David
c492ef74d1
A capability is a descriptor plus what the endpoint declares (device/Capability.ts), and its discovery object is
generated from the two. The new API is bridge.addDevice({ endpointId, name, categories, ... }) and
device.add(PowerController, options): both throw a DeclarationError that names the endpoint, the interface and
the instance (device/validate.ts), and leave the bridge and the device as they were. AlexaInterface is the same
Capability with the 1.x methods on it; it, ActionMapping and the enums moved to src/compat/, Device to src/device/.
What a 1.x caller can observe:
- every endpoint ends with { type: "AlexaInterface", interface: "Alexa", version: "3" } (alexa-interface.html);
new Alex2MQTT(..., { alexaInterface: false }) leaves it out
- the fields of a capability object come in the order of Amazon's examples; their content is unchanged
- addCapability() with a name that is not an interface throws (1.5.2 announced it with the version "UNKNOWN")
- ActionMapping takes the payload as an object; a JSON string is parsed (1.5.2 sent the string), any other throws
- what Alexa would reject in a 1.x declaration is not refused: device.check() lists it and the bridge logs each
line once, as "warning: ..." through the log hook, when it answers a discovery
- a device whose JSON cannot be built is left out of the answer and reported as an error event
- PowerController and EndpointHealth are the descriptors and keep ON/OFF and OK/UNREACHABLE; PowerState is new
Tests: six zoo devices declared the 1.x way give the JSON that Alexa accepted from 1.5.2 on 2026-09-28, plus the
Alexa capability. npm test: 85 pass (was 57) in 10-12 s, also on Node 18.20.8 and 20.20.2.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
49 lines
2.1 KiB
TypeScript
49 lines
2.1 KiB
TypeScript
import { Capability } from "../device/Capability.js";
|
|
import type { CapabilityJson } from "../device/Capability.js";
|
|
import type { Semantics } from "../registry/types.js";
|
|
import type { ActionMapping } from "./ActionMapping.js";
|
|
import type { AlexaInterfaceType } from "./enums.js";
|
|
/** One ModeController mode as discovery lists it (configuration.supportedModes): the value plus its friendly names. */
|
|
export interface SupportedMode {
|
|
value: string;
|
|
modeResources?: {
|
|
friendlyNames: Array<{
|
|
"@type": string;
|
|
value: {
|
|
text?: string;
|
|
locale?: string;
|
|
assetId?: string;
|
|
};
|
|
}>;
|
|
};
|
|
}
|
|
interface Options {
|
|
semantics?: Semantics;
|
|
supportedModes?: Array<string | SupportedMode>;
|
|
}
|
|
/**
|
|
* A capability as 1.x declares it: device.addCapability(type, options), then the add and set methods below. It is a
|
|
* Capability, so what the methods set reaches discovery the same way as the options of device.add(). Nothing is
|
|
* refused here except an interface name the registry does not have: what Alexa would reject is logged by the bridge
|
|
* when it answers a discovery.
|
|
*/
|
|
export declare class AlexaInterface extends Capability<any, any, Options> {
|
|
constructor(type: AlexaInterfaceType | string, retrievable?: boolean, proactivelyReported?: boolean, instance?: string);
|
|
/** The namespace of the interface, which is the value of its AlexaInterfaceType member. */
|
|
get type(): AlexaInterfaceType;
|
|
addActionMapping(mapping: ActionMapping): void;
|
|
addFriendlyName(name: string, locale: string): void;
|
|
/** The modes of a ModeController, as { value, modeResources } objects. */
|
|
addSupportedModes(modes: Array<string | SupportedMode>): void;
|
|
setInstance(name: string): void;
|
|
getType(): AlexaInterfaceType;
|
|
/** @deprecated The same as getType(). */
|
|
getTypeString(): string;
|
|
/** @deprecated Read descriptor.version. */
|
|
getVersion(): string;
|
|
/** @deprecated Read the keys of descriptor.properties. */
|
|
getProps(): string[];
|
|
getJSON(): CapabilityJson;
|
|
toJSON(): CapabilityJson;
|
|
}
|
|
export {};
|