.

Licensing and safety

The repository is MIT-licensed. Examples use synthetic state. See each runbook for mode labels, approval boundaries, reset limitations, and release gates. Do not apply the legacy CronJob example or connect a default Kubernetes context.