How-to Guides
The rest of the documentation explains what System9 is and why it works the way it does. This hub is the task layer: short, ordered runbooks that get one job done, using real commands and the surfaces the platform actually ships.
Every guide states its prerequisites, names the exact command or endpoint, and, where an operator surface is not yet wired end-to-end, says so plainly instead of inventing a flag. Read the linked concept and reference pages when you want the model behind a step.
Start here
If you are new, do the quickstart first. It stands up a cell from the machine image and confirms it is serving. The device quickstart then attaches an Edge Agent to it.
Then pick the job you have:
| Guide | Type | What you get |
|---|---|---|
| Bring up a node | Runbook | Stand up a node and its cells across all four deployment shapes, starting with the bootstrap token most first attempts miss |
| Connect an identity provider | How-to | Federate Entra ID, Okta, Google or any OIDC or SAML provider, map groups to roles, and prove a real person logs in with the roles you intended |
| Connect a BI or reporting tool | How-to | Point Power BI or another reporting tool at the data platform over the SQL, Flight SQL, or Delta Sharing surfaces |
| Set up data-platform resources | How-to | Create the resources a data platform is made of: catalogs, workspaces, clean rooms, data products, and connectors |
| Use workspaces and clean rooms | How-to | Create, populate, invite into, and govern a workspace and a clean room |
| Connect a data source | How-to | Point the platform at a warehouse, object store, stream or website, and confirm the data path rather than just the handshake |
| Scan an estate | How-to | Scan someone else's data, cloud or virtualization estate in place, under signed scope, and produce a signed deliverable |
| Connect an Edge Agent | How-to | Enroll and connect an Edge Agent, run it as a daemon or in chat mode, and ingest to a cell |
| Build agents and workflows | How-to | Set up an agent, give it tools and a model, and compose agents into a governed workflow graph |
Two more task guides live with the capability they belong to: applying and updating a licence, and when the meter cannot report for what happens across a disconnection.
The triad these guides sit in
Figure: where the how-to layer sits. Get Started orients you (neutral); a how-to guide (green) does one job; the reference and concept pages (filled) ground every step and are linked from each guide. Ceiling: these guides use the shipped commands and surfaces; an operator may rebind ports and disable surfaces, and anything not wired end-to-end is called out in place.
Table equivalent
| Layer | Answers | Where |
|---|---|---|
| Get Started | "What is this, and how do I run one?" | Get Started hub |
| How-to guides | "How do I do this specific job?" | this hub |
| Reference & concepts | "What exactly is a cell / port / workflow node?" | the four topic hubs |
The rules every guide follows
- Grounded in real surfaces. Ports, protocols, and commands are the shipped ones, cross-referenced to the Ports & protocols reference and the Edge command reference.
- Honest about what is wired. Where the machinery exists but the operator surface is not connected end-to-end, the guide says "owed" and points at what does work today, rather than documenting a command that does not run.
- Matrix-aware. Each guide notes how the job behaves single-cell, multi-cell, sovereign/air-gapped, and cloud-enabled, because a step that assumes connectivity fails the case the product is sold on.
- It ends by proving itself. Every guide has a verify it worked section, and the check that matters is never "did the call succeed" but "did the thing actually happen". A connector reporting healthy proves a handshake, not a data path.
- It says what to do when it fails. Every guide has a when it does not work table, including the cases where a refusal is the correct behaviour rather than a fault. Knowing which is which is most of operating this.