@agimon-ai/doompi-voice 0.0.1-alpha.77 → 0.0.1-alpha.78

package/README.md

@@ -22,7 +22,7 @@ Voice owns these authenticated server endpoints:
 | same                                    | `POST /control`  | Send `{action, target?}`                                 |
 
 Control actions are `status`, `manual`, `activate`, `deactivate`, `mute`, `unmute`, `interrupt`,
-and `transfer` (with a target catalog order). The typed session method `voice.control` uses the
+and `transfer` (with a target catalog order and its catalog revision). The typed session method `voice.control` uses the
 same operations as commands and minor-mode actions. Browser recording uploads continue to use
 `/api/sessions/:sessionId/plugin/voice-media/manual/transcribe`.
 
@@ -204,7 +204,7 @@ limited to 4,096 characters and returns `completed`, `interrupted`, `superseded`
 Once confirmed user speech begins, new narration waits in memory until that turn is synchronously delivered to Pi. Multiple waiting requests use the configured `autoCapture.model` to become one bounded spoken update, with one bounded deterministic utterance if compaction fails. Narration that was already playing is not stopped by this queue gate; the existing ranked barge-in policy remains responsible for interruption.
 Narration fails closed while Voice is starting or draining, during shutdown, reload, or
 deactivation, and when the request belongs to a stale session.
-Only the currently active TUI session receives Voice-owned tools.
+Only the currently active TUI session receives Voice-owned tools. Paired-host targets are discovered through authenticated Voice grants and appear in the same revision-bound transfer catalog. A handoff prepares the target, releases the source, activates the target, and waits for usable media readiness. It does not transfer agent context, transcripts, pending work, or buffered source audio.
 
 If no `narrate` attempt is made before a final response, the active Voice session can produce one
 bounded fallback utterance. Short final responses use deterministic text; longer ones may use

package/dist/extensions/web.mjs

@@ -224,7 +224,7 @@ function parseBrowserVoiceOwnershipPayload(value) {
 		"type",
 		"version",
 		"activeSessionId"
-	]) || input.type !== "browser-media-session" || input.version !== 2 || input.activeSessionId !== null && !id(input.activeSessionId, 200)) return void 0;
+	]) || input.type !== "browser-media-session" || input.version !== 3 || input.activeSessionId !== null && (typeof input.activeSessionId !== "string" || !id(input.activeSessionId, 200) && !/^peer\/[A-Za-z0-9][A-Za-z0-9._-]{0,127}\/[A-Za-z0-9][A-Za-z0-9._-]{0,127}$/u.test(input.activeSessionId))) return void 0;
 	return input;
 }
 //#endregion
@@ -893,6 +893,17 @@ var apiRoutes_default = defineApiRoutes({
 		path: "/readiness",
 		response: apiResponse()
 	},
+	relayBinding: {
+		method: "POST",
+		path: "/relay-binding",
+		response: apiResponse()
+	},
+	relay: {
+		method: "POST",
+		path: "/relay",
+		query: ["binding"],
+		response: apiResponse()
+	},
 	clientConnect: {
 		method: "POST",
 		path: VOICE_MEDIA_ROUTES.clientConnect,
@@ -972,7 +983,7 @@ var apiRoutes_default = defineApiRoutes({
 		response: apiResponse()
 	}
 });
-createApiClient(apiRoutes_default, {
+const voice = createApiClient(apiRoutes_default, {
 	scopes: [
 		"global",
 		"workspace",
@@ -2720,12 +2731,17 @@ function jsonBody(value) {
 		body: JSON.stringify(value)
 	};
 }
+function route(path, query = {}) {
+	const url = new URL(path, "http://doompi.local");
+	for (const [name, value] of Object.entries(query)) if (value !== void 0) url.searchParams.set(name, String(value));
+	return `${url.pathname}${url.search}`;
+}
 function cancelResponseBody(response) {
 	try {
 		response?.body?.cancel().catch(() => void 0);
 	} catch {}
 }
-async function boundedControlRequest(input, init, deadlineMs, consume, parentSignal) {
+async function boundedControlRequest(input, init, deadlineMs, consume, parentSignal, transport = (input, init) => sealedTransport.fetch(input, init)) {
 	const deadlineController = new AbortController();
 	const signal = parentSignal ?? deadlineController.signal;
 	let response;
@@ -2748,7 +2764,7 @@ async function boundedControlRequest(input, init, deadlineMs, consume, parentSig
 	parentSignal?.addEventListener("abort", abortParent, { once: true });
 	let fetched;
 	try {
-		fetched = sealedTransport.fetch(input, {
+		fetched = transport(input, {
 			...init,
 			signal
 		});
@@ -2777,6 +2793,7 @@ var BrowserVoiceMediaTransport = class {
 	sessionId;
 	push;
 	controlLocation;
+	binding;
 	constructor(sessionId) {
 		this.sessionId = sessionId;
 	}
@@ -2789,6 +2806,79 @@ var BrowserVoiceMediaTransport = class {
 	media() {
 		return voiceMedia.session(this.sessionId);
 	}
+	remote() {
+		return /^peer\/[A-Za-z0-9][A-Za-z0-9._-]{0,127}\/[A-Za-z0-9][A-Za-z0-9._-]{0,127}$/u.test(this.sessionId);
+	}
+	async fetch(path, directUrl, init = {}) {
+		if (!this.remote()) return sealedTransport.fetch(directUrl, init);
+		const request = new Request(new URL(path, "http://doompi.local"), init);
+		const bytes = new Uint8Array(await request.arrayBuffer());
+		let connectionId = new URL(request.url).searchParams.get("connectionId") ?? void 0;
+		if (!connectionId && bytes.byteLength > 0) try {
+			const value = JSON.parse(new TextDecoder().decode(bytes));
+			if (typeof value.connectionId === "string") connectionId = value.connectionId;
+		} catch {
+			connectionId = void 0;
+		}
+		if (!connectionId) throw new Error("Paired Voice request has no connection binding.");
+		if (!this.binding || this.binding.connectionId !== connectionId || this.binding.expiresAt <= Date.now() + CONTROL_REQUEST_DEADLINE_MS) {
+			const issued = await voice.global.relayBinding({ body: {
+				target: this.sessionId,
+				connectionId
+			} });
+			if (!issued.ok) throw resultError(issued);
+			const value = issued.data;
+			if (typeof value.binding !== "string" || !Number.isSafeInteger(value.expiresAt)) throw new Error("Paired Voice media binding is invalid.");
+			this.binding = {
+				id: value.binding,
+				connectionId,
+				expiresAt: value.expiresAt
+			};
+		}
+		let binary = "";
+		for (const byte of bytes) binary += String.fromCharCode(byte);
+		const result = await voice.global.relay({
+			query: { binding: this.binding.id },
+			body: {
+				targetSessionId: "",
+				method: request.method === "GET" ? "GET" : "POST",
+				path,
+				headers: [...request.headers.entries()],
+				body: btoa(binary)
+			}
+		});
+		if (!result.ok) throw resultError(result);
+		const envelope = result.data;
+		if (!Number.isSafeInteger(envelope.status) || !Array.isArray(envelope.headers) || typeof envelope.body !== "string") throw new Error("Paired Voice relay response is invalid.");
+		const decoded = atob(envelope.body);
+		const body = Uint8Array.from(decoded, (character) => character.charCodeAt(0));
+		return new Response(body.byteLength === 0 ? null : body, {
+			status: envelope.status,
+			headers: envelope.headers
+		});
+	}
+	async remoteResult(path, directUrl, init) {
+		const response = await this.fetch(path, directUrl, init);
+		let data;
+		try {
+			data = response.status === 204 ? void 0 : await response.json();
+		} catch {
+			data = void 0;
+		}
+		if (response.ok) return {
+			ok: true,
+			status: response.status,
+			data,
+			response
+		};
+		return {
+			ok: false,
+			status: response.status,
+			error: typeof data === "object" && data !== null && "error" in data && typeof data.error === "string" ? data.error : `Voice media request failed with status ${String(response.status)}.`,
+			data,
+			response
+		};
+	}
 	async connect(clientId, connectionId, capabilities) {
 		this.push = void 0;
 		const controlLocation = sealedTransport.active() ? "remote" : "local";
@@ -2807,14 +2897,16 @@ var BrowserVoiceMediaTransport = class {
 	}
 	async disconnect(clientId, connectionId) {
 		try {
-			const result = await this.media().clientDisconnect({ body: {
+			const body = {
 				clientId,
 				connectionId
-			} });
+			};
+			const result = this.remote() ? await this.remoteResult(VOICE_MEDIA_ROUTES.clientDisconnect, "", jsonBody(body)) : await this.media().clientDisconnect({ body });
 			if (!result.ok && result.status !== 409) throw resultError(result);
 		} finally {
 			this.push = void 0;
 			this.controlLocation = void 0;
+			this.binding = void 0;
 		}
 	}
 	async nextEvent(clientId, connectionId, after, signal) {
@@ -2832,29 +2924,39 @@ var BrowserVoiceMediaTransport = class {
 		}
 	}
 	declareClient(clientId, connectionId, capabilities, controlLocation) {
-		return this.media().clientConnect({ body: {
+		const body = {
 			version: 6,
 			clientId,
 			connectionId,
 			clientKind: "browser",
 			controlLocation,
 			capabilities
-		} });
+		};
+		return this.remote() ? this.remoteResult(VOICE_MEDIA_ROUTES.clientConnect, "", jsonBody(body)) : this.media().clientConnect({ body });
 	}
 	async fetchEvent(clientId, connectionId, after, signal, nonblocking) {
-		return boundedControlRequest(this.media().clientEvents.url({ query: {
+		const path = route(VOICE_MEDIA_ROUTES.clientEvents, {
+			clientId,
+			connectionId,
+			after,
+			...nonblocking ? { wait: "0" } : {}
+		});
+		const directUrl = this.remote() ? "" : this.media().clientEvents.url({ query: {
 			clientId,
 			connectionId,
 			after,
 			...nonblocking ? { wait: "0" } : {}
-		} }), {}, CONTROL_REQUEST_DEADLINE_MS, async (response) => {
+		} });
+		return boundedControlRequest(directUrl, {}, CONTROL_REQUEST_DEADLINE_MS, async (response) => {
 			if (response.status === 204) return void 0;
 			if (!response.ok) throw await responseError(response);
 			return await response.json();
-		}, signal);
+		}, signal, (_input, init) => this.fetch(path, directUrl, init));
 	}
 	heartbeat(clientId, connectionId) {
-		return boundedControlRequest(this.media().clientHeartbeat.url(), jsonBody({
+		const path = VOICE_MEDIA_ROUTES.clientHeartbeat;
+		const directUrl = this.remote() ? "" : this.media().clientHeartbeat.url();
+		return boundedControlRequest(directUrl, jsonBody({
 			clientId,
 			connectionId
 		}), CONTROL_REQUEST_DEADLINE_MS, async (response) => {
@@ -2862,37 +2964,45 @@ var BrowserVoiceMediaTransport = class {
 			const wake = parseVoiceMediaWakePayload(await response.json());
 			if (wake === null) throw new Error("Voice media heartbeat response is invalid.");
 			return wake;
-		});
+		}, void 0, (_input, init) => this.fetch(path, directUrl, init));
 	}
 	async sendAudio(clientId, connectionId, captureId, pcm, activity) {
-		const result = await this.media().clientAudio({
-			query: {
-				clientId,
-				connectionId,
-				captureId
-			},
-			headers: {
-				"content-type": VOICE_MEDIA_CONTENT_TYPE,
-				...activity === void 0 ? {} : {
-					[VOICE_MEDIA_ACTIVITY_STATE_HEADER]: activity.state,
-					[VOICE_MEDIA_ACTIVITY_LEVEL_HEADER]: String(activity.levelDbfs),
-					[VOICE_MEDIA_ACTIVITY_ELAPSED_HEADER]: String(activity.elapsedMs),
-					...activity.epoch === void 0 ? {} : { [VOICE_MEDIA_ACTIVITY_EPOCH_HEADER]: String(activity.epoch) },
-					...activity.classifiedSpeechMs === void 0 ? {} : { [VOICE_MEDIA_ACTIVITY_SPEECH_MS_HEADER]: String(activity.classifiedSpeechMs) },
-					...activity.echoDiscriminatedSpeechMs === void 0 ? {} : { [VOICE_MEDIA_ACTIVITY_ECHO_SPEECH_MS_HEADER]: String(activity.echoDiscriminatedSpeechMs) }
-				}
-			},
-			body: new Blob([new Uint8Array(pcm)], { type: VOICE_MEDIA_CONTENT_TYPE })
+		const query = {
+			clientId,
+			connectionId,
+			captureId
+		};
+		const headers = {
+			"content-type": VOICE_MEDIA_CONTENT_TYPE,
+			...activity === void 0 ? {} : {
+				[VOICE_MEDIA_ACTIVITY_STATE_HEADER]: activity.state,
+				[VOICE_MEDIA_ACTIVITY_LEVEL_HEADER]: String(activity.levelDbfs),
+				[VOICE_MEDIA_ACTIVITY_ELAPSED_HEADER]: String(activity.elapsedMs),
+				...activity.epoch === void 0 ? {} : { [VOICE_MEDIA_ACTIVITY_EPOCH_HEADER]: String(activity.epoch) },
+				...activity.classifiedSpeechMs === void 0 ? {} : { [VOICE_MEDIA_ACTIVITY_SPEECH_MS_HEADER]: String(activity.classifiedSpeechMs) },
+				...activity.echoDiscriminatedSpeechMs === void 0 ? {} : { [VOICE_MEDIA_ACTIVITY_ECHO_SPEECH_MS_HEADER]: String(activity.echoDiscriminatedSpeechMs) }
+			}
+		};
+		const body = new Blob([new Uint8Array(pcm)], { type: VOICE_MEDIA_CONTENT_TYPE });
+		const result = this.remote() ? await this.remoteResult(route(VOICE_MEDIA_ROUTES.clientAudio, query), "", {
+			method: "POST",
+			headers,
+			body
+		}) : await this.media().clientAudio({
+			query,
+			headers,
+			body
 		});
 		if (!result.ok) throw resultError(result);
 	}
 	async captureStopped(clientId, connectionId, captureId, error) {
-		const result = await this.media().clientCaptureStopped({ body: {
+		const body = {
 			clientId,
 			connectionId,
 			captureId,
 			...error === void 0 ? {} : { error }
-		} });
+		};
+		const result = this.remote() ? await this.remoteResult(VOICE_MEDIA_ROUTES.clientCaptureStopped, "", jsonBody(body)) : await this.media().clientCaptureStopped({ body });
 		if (!result.ok) throw resultError(result);
 	}
 	/**
@@ -2906,13 +3016,18 @@ var BrowserVoiceMediaTransport = class {
 	async receivePlaybackAudio(clientId, connectionId, playbackId, signal) {
 		const chunks = [];
 		let byteLength = 0;
-		const url = this.media().clientPlaybackAudio.url({ query: {
+		const path = route(VOICE_MEDIA_ROUTES.clientPlaybackAudio, {
+			clientId,
+			connectionId,
+			playbackId
+		});
+		const url = this.remote() ? "" : this.media().clientPlaybackAudio.url({ query: {
 			clientId,
 			connectionId,
 			playbackId
 		} });
 		while (!signal.aborted) {
-			const response = await sealedTransport.fetch(url, { signal });
+			const response = await this.fetch(path, url, { signal });
 			if (response.status === 200 && response.headers.get("content-type")?.startsWith("application/vnd.doompi.pcm-s16le")) {
 				const chunk = new Uint8Array(await response.arrayBuffer());
 				chunks.push(chunk);
@@ -2934,15 +3049,18 @@ var BrowserVoiceMediaTransport = class {
 		throw new DOMException("Voice playback audio was aborted.", "AbortError");
 	}
 	async playbackFinished(clientId, connectionId, result) {
-		const answer = await this.media().clientPlaybackResult({ body: {
+		const body = {
 			clientId,
 			connectionId,
 			...result
-		} });
+		};
+		const answer = this.remote() ? await this.remoteResult(VOICE_MEDIA_ROUTES.clientPlaybackResult, "", jsonBody(body)) : await this.media().clientPlaybackResult({ body });
 		if (!answer.ok) throw resultError(answer);
 	}
 	async realtimeNegotiate(clientId, connectionId, activationId, sdp, signal) {
-		return boundedControlRequest(this.media().realtimeNegotiate.url(), jsonBody({
+		const path = REALTIME_ROUTES.clientNegotiate;
+		const directUrl = this.remote() ? "" : this.media().realtimeNegotiate.url();
+		return boundedControlRequest(directUrl, jsonBody({
 			clientId,
 			connectionId,
 			activationId,
@@ -2952,10 +3070,10 @@ var BrowserVoiceMediaTransport = class {
 			const body = await response.json();
 			if (typeof body !== "object" || body === null || typeof body.sdp !== "string") throw new Error("Realtime negotiation response is invalid.");
 			return body.sdp;
-		}, signal);
+		}, signal, (_input, init) => this.fetch(path, directUrl, init));
 	}
 	async realtimeEvent(clientId, connectionId, activationId, event) {
-		await this.postRealtime(this.media().realtimeEvent.url(), {
+		await this.postRealtime(REALTIME_ROUTES.clientEvent, this.remote() ? "" : this.media().realtimeEvent.url(), {
 			clientId,
 			connectionId,
 			activationId,
@@ -2963,17 +3081,17 @@ var BrowserVoiceMediaTransport = class {
 		});
 	}
 	async realtimeState(clientId, connectionId, activationId, state) {
-		await this.postRealtime(this.media().realtimeState.url(), {
+		await this.postRealtime(REALTIME_ROUTES.clientState, this.remote() ? "" : this.media().realtimeState.url(), {
 			clientId,
 			connectionId,
 			activationId,
 			state
 		});
 	}
-	postRealtime(url, body) {
-		return boundedControlRequest(url, jsonBody(body), CONTROL_REQUEST_DEADLINE_MS, async (response) => {
+	postRealtime(path, directUrl, body) {
+		return boundedControlRequest(directUrl, jsonBody(body), CONTROL_REQUEST_DEADLINE_MS, async (response) => {
 			if (!response.ok) throw await responseError(response);
-		});
+		}, void 0, (_input, init) => this.fetch(path, directUrl, init));
 	}
 };
 //#endregion