Build a pattern scaffold in Studio

The Studio workspace is where Atlas turns intent into artifacts. Every time you open /studio, you step through the same five-station pipeline — Classify → Plan → Generate → Validate → Transport. This tutorial walks the first four stations against the sandbox.

Studio pipeline — Atlas turns a task description into a transported CDS view through six gated stages.

What you will build

A task graph, a set of ABAP / CDS scaffolds, and a validation run — all anchored to a single decision the classifier makes and every step of it traceable on the graph canvas.

1 · Open Studio

Click Studio in the left nav. The canvas fills the view; an inspector pane sits on the right. At the top is the classify strip, a single-line band that shows the current pattern decision.

The task input field at the bottom of the canvas accepts a one-sentence description. Type or paste:

list top customers by net sales, last four quarters, ranked

Press Run. Atlas calls POST /api/atlas/studio/classify and the classify strip populates.

2 · Classify — read the decision strip

The classify strip — five elements on a single line, each one reviewable.

The classify strip shows the chosen pattern, a score, a keyword-coverage honesty chip (Atlas’s classifier is heuristic, not probabilistic), the reference artifacts the classifier leaned on, and a Reclassify button.

If the decision is wrong — it may pick transactional when you meant analytical, or vice versa — click Reclassify and pick the right shape from the menu. The rest of the pipeline re-runs against the new pattern.

3 · Plan — read the task graph

Below the classify strip the canvas renders the task graph: nodes for each artifact the plan will need, edges for dependencies. Released C2 views show green, C1 amber, C0 red. A consumption view sits on top; composites and basics fan down from it.

Drag to pan, scroll to zoom, click a node to pin the inspector. The inspector shows the node’s type, its stability contract, and the evidence that links it to the Resolve stream.

→

Starter-entity warning

If the plan needs a view Atlas cannot resolve to a released CDS, it paints the node with a starter-entity warning and the inspector suggests a replacement. Swap the binding in the inspector before generating and the warning clears.

4 · Generate — emit the scaffolds

Once the plan is approved, hit Generate. Atlas calls POST /api/atlas/studio/scaffold. The right pane fills with a Monaco editor showing the generated files — DDL for the CDS views, ABAP for any behavior definitions, metadata for service bindings. Each file is annotated with the evidence that led to it.

You can copy the text out of Monaco, read it, and leave it there. No export or push today — the generated text is display-only.

5 · Validate — read the gate results

Click through to Validate. Atlas runs the gate suite against the generated plan and shows results per gate:

• green — passed
• amber — starter-entity warning (see Step 3)
• red — would block transport (e.g. binding to a P_ view)

Resolving a red or amber gate means going back to the plan or the binding the gate is flagging and adjusting it; the gate re-runs when you regenerate. For the step-by-step shape of the fix, see Resolve a gate; for what each gate actually checks, see the Gate catalog.

6 · Transport — the handoff marker

The fifth station is the customer-side hand-off. Atlas assembles the validated artifacts and evidence into a bundle; from there the customer’s dev system and CTS pipeline carry it through the landscape. The Studio marks the plan as handed off once the bundle is ready.

Where to go next

• Run your first investigation if you skipped the investigation side.
• VDM layer architecture — the hierarchy the classify strip and the task graph are both placing nodes inside.
• The Studio pipeline for the conceptual story behind the five stations.