346 hands-on deployment labs, 523 module configuration guides and 41 certification study guides. Counted 2026-08-17, all published at docs.radmodules.dev.
Three separate bodies of writing, each with a different job. All of it is public, none of it is behind a sign-in.
One for each of the 346 application options. A lab takes you from an empty project to a running, verified workload, then through day-2 operations and a clean teardown. It ends with something running, not with a diagram.
One for each of the 346 application options, plus 177 shared foundation guides those options build on. They document the Google Cloud services a module provisions and every input the deployment form will ask you for, grouped and with defaults.
Study material aligned to seven Google Cloud certification tracks, written against the foundation modules you can actually deploy — Services_GCP, App_CloudRun and App_GKE — so the revision and the practice are the same environment.
Those counts cover the 346 application options in the catalogue. The other eight options — the migration, service-mesh and reference-architecture modules RAD publishes itself — carry a configuration guide and a lab each in the public repository they are published from, rather than on the documentation site.
Every lab follows the same shape, so once you have done one you know where to look in all of them.
Stated at the top, before you commit an afternoon. The Elasticsearch on GKE Autopilot lab, for example, estimates 45–90 minutes.
What you will be able to do at the end, what has to exist first, and which shell variables the rest of the lab reuses. Where a module depends on shared infrastructure, the lab says whether the platform provisions it for you.
The deploy step runs through the RAD form; the verify step is command line. You end holding a real endpoint and a health response, not a screenshot of one.
Inspect, scale and update the workload; read its logs and metrics in Cloud Logging and Cloud Monitoring. This is the part most tutorials stop before.
The failures that actually happen on that module, with their causes — then how to destroy everything you created. Nothing is left running by accident.
Open one now — no account is needed to read it, and each one names the module it deploys.
A stateful workload done properly: a StatefulSet with an SSD volume that survives restarts, single-node discovery enforced at plan time, and the endpoint another application will later consume.
A managed-container deployment of a publishing platform, for the reader who wants to see the Cloud Run operating model rather than Kubernetes. Ghost also ships as a GKE variant.
Self-hosted Git, and one half of a pair the catalogue knows about: when Gitea and Woodpecker CI are deployed together as a solution, Gitea's service URL is designed to be written into Woodpecker's configuration as the forge it builds from. The OAuth application is still one you register by hand.
The reference half of the pair. If the lab is what you do, the guide is what the form is asking you.
Capability by capability — compute, persistent storage, secrets, ingress, image registry — named as the specific Google Cloud service, including the ones a module deliberately does not use.
The same grouping the deployment form uses, so you can read the guide beside the form. Basic mode asks only the mandatory inputs; Advanced exposes the full set documented here.
Where a module refuses an invalid combination at plan time, the guide says so and says why — for example the heap-to-memory ratio Elasticsearch is held to, because violating it produces out-of-memory kills under load.
Mechanics common to every application on a given operating model — workload identity, ingress, autoscaling, backups, the deployment lifecycle — live in one foundation guide rather than being repeated 300 times and drifting apart.
41 study guides across seven Google Cloud certification tracks. Each track has an overview guide plus section-by-section exploration guides.
The link runs from the guides to the platform, not the other way round: the labs carry no certification tag. Each track's overview guide defines deployment profiles — which foundation modules to deploy and which variables to set for a given study session — and each section guide then maps one exam domain onto those settings, with the observation to make in the console and the command-line call that confirms it. A coverage legend marks what the modules demonstrate fully, what they demonstrate partly, and what you will have to study elsewhere.
How to deploy, update, tear down, read build logs and understand what credits are being spent on — written per role, because the console shows you only what your role can do.
The shared overview: signing in, finding your way around, and the core concepts — modules, deployments, credits and billing. Start here whichever role you hold.
Browsing the catalogue, deploying through the guided form, tracking your own deployments, reading their logs, updating and tearing them down, and managing your credit balance.
For module authors: connecting your own GitHub repository, syncing your modules into the catalogue, and what happens when someone else deploys one.
A worked route through the self-hosted generative-AI modules — model serving, chat interfaces, gateways, retrieval pipelines and vector storage — with the lab and the configuration guide linked for each.
There are separate guides for the administrator, finance, support and agent roles, published alongside these.
The labs were written to be run by one person, and they hold up when thirty people run them at once in thirty separate Google Cloud projects.
Cohort environments are created in RAD's lab tier, which carries the tightest guardrail set — external IP addresses denied, default networks suppressed, service-account key creation blocked, and the Google Cloud services that can be enabled restricted to an allowlist. Choose labs whose modules live comfortably inside that.
Each lab states its own estimated time, and more than half of recorded deployments finish inside 20 minutes with more than nine in ten inside 30 — so a session can be planned around the provisioning wait rather than interrupted by it.
A pre-composed solution deploys several applications as one dependency-ordered unit, which makes a plausible final exercise once the single-module labs are done.
Cohort provisioning shipped in August 2026 and is early access — newer than the rest of the platform. We are looking for pilot partners, and we would rather tell you that here than have you find out afterwards.
Taken from the documentation site's own sitemap, read on 2026-08-17. We publish what the dates say, not a narrative.
Open the lab for the application you care about, decide whether the mechanism convinces you, and only then create an account. Signing up gives you 300 credits and asks for no payment method.
RAD is in beta. The platform, catalogue and documentation described here are built and running; features marked as early access are newer and we say so where that is true. Catalogue figures counted 2026-08-17.