Skip to content

The Canvas

Open any product from Journey Maps and you land on its canvas. This is where the scan's output becomes something you can read, question, and act on: the map autopilots are built from.

The overview

The default view shows all workflows laid out under their stages, with the stage rail on the left and the product's properties on the right: source status, framework, last scan, and the counts (routes, workflows, components, copy strings).

Each workflow card carries its autopilot state once one exists: Workflow ✓ (synthetic users completed it), ✦ Certified (the handover works too), or ✦ Live (the offer is on for real users).

Inside a workflow

Click a workflow in the stage rail to open its flow. The canvas draws three layers.

The journey. Steps, decisions, and the screens involved, as the user experiences them. Page nodes show the real screen, rendered from your code, so the flow stays grounded in the product.

The autopilot. A header card at the top of every workflow. It carries the "Do this for me" pill, the autopilot's state (candidate, draft, workflow certified, certified, live, failing), both certification levels with their run counts, and, once certified, how long a handover takes. Entry pages show where the offer appears. Once a certified operating profile exists, page nodes show what the autopilot does there: asks where it asks the user a question, you type where the user fills a field themselves, Allow where a consequential click waits for the user.

Click the header card to edit the autopilot in the properties panel: the name on the offer, the step budget, the certification record, what it asks and waits on, and the live toggle. Both certification levels come first; the toggle stays off until they pass.

Friction. Points where a user stalls, hits an error, or drops off, drawn as small dashed cards beside the step. They are hypotheses that seed rehearsal scenarios; nothing in production reacts to them. Show or hide the layer with the Friction button in the top bar. Right-click any step to add a friction point or start a rehearsal scenario from it.

Rehearsals on the map

Scenarios are authored on the canvas, anchored to the step they start from. A step with scenarios shows a small ▶ badge; after a run, the badge turns green or red with the latest verdict. Select a scenario to see the verdict, open the run, or watch the clip.

Handover scenarios (the synthetic user hands the task to the autopilot) certify the autopilot itself. They run from the Autopilots page; the header card links there.

Coverage gaps

The properties panel lists what the scan could not reach in the codebase. Gaps are normal after a first scan. Re-scan after the code moves, or add the workflow by hand.

Re-scanning

Code moved? Click Re-scan in the properties panel, or run /scan again from the repo. The map updates in place and changed rows are flagged on the canvas so you can review what moved. Re-scans version the artifact; nothing you configured or deployed is lost.

Tip

Scans bind to the repo through .holostaff/source.json, so every re-scan lands on the same source. See the CLI reference.