YAML editor

YAML editor

When you edit an object or create resources from YAML, the editor knows the fields of each kind from the cluster itself: it completes field names and allowed values, underlines what the cluster would not accept, shows a field's documentation when you point at it, and opens an Explain panel like kubectl explain. It works for built-in kinds and for custom resources, at the versions your cluster serves.

Contents: Where it works · Completion · Problems · Documentation and Explain · Where the schema comes from · Without a schema · Keyboard

Where it works

• Edit (details panel or row menu): the object's YAML in a dock tab.
• Create resource (+ in the dock's bar): YAML you paste or type, also several documents separated by ---. Each document is checked against the schema of its own apiVersion and kind.

Helm's Edit values and the editors of extensions do not use the schema: their text is not a Kubernetes object.

Completion

Start typing a key, or press Ctrl+Space:

• Fields of the object at the cursor: required ones first, fields already written left out, each with its type and the first line of its description.
• Values: the allowed values of a field that has a fixed list (for example imagePullPolicy, restartPolicy, a Service's type), and true/false for booleans.

The path at the cursor is worked out from the indentation, so completion works while the text is still half written: in block maps and lists, after ---, around block scalars (|, >), and in flow style ({…}, […]) on the cursor's line.

Problems

A short while after you stop typing, the editor underlines:

Problem	Example
Unknown field	spec.replica instead of spec.replicas
Wrong type	replicas: "3" where a number is expected
Missing required field	a container without name
Value not allowed	imagePullPolicy: Sometimes
YAML error	bad indentation, a tab, an unclosed quote

Fields whose schema accepts anything (for example a ConfigMap's data keys, x-kubernetes-preserve-unknown-fields, embedded objects) are not reported as unknown. The status of an object is not checked.

The problems are advice: they never block Save or Apply. The API server's answer decides, and Preview changes shows what it would store, with its own validation and admission webhooks.

Documentation and Explain

• Hover over a field to see its type, whether it is required, its default and allowed values, and its description.
• Explain — press F1, or choose Explain in the hover — opens a panel under the editor with the field at the cursor: its documentation and the list of its fields; choose a field to go down, Up to go back to the parent, Close to close it. It shows what kubectl explain <kind>.<path> would, for the version your cluster serves.

Where the schema comes from

From the cluster's OpenAPI v3 (/openapi/v3), read by Sextant's backend with your credentials, one API group and version at a time, when an editor first needs a kind of it. Sextant keeps a compact form of it in memory (the last 12 group/versions of the last 4 clusters), and the editor gets only the one kind it edits, so a large document never reaches the window. The schema is read again after a reconnect and when the cluster's kinds change (a CRD was installed or updated). Nothing is written to disk.

Kinds the cluster does not serve are never asked for, so a half-typed kind: does nothing until it names a real kind.

Without a schema

The editor stays a normal YAML editor, and everything else works as before, when:

• the cluster does not publish OpenAPI v3 (Kubernetes before 1.24 by default), or the kind is not served — a note says The cluster publishes no schema for …;
• reading the schema failed (Sextant tries again after 30 seconds);
• the cluster is disconnected and the kind was never read. A schema read earlier keeps working while disconnected.

Keyboard

Key	What it does
Ctrl+Space	Show completions
Enter	Accept the completion
F1	Explain the field at the cursor
Escape	Close the completion list

AI agents (MCP) and extensions get nothing new from this feature.