@xopcai/xopc 0.0.286 → 0.0.291

package/dist/packages/computer-control-contract/src/index.js

@@ -0,0 +1,178 @@
+import { __esmMin } from "../../../_virtual/_rolldown/runtime.js";
+import { z } from "zod";
+//#region packages/computer-control-contract/src/index.ts
+var COMPUTER_CONTROL_TOOL, ComputerProfileSchema, Id, Point, ComputerActionSchema, ComputerTargetSchema, ComputerObservationSchema, ComputerModelBindingSchema, ActionEnvelopeSchema, ComputerAppSchema, ComputerWindowSchema, ComputerCommandSchema, COMPUTER_INPUT_SCHEMA, COMPUTER_OUTPUT_SCHEMA;
+var init_src = __esmMin((() => {
+	COMPUTER_CONTROL_TOOL = "desktop.computer.control";
+	ComputerProfileSchema = z.enum(["gui-plus-2026-02-26", "structured-tools-v1"]);
+	Id = z.string().min(1).max(200);
+	Point = z.object({
+		x: z.number().finite().min(0),
+		y: z.number().finite().min(0)
+	}).strict();
+	ComputerActionSchema = z.discriminatedUnion("kind", [
+		z.object({
+			kind: z.literal("click"),
+			point: Point,
+			button: z.enum(["left", "right"]),
+			count: z.union([z.literal(1), z.literal(2)])
+		}).strict(),
+		z.object({
+			kind: z.literal("setValue"),
+			ref: Id,
+			text: z.string().max(16384)
+		}).strict(),
+		z.object({
+			kind: z.literal("typeText"),
+			text: z.string().max(16384),
+			point: Point.optional()
+		}).strict(),
+		z.object({
+			kind: z.literal("pressKeys"),
+			keys: z.array(z.string().regex(/^[a-zA-Z0-9_]+$/).max(40)).min(1).max(4)
+		}).strict(),
+		z.object({
+			kind: z.literal("scroll"),
+			point: Point,
+			deltaX: z.number().int().min(-2e3).max(2e3),
+			deltaY: z.number().int().min(-2e3).max(2e3)
+		}).strict(),
+		z.object({
+			kind: z.literal("wait"),
+			durationMs: z.number().int().min(0).max(2e3)
+		}).strict()
+	]);
+	ComputerTargetSchema = z.object({
+		appId: Id,
+		pid: z.number().int().positive(),
+		processIdentity: Id,
+		windowId: Id,
+		width: z.number().int().positive().max(16384),
+		height: z.number().int().positive().max(16384),
+		geometryRevision: Id
+	}).strict();
+	ComputerObservationSchema = z.object({
+		id: Id,
+		sessionId: Id,
+		brokerEpoch: Id,
+		generation: z.number().int().nonnegative(),
+		target: ComputerTargetSchema,
+		capturedAt: z.number().int().positive(),
+		imageWidth: z.number().int().positive(),
+		imageHeight: z.number().int().positive(),
+		summary: z.string().max(12e3),
+		focusedEditableRef: Id.optional(),
+		stateDigest: Id
+	}).strict();
+	ComputerModelBindingSchema = z.object({
+		modelRef: Id,
+		profile: ComputerProfileSchema,
+		origin: z.string().url(),
+		runtimeLocation: z.enum([
+			"local",
+			"remote",
+			"unknown"
+		]),
+		upstreamOrigin: z.string().url().optional()
+	}).strict();
+	ActionEnvelopeSchema = z.object({
+		actionId: Id,
+		sessionId: Id,
+		observationId: Id,
+		brokerEpoch: Id,
+		generation: z.number().int().nonnegative(),
+		grantId: Id,
+		deadlineAt: z.number().int().positive(),
+		action: ComputerActionSchema
+	}).strict();
+	ComputerAppSchema = z.object({
+		appRef: Id,
+		name: z.string().max(300),
+		running: z.boolean()
+	}).strict();
+	ComputerWindowSchema = z.object({
+		windowRef: Id,
+		title: z.string().max(300),
+		visible: z.boolean()
+	}).strict();
+	ComputerCommandSchema = z.discriminatedUnion("op", [
+		z.object({
+			op: z.literal("discover"),
+			sessionId: Id,
+			owner: Id,
+			query: z.string().max(200)
+		}).strict(),
+		z.object({
+			op: z.literal("open"),
+			sessionId: Id,
+			owner: Id,
+			appRef: Id,
+			windowRef: Id.optional(),
+			mode: z.enum(["observe", "control"]),
+			prepare: z.boolean(),
+			model: ComputerModelBindingSchema
+		}).strict(),
+		z.object({
+			op: z.literal("status"),
+			sessionId: Id,
+			owner: Id
+		}).strict(),
+		z.object({
+			op: z.literal("observe"),
+			sessionId: Id,
+			owner: Id
+		}).strict(),
+		z.object({
+			op: z.literal("act"),
+			sessionId: Id,
+			owner: Id,
+			envelope: ActionEnvelopeSchema
+		}).strict(),
+		z.object({
+			op: z.literal("release"),
+			sessionId: Id,
+			owner: Id
+		}).strict()
+	]);
+	COMPUTER_INPUT_SCHEMA = z.toJSONSchema(ComputerCommandSchema);
+	COMPUTER_OUTPUT_SCHEMA = {
+		type: "array",
+		minItems: 1,
+		maxItems: 2,
+		items: { oneOf: [{
+			type: "object",
+			additionalProperties: false,
+			required: ["type", "value"],
+			properties: {
+				type: { const: "json" },
+				value: { type: "object" }
+			}
+		}, {
+			type: "object",
+			additionalProperties: false,
+			required: [
+				"type",
+				"fileId",
+				"name",
+				"mimeType",
+				"size",
+				"sha256"
+			],
+			properties: {
+				type: { const: "file" },
+				fileId: { type: "string" },
+				name: { type: "string" },
+				mimeType: { enum: ["image/png", "image/jpeg"] },
+				size: {
+					type: "integer",
+					minimum: 1,
+					maximum: 5242880
+				},
+				sha256: { type: "string" }
+			}
+		}] }
+	};
+}));
+//#endregion
+init_src();
+export { ActionEnvelopeSchema, COMPUTER_CONTROL_TOOL, COMPUTER_INPUT_SCHEMA, COMPUTER_OUTPUT_SCHEMA, ComputerActionSchema, ComputerAppSchema, ComputerCommandSchema, ComputerObservationSchema, ComputerProfileSchema, ComputerWindowSchema, init_src };

package/dist/packages/realtime-protocol/src/voice-audio.js

@@ -1,13 +1,5 @@
+import { __esmMin } from "../../../_virtual/_rolldown/runtime.js";
 //#region packages/realtime-protocol/src/voice-audio.ts
-const HEADER_BYTES = 32;
-const MAGIC = 1481592883;
-const VERSION = 3;
-const UPLINK = 1;
-const DOWNLINK = 2;
-const PCM_S16LE = 1;
-const MAX_FRAME_BYTES = 64 * 1024;
-const encoder = new TextEncoder();
-const decoder = new TextDecoder("utf-8", { fatal: true });
 function encode(frame) {
 	const id = encoder.encode(frame.id);
 	if (!id.length || id.length > 160 || !Number.isInteger(frame.connectionEpoch) || frame.connectionEpoch < 1 || frame.connectionEpoch > 4294967295 || !Number.isInteger(frame.seq) || frame.seq < 1 || frame.seq > 4294967295 || !Number.isFinite(frame.timestampMs) || frame.timestampMs < 0 || frame.durationMs !== 20 || frame.flags < 0 || frame.flags > 3 || !frame.audio.length || frame.audio.length % 2 || HEADER_BYTES + id.length + frame.audio.length > MAX_FRAME_BYTES) throw new Error("Invalid voice media frame");
@@ -77,5 +69,18 @@ function decodeVoiceUplinkAudioFrame(bytes) {
 		audio: frame.audio
 	};
 }
+var HEADER_BYTES, MAGIC, VERSION, UPLINK, DOWNLINK, PCM_S16LE, MAX_FRAME_BYTES, encoder, decoder;
+var init_voice_audio = __esmMin((() => {
+	HEADER_BYTES = 32;
+	MAGIC = 1481592883;
+	VERSION = 3;
+	UPLINK = 1;
+	DOWNLINK = 2;
+	PCM_S16LE = 1;
+	MAX_FRAME_BYTES = 64 * 1024;
+	encoder = new TextEncoder();
+	decoder = new TextDecoder("utf-8", { fatal: true });
+}));
 //#endregion
-export { decodeVoiceUplinkAudioFrame, encodeVoiceAudioFrame };
+init_voice_audio();
+export { decodeVoiceUplinkAudioFrame, encodeVoiceAudioFrame, init_voice_audio };

package/dist/packages/realtime-protocol/src/voice-catalog.js

@@ -0,0 +1,61 @@
+import { __esmMin } from "../../../_virtual/_rolldown/runtime.js";
+import { z } from "zod";
+//#region packages/realtime-protocol/src/voice-catalog.ts
+var voiceModeCapabilitySchema, voiceManifestSchema, voiceSelectionSchema;
+var init_voice_catalog = __esmMin((() => {
+	voiceModeCapabilitySchema = z.enum([
+		"transcription",
+		"transcription.stream",
+		"speech",
+		"speech.stream",
+		"conversation"
+	]);
+	voiceManifestSchema = z.strictObject({
+		protocolVersion: z.literal(1),
+		serviceVersion: z.number().int().positive(),
+		modes: z.array(voiceModeCapabilitySchema),
+		transport: z.literal("websocket-pcm"),
+		inputFormat: z.strictObject({
+			encoding: z.literal("pcm_s16le"),
+			sampleRate: z.literal(16e3),
+			channels: z.literal(1)
+		}),
+		outputFormat: z.strictObject({
+			encoding: z.literal("pcm_s16le"),
+			sampleRate: z.literal(24e3),
+			channels: z.literal(1)
+		}),
+		turnDetection: z.array(z.literal("server_vad")),
+		bargeIn: z.boolean(),
+		tools: z.boolean(),
+		resumable: z.literal(false),
+		voices: z.array(z.strictObject({
+			id: z.string().min(1),
+			name: z.string(),
+			languages: z.array(z.string())
+		})),
+		defaultVoice: z.string().min(1).optional(),
+		limits: z.strictObject({
+			maxFrameBytes: z.number().int().positive().max(65536),
+			maxSessionSeconds: z.number().int().positive()
+		})
+	});
+	voiceSelectionSchema = z.strictObject({
+		mode: voiceModeCapabilitySchema,
+		model: z.string().min(1).max(200),
+		voice: z.string().min(1).max(200).optional()
+	});
+	z.strictObject({
+		revision: z.string().min(1),
+		catalogVersion: z.string().nullable(),
+		selections: z.array(voiceSelectionSchema),
+		models: z.array(z.strictObject({
+			id: z.string(),
+			name: z.string(),
+			voice: voiceManifestSchema
+		}))
+	});
+}));
+//#endregion
+init_voice_catalog();
+export { init_voice_catalog, voiceManifestSchema, voiceModeCapabilitySchema, voiceSelectionSchema };