@happyvertical/speech 0.98.0 → 0.99.1

package/dist/conversation.js

@@ -0,0 +1,346 @@
+import { i as SpeechProviderError, n as SpeechConfigurationError } from "./chunks/errors-CQBX3OPS.js";
+//#region src/conversation/openai-webrtc.ts
+var ADAPTER = "openai-conversation";
+/** Browser WebRTC transport. No credentials, Node dependencies or globals are read at import. */
+function createOpenAIWebRTCVoiceSession(options) {
+	let state = "idle";
+	let pending;
+	let peer;
+	let channel;
+	let microphone;
+	let audio;
+	let micMuted = options.micMuted ?? false;
+	let outputMuted = options.outputMuted ?? false;
+	let responseId;
+	let responseActive = false;
+	let speaking = false;
+	const abort = new AbortController();
+	const listeners = /* @__PURE__ */ new Map();
+	const turns = /* @__PURE__ */ new Map();
+	const disposers = [];
+	function emit(event, value) {
+		for (const listener of listeners.get(event) ?? []) listener(value);
+	}
+	function change(next) {
+		state = next;
+		emit("state", next);
+	}
+	function playback(value) {
+		if (speaking === value) return;
+		speaking = value;
+		emit("speaking", value);
+	}
+	function cleanup() {
+		abort.abort();
+		for (const dispose of disposers.splice(0)) dispose();
+		microphone?.getTracks().forEach((track) => {
+			track.stop();
+		});
+		microphone = void 0;
+		if (channel) {
+			channel.onopen = channel.onmessage = channel.onclose = channel.onerror = null;
+			channel.close();
+		}
+		if (peer) {
+			peer.ontrack = peer.onconnectionstatechange = null;
+			peer.close();
+		}
+		if (audio) {
+			audio.pause();
+			audio.srcObject = null;
+		}
+		playback(false);
+		responseActive = false;
+		turns.clear();
+	}
+	function fail(message) {
+		if (state === "closed" || state === "failed") return;
+		state = "failed";
+		cleanup();
+		emit("state", state);
+		emit("error", new SpeechProviderError(ADAPTER, message));
+	}
+	function send(event) {
+		if (state !== "connected" || channel?.readyState !== "open") throw new SpeechConfigurationError("Voice session is not connected", ADAPTER);
+		channel.send(JSON.stringify(event));
+	}
+	function transcript(itemId, role, text, final, delta = false) {
+		if (typeof itemId !== "string" || typeof text !== "string") return;
+		const prior = turns.get(itemId);
+		if (prior?.final || prior && prior.role !== role) return;
+		const turn = {
+			itemId,
+			role,
+			text: delta ? (prior?.text ?? "") + text : text,
+			final
+		};
+		turns.set(itemId, turn);
+		emit("transcript", turn);
+	}
+	function receive(data) {
+		let event;
+		try {
+			const parsed = typeof data === "string" ? JSON.parse(data) : data;
+			if (!parsed || typeof parsed !== "object" || Array.isArray(parsed)) return;
+			event = parsed;
+		} catch {
+			return;
+		}
+		switch (event.type) {
+			case "conversation.item.input_audio_transcription.delta":
+				transcript(event.item_id, "user", event.delta, false, true);
+				break;
+			case "conversation.item.input_audio_transcription.completed":
+				transcript(event.item_id, "user", event.transcript, true);
+				break;
+			case "conversation.item.input_audio_transcription.failed":
+				emit("error", new SpeechProviderError(ADAPTER, "Input transcription failed"));
+				break;
+			case "response.output_audio_transcript.delta":
+				transcript(event.item_id, "assistant", event.delta, false, true);
+				break;
+			case "response.output_audio_transcript.done":
+				transcript(event.item_id, "assistant", event.transcript, true);
+				break;
+			case "response.output_text.delta":
+				transcript(event.item_id, "assistant", event.delta, false, true);
+				break;
+			case "response.output_text.done":
+				transcript(event.item_id, "assistant", event.text, true);
+				break;
+			case "input_audio_buffer.speech_started":
+				emit("inputSpeech", true);
+				break;
+			case "input_audio_buffer.speech_stopped":
+				emit("inputSpeech", false);
+				break;
+			case "output_audio_buffer.started":
+				playback(!outputMuted);
+				break;
+			case "output_audio_buffer.stopped":
+			case "output_audio_buffer.cleared":
+				playback(false);
+				break;
+			case "response.created": {
+				const response = record(event.response);
+				responseId = typeof response.id === "string" ? response.id : void 0;
+				responseActive = true;
+				emit("response", {
+					active: true,
+					responseId
+				});
+				break;
+			}
+			case "response.done": {
+				const response = record(event.response);
+				responseActive = false;
+				emit("response", {
+					active: false,
+					responseId
+				});
+				if (response.usage && typeof response.usage === "object") emit("usage", record(response.usage));
+				if (response.status === "failed" || response.status === "incomplete") emit("error", new SpeechProviderError(ADAPTER, response.status === "incomplete" ? "Voice response incomplete" : "Voice response failed"));
+				break;
+			}
+			case "response.function_call_arguments.done":
+				if (typeof event.call_id === "string" && typeof event.name === "string" && typeof event.arguments === "string") emit("toolCall", {
+					callId: event.call_id,
+					name: event.name,
+					arguments: event.arguments
+				});
+				break;
+			case "error":
+				emit("error", new SpeechProviderError(ADAPTER, "Realtime provider rejected a conversation event"));
+				break;
+		}
+	}
+	const session = {
+		get state() {
+			return state;
+		},
+		connect() {
+			if (pending) return pending;
+			if (state === "connected") return Promise.resolve();
+			if (state !== "idle") return Promise.reject(new SpeechConfigurationError("Create a new voice session after close or failure", ADAPTER));
+			const timeoutMs = options.connectTimeoutMs ?? 2e4;
+			if (!Number.isFinite(timeoutMs) || timeoutMs <= 0) return Promise.reject(new SpeechConfigurationError("connectTimeoutMs must be positive", ADAPTER));
+			change("connecting");
+			const signal = AbortSignal.any([abort.signal, ...options.signal ? [options.signal] : []]);
+			pending = (async () => {
+				const timeout = setTimeout(() => fail("Voice connection timed out"), timeoutMs);
+				const onAbort = () => {
+					if (state !== "closed" && state !== "failed") fail("Voice connection cancelled");
+				};
+				signal.addEventListener("abort", onAbort, { once: true });
+				disposers.push(() => signal.removeEventListener("abort", onAbort));
+				try {
+					signal.throwIfAborted();
+					const stream = await abortable(options.getMicrophone().then((stream) => {
+						if (signal.aborted) stream.getTracks().forEach((track) => {
+							track.stop();
+						});
+						return stream;
+					}), signal);
+					signal.throwIfAborted();
+					microphone = stream;
+					stream.getAudioTracks().forEach((track) => {
+						track.enabled = !micMuted;
+						const ended = () => fail("Microphone ended; allow access and start a new session");
+						track.addEventListener("ended", ended);
+						disposers.push(() => track.removeEventListener("ended", ended));
+						if (track.readyState === "ended") throw new SpeechConfigurationError("Microphone stream has ended", ADAPTER);
+					});
+					if (!stream.getAudioTracks().length) throw new SpeechConfigurationError("Microphone stream has no audio tracks", ADAPTER);
+					peer = options.createPeerConnection?.() ?? new RTCPeerConnection();
+					audio = options.audio ?? new Audio();
+					audio.autoplay = true;
+					audio.muted = outputMuted;
+					peer.ontrack = (event) => {
+						if (!audio || signal.aborted) return;
+						audio.srcObject = event.streams[0] ?? new MediaStream([event.track]);
+						audio.play().catch(() => emit("error", new SpeechProviderError(ADAPTER, "Audio playback needs a user gesture")));
+					};
+					peer.onconnectionstatechange = () => {
+						if (peer?.connectionState === "failed" || peer?.connectionState === "disconnected") fail("Voice connection lost; start a new session");
+					};
+					for (const track of stream.getAudioTracks()) peer.addTrack(track, stream);
+					channel = peer.createDataChannel("oai-events");
+					channel.onmessage = (event) => receive(event.data);
+					const opened = new Promise((resolve, reject) => {
+						const cancel = () => reject(new SpeechProviderError(ADAPTER, "Voice connection cancelled"));
+						signal.addEventListener("abort", cancel, { once: true });
+						disposers.push(() => signal.removeEventListener("abort", cancel));
+						if (channel) {
+							channel.onopen = () => {
+								signal.removeEventListener("abort", cancel);
+								resolve();
+							};
+							channel.onclose = () => {
+								fail("Voice channel closed");
+								reject(new SpeechProviderError(ADAPTER, "Voice channel closed"));
+							};
+							channel.onerror = () => {
+								fail("Voice channel failed");
+								reject(new SpeechProviderError(ADAPTER, "Voice channel failed"));
+							};
+						}
+					});
+					opened.catch(() => void 0);
+					const offer = await abortable(peer.createOffer(), signal);
+					await abortable(peer.setLocalDescription(offer), signal);
+					signal.throwIfAborted();
+					if (!offer.sdp) throw new SpeechConfigurationError("WebRTC did not produce an SDP offer", ADAPTER);
+					const answer = await abortable(options.negotiate(offer.sdp, signal), signal);
+					signal.throwIfAborted();
+					await abortable(peer.setRemoteDescription({
+						type: "answer",
+						sdp: answer
+					}), signal);
+					await opened;
+					signal.throwIfAborted();
+					change("connected");
+				} catch (error) {
+					fail("Unable to connect voice session");
+					throw error;
+				} finally {
+					clearTimeout(timeout);
+				}
+			})().finally(() => {
+				pending = void 0;
+			});
+			return pending;
+		},
+		sendText(text) {
+			const trimmed = text.trim();
+			if (!trimmed || trimmed.length > 16e3) throw new SpeechConfigurationError("Text must contain 1–16000 characters", ADAPTER);
+			const id = `hv_${crypto.randomUUID().replaceAll("-", "").slice(0, 24)}`;
+			send({
+				type: "conversation.item.create",
+				item: {
+					id,
+					type: "message",
+					role: "user",
+					content: [{
+						type: "input_text",
+						text: trimmed
+					}]
+				}
+			});
+			transcript(id, "user", trimmed, true);
+			session.respond();
+			return id;
+		},
+		commitInput(respond = true) {
+			send({ type: "input_audio_buffer.commit" });
+			if (respond) session.respond();
+		},
+		respond(instructions) {
+			send({
+				type: "response.create",
+				...instructions ? { response: { instructions } } : {}
+			});
+		},
+		interrupt() {
+			if (state !== "connected") return;
+			if (responseActive) send({
+				type: "response.cancel",
+				...responseId ? { response_id: responseId } : {}
+			});
+			send({ type: "output_audio_buffer.clear" });
+			playback(false);
+		},
+		setMicMuted(muted) {
+			micMuted = muted;
+			microphone?.getAudioTracks().forEach((track) => {
+				track.enabled = !muted;
+			});
+			if (muted && state === "connected") send({ type: "input_audio_buffer.clear" });
+		},
+		setOutputMuted(muted) {
+			outputMuted = muted;
+			if (audio) audio.muted = muted;
+			if (muted) playback(false);
+		},
+		submitToolResult(callId, result) {
+			send({
+				type: "conversation.item.create",
+				item: {
+					type: "function_call_output",
+					call_id: callId,
+					output: JSON.stringify(result)
+				}
+			});
+		},
+		close() {
+			if (state === "closed") return;
+			state = "closed";
+			cleanup();
+			emit("state", state);
+			listeners.clear();
+		},
+		on(event, listener) {
+			const set = listeners.get(event) ?? /* @__PURE__ */ new Set();
+			set.add(listener);
+			listeners.set(event, set);
+			return () => {
+				set.delete(listener);
+			};
+		}
+	};
+	return session;
+}
+function record(value) {
+	return value && typeof value === "object" && !Array.isArray(value) ? value : {};
+}
+function abortable(promise, signal) {
+	if (signal.aborted) return Promise.reject(signal.reason);
+	return new Promise((resolve, reject) => {
+		const cancel = () => reject(signal.reason);
+		signal.addEventListener("abort", cancel, { once: true });
+		promise.then(resolve, reject).finally(() => signal.removeEventListener("abort", cancel));
+	});
+}
+//#endregion
+export { createOpenAIWebRTCVoiceSession };
+
+//# sourceMappingURL=conversation.js.map

package/dist/conversation.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"conversation.js","names":[],"sources":["../src/conversation/openai-webrtc.ts"],"sourcesContent":["import {\n  SpeechConfigurationError,\n  SpeechProviderError,\n} from '../shared/errors.js';\nimport type {\n  OpenAIWebRTCVoiceOptions,\n  VoiceSession,\n  VoiceSessionEvents,\n  VoiceSessionState,\n  VoiceTranscript,\n} from './types.js';\n\nconst ADAPTER = 'openai-conversation';\n\n/** Browser WebRTC transport. No credentials, Node dependencies or globals are read at import. */\nexport function createOpenAIWebRTCVoiceSession(\n  options: OpenAIWebRTCVoiceOptions,\n): VoiceSession {\n  let state: VoiceSessionState = 'idle';\n  let pending: Promise<void> | undefined;\n  let peer: RTCPeerConnection | undefined;\n  let channel: RTCDataChannel | undefined;\n  let microphone: MediaStream | undefined;\n  let audio: HTMLAudioElement | undefined;\n  let micMuted = options.micMuted ?? false;\n  let outputMuted = options.outputMuted ?? false;\n  let responseId: string | undefined;\n  let responseActive = false;\n  let speaking = false;\n  const abort = new AbortController();\n  const listeners = new Map<\n    keyof VoiceSessionEvents,\n    Set<(value: never) => void>\n  >();\n  const turns = new Map<string, VoiceTranscript>();\n  const disposers: (() => void)[] = [];\n\n  function emit<E extends keyof VoiceSessionEvents>(\n    event: E,\n    value: VoiceSessionEvents[E],\n  ): void {\n    for (const listener of listeners.get(event) ?? []) listener(value as never);\n  }\n  function change(next: VoiceSessionState): void {\n    state = next;\n    emit('state', next);\n  }\n  function playback(value: boolean): void {\n    if (speaking === value) return;\n    speaking = value;\n    emit('speaking', value);\n  }\n  function cleanup(): void {\n    abort.abort();\n    for (const dispose of disposers.splice(0)) dispose();\n    microphone?.getTracks().forEach((track) => {\n      track.stop();\n    });\n    microphone = undefined;\n    if (channel) {\n      channel.onopen =\n        channel.onmessage =\n        channel.onclose =\n        channel.onerror =\n          null;\n      channel.close();\n    }\n    if (peer) {\n      peer.ontrack = peer.onconnectionstatechange = null;\n      peer.close();\n    }\n    if (audio) {\n      audio.pause();\n      audio.srcObject = null;\n    }\n    playback(false);\n    responseActive = false;\n    turns.clear();\n  }\n  function fail(message: string): void {\n    if (state === 'closed' || state === 'failed') return;\n    state = 'failed';\n    cleanup();\n    emit('state', state);\n    emit('error', new SpeechProviderError(ADAPTER, message));\n  }\n  function send(event: Record<string, unknown>): void {\n    if (state !== 'connected' || channel?.readyState !== 'open')\n      throw new SpeechConfigurationError(\n        'Voice session is not connected',\n        ADAPTER,\n      );\n    channel.send(JSON.stringify(event));\n  }\n  function transcript(\n    itemId: unknown,\n    role: VoiceTranscript['role'],\n    text: unknown,\n    final: boolean,\n    delta = false,\n  ): void {\n    if (typeof itemId !== 'string' || typeof text !== 'string') return;\n    const prior = turns.get(itemId);\n    // Late duplicates/deltas must not overwrite a final turn or its role.\n    if (prior?.final || (prior && prior.role !== role)) return;\n    const turn = {\n      itemId,\n      role,\n      text: delta ? (prior?.text ?? '') + text : text,\n      final,\n    };\n    turns.set(itemId, turn);\n    emit('transcript', turn);\n  }\n  function receive(data: unknown): void {\n    let event: Record<string, unknown>;\n    try {\n      const parsed: unknown =\n        typeof data === 'string' ? JSON.parse(data) : data;\n      if (!parsed || typeof parsed !== 'object' || Array.isArray(parsed))\n        return;\n      event = parsed as Record<string, unknown>;\n    } catch {\n      return;\n    }\n    switch (event.type) {\n      case 'conversation.item.input_audio_transcription.delta':\n        transcript(event.item_id, 'user', event.delta, false, true);\n        break;\n      case 'conversation.item.input_audio_transcription.completed':\n        transcript(event.item_id, 'user', event.transcript, true);\n        break;\n      case 'conversation.item.input_audio_transcription.failed':\n        emit(\n          'error',\n          new SpeechProviderError(ADAPTER, 'Input transcription failed'),\n        );\n        break;\n      case 'response.output_audio_transcript.delta':\n        transcript(event.item_id, 'assistant', event.delta, false, true);\n        break;\n      case 'response.output_audio_transcript.done':\n        transcript(event.item_id, 'assistant', event.transcript, true);\n        break;\n      case 'response.output_text.delta':\n        transcript(event.item_id, 'assistant', event.delta, false, true);\n        break;\n      case 'response.output_text.done':\n        transcript(event.item_id, 'assistant', event.text, true);\n        break;\n      case 'input_audio_buffer.speech_started':\n        emit('inputSpeech', true);\n        break;\n      case 'input_audio_buffer.speech_stopped':\n        emit('inputSpeech', false);\n        break;\n      case 'output_audio_buffer.started':\n        playback(!outputMuted);\n        break;\n      case 'output_audio_buffer.stopped':\n      case 'output_audio_buffer.cleared':\n        playback(false);\n        break;\n      case 'response.created': {\n        const response = record(event.response);\n        responseId = typeof response.id === 'string' ? response.id : undefined;\n        responseActive = true;\n        emit('response', { active: true, responseId });\n        break;\n      }\n      case 'response.done': {\n        const response = record(event.response);\n        responseActive = false;\n        emit('response', { active: false, responseId });\n        if (response.usage && typeof response.usage === 'object')\n          emit('usage', record(response.usage));\n        if (response.status === 'failed' || response.status === 'incomplete')\n          emit(\n            'error',\n            new SpeechProviderError(\n              ADAPTER,\n              response.status === 'incomplete'\n                ? 'Voice response incomplete'\n                : 'Voice response failed',\n            ),\n          );\n        break;\n      }\n      case 'response.function_call_arguments.done':\n        if (\n          typeof event.call_id === 'string' &&\n          typeof event.name === 'string' &&\n          typeof event.arguments === 'string'\n        )\n          emit('toolCall', {\n            callId: event.call_id,\n            name: event.name,\n            arguments: event.arguments,\n          });\n        break;\n      case 'error':\n        emit(\n          'error',\n          new SpeechProviderError(\n            ADAPTER,\n            'Realtime provider rejected a conversation event',\n          ),\n        );\n        break;\n    }\n  }\n\n  const session: VoiceSession = {\n    get state() {\n      return state;\n    },\n    connect() {\n      if (pending) return pending;\n      if (state === 'connected') return Promise.resolve();\n      if (state !== 'idle')\n        return Promise.reject(\n          new SpeechConfigurationError(\n            'Create a new voice session after close or failure',\n            ADAPTER,\n          ),\n        );\n      const timeoutMs = options.connectTimeoutMs ?? 20_000;\n      if (!Number.isFinite(timeoutMs) || timeoutMs <= 0)\n        return Promise.reject(\n          new SpeechConfigurationError(\n            'connectTimeoutMs must be positive',\n            ADAPTER,\n          ),\n        );\n      change('connecting');\n      const signal = AbortSignal.any([\n        abort.signal,\n        ...(options.signal ? [options.signal] : []),\n      ]);\n      pending = (async () => {\n        const timeout = setTimeout(\n          () => fail('Voice connection timed out'),\n          timeoutMs,\n        );\n        const onAbort = () => {\n          if (state !== 'closed' && state !== 'failed')\n            fail('Voice connection cancelled');\n        };\n        signal.addEventListener('abort', onAbort, { once: true });\n        disposers.push(() => signal.removeEventListener('abort', onAbort));\n        try {\n          signal.throwIfAborted();\n          // Permission acquisition cannot be aborted; dispose even if it resolves after close.\n          const acquisition = options.getMicrophone().then((stream) => {\n            if (signal.aborted)\n              stream.getTracks().forEach((track) => {\n                track.stop();\n              });\n            return stream;\n          });\n          const stream = await abortable(acquisition, signal);\n          signal.throwIfAborted();\n          microphone = stream;\n          stream.getAudioTracks().forEach((track) => {\n            track.enabled = !micMuted;\n            const ended = () =>\n              fail('Microphone ended; allow access and start a new session');\n            track.addEventListener('ended', ended);\n            disposers.push(() => track.removeEventListener('ended', ended));\n            if (track.readyState === 'ended')\n              throw new SpeechConfigurationError(\n                'Microphone stream has ended',\n                ADAPTER,\n              );\n          });\n          if (!stream.getAudioTracks().length)\n            throw new SpeechConfigurationError(\n              'Microphone stream has no audio tracks',\n              ADAPTER,\n            );\n          peer = options.createPeerConnection?.() ?? new RTCPeerConnection();\n          audio = options.audio ?? new Audio();\n          audio.autoplay = true;\n          audio.muted = outputMuted;\n          peer.ontrack = (event) => {\n            if (!audio || signal.aborted) return;\n            audio.srcObject =\n              event.streams[0] ?? new MediaStream([event.track]);\n            void audio\n              .play()\n              .catch(() =>\n                emit(\n                  'error',\n                  new SpeechProviderError(\n                    ADAPTER,\n                    'Audio playback needs a user gesture',\n                  ),\n                ),\n              );\n          };\n          peer.onconnectionstatechange = () => {\n            if (\n              peer?.connectionState === 'failed' ||\n              peer?.connectionState === 'disconnected'\n            )\n              fail('Voice connection lost; start a new session');\n          };\n          for (const track of stream.getAudioTracks())\n            peer.addTrack(track, stream);\n          channel = peer.createDataChannel('oai-events');\n          channel.onmessage = (event) => receive(event.data);\n          // Attach before SDP negotiation; some transports open during setRemoteDescription.\n          const opened = new Promise<void>((resolve, reject) => {\n            const cancel = () =>\n              reject(\n                new SpeechProviderError(ADAPTER, 'Voice connection cancelled'),\n              );\n            signal.addEventListener('abort', cancel, { once: true });\n            disposers.push(() => signal.removeEventListener('abort', cancel));\n            if (channel) {\n              channel.onopen = () => {\n                signal.removeEventListener('abort', cancel);\n                resolve();\n              };\n              channel.onclose = () => {\n                fail('Voice channel closed');\n                reject(\n                  new SpeechProviderError(ADAPTER, 'Voice channel closed'),\n                );\n              };\n              channel.onerror = () => {\n                fail('Voice channel failed');\n                reject(\n                  new SpeechProviderError(ADAPTER, 'Voice channel failed'),\n                );\n              };\n            }\n          });\n          // Keep an early rejection handled while negotiation is pending.\n          void opened.catch(() => undefined);\n          const offer = await abortable(peer.createOffer(), signal);\n          await abortable(peer.setLocalDescription(offer), signal);\n          signal.throwIfAborted();\n          if (!offer.sdp)\n            throw new SpeechConfigurationError(\n              'WebRTC did not produce an SDP offer',\n              ADAPTER,\n            );\n          const answer = await abortable(\n            options.negotiate(offer.sdp, signal),\n            signal,\n          );\n          signal.throwIfAborted();\n          await abortable(\n            peer.setRemoteDescription({ type: 'answer', sdp: answer }),\n            signal,\n          );\n          await opened;\n          signal.throwIfAborted();\n          change('connected');\n        } catch (error) {\n          fail('Unable to connect voice session');\n          throw error;\n        } finally {\n          clearTimeout(timeout);\n        }\n      })().finally(() => {\n        pending = undefined;\n      });\n      return pending;\n    },\n    sendText(text) {\n      const trimmed = text.trim();\n      if (!trimmed || trimmed.length > 16_000)\n        throw new SpeechConfigurationError(\n          'Text must contain 1–16000 characters',\n          ADAPTER,\n        );\n      const id = `hv_${crypto.randomUUID().replaceAll('-', '').slice(0, 24)}`;\n      send({\n        type: 'conversation.item.create',\n        item: {\n          id,\n          type: 'message',\n          role: 'user',\n          content: [{ type: 'input_text', text: trimmed }],\n        },\n      });\n      transcript(id, 'user', trimmed, true);\n      session.respond();\n      return id;\n    },\n    commitInput(respond = true) {\n      send({ type: 'input_audio_buffer.commit' });\n      if (respond) session.respond();\n    },\n    respond(instructions) {\n      send({\n        type: 'response.create',\n        ...(instructions ? { response: { instructions } } : {}),\n      });\n    },\n    interrupt() {\n      if (state !== 'connected') return;\n      if (responseActive)\n        send({\n          type: 'response.cancel',\n          ...(responseId ? { response_id: responseId } : {}),\n        });\n      send({ type: 'output_audio_buffer.clear' });\n      playback(false);\n    },\n    setMicMuted(muted) {\n      micMuted = muted;\n      microphone?.getAudioTracks().forEach((track) => {\n        track.enabled = !muted;\n      });\n      if (muted && state === 'connected')\n        send({ type: 'input_audio_buffer.clear' });\n    },\n    setOutputMuted(muted) {\n      outputMuted = muted;\n      if (audio) audio.muted = muted;\n      if (muted) playback(false);\n    },\n    submitToolResult(callId, result) {\n      send({\n        type: 'conversation.item.create',\n        item: {\n          type: 'function_call_output',\n          call_id: callId,\n          output: JSON.stringify(result),\n        },\n      });\n    },\n    close() {\n      if (state === 'closed') return;\n      state = 'closed';\n      cleanup();\n      emit('state', state);\n      listeners.clear();\n    },\n    on(event, listener) {\n      const set = listeners.get(event) ?? new Set();\n      set.add(listener as (value: never) => void);\n      listeners.set(event, set);\n      return () => {\n        set.delete(listener as (value: never) => void);\n      };\n    },\n  };\n  return session;\n}\n\nfunction record(value: unknown): Record<string, unknown> {\n  return value && typeof value === 'object' && !Array.isArray(value)\n    ? (value as Record<string, unknown>)\n    : {};\n}\n\nfunction abortable<T>(promise: Promise<T>, signal: AbortSignal): Promise<T> {\n  if (signal.aborted) return Promise.reject(signal.reason);\n  return new Promise((resolve, reject) => {\n    const cancel = () => reject(signal.reason);\n    signal.addEventListener('abort', cancel, { once: true });\n    promise\n      .then(resolve, reject)\n      .finally(() => signal.removeEventListener('abort', cancel));\n  });\n}\n"],"mappings":";;AAYA,IAAM,UAAU;;AAGhB,SAAgB,+BACd,SACc;CACd,IAAI,QAA2B;CAC/B,IAAI;CACJ,IAAI;CACJ,IAAI;CACJ,IAAI;CACJ,IAAI;CACJ,IAAI,WAAW,QAAQ,YAAY;CACnC,IAAI,cAAc,QAAQ,eAAe;CACzC,IAAI;CACJ,IAAI,iBAAiB;CACrB,IAAI,WAAW;CACf,MAAM,QAAQ,IAAI,gBAAgB;CAClC,MAAM,4BAAY,IAAI,IAGpB;CACF,MAAM,wBAAQ,IAAI,IAA6B;CAC/C,MAAM,YAA4B,CAAC;CAEnC,SAAS,KACP,OACA,OACM;EACN,KAAK,MAAM,YAAY,UAAU,IAAI,KAAK,KAAK,CAAC,GAAG,SAAS,KAAc;CAC5E;CACA,SAAS,OAAO,MAA+B;EAC7C,QAAQ;EACR,KAAK,SAAS,IAAI;CACpB;CACA,SAAS,SAAS,OAAsB;EACtC,IAAI,aAAa,OAAO;EACxB,WAAW;EACX,KAAK,YAAY,KAAK;CACxB;CACA,SAAS,UAAgB;EACvB,MAAM,MAAM;EACZ,KAAK,MAAM,WAAW,UAAU,OAAO,CAAC,GAAG,QAAQ;EACnD,YAAY,UAAU,CAAC,CAAC,SAAS,UAAU;GACzC,MAAM,KAAK;EACb,CAAC;EACD,aAAa,KAAA;EACb,IAAI,SAAS;GACX,QAAQ,SACN,QAAQ,YACR,QAAQ,UACR,QAAQ,UACN;GACJ,QAAQ,MAAM;EAChB;EACA,IAAI,MAAM;GACR,KAAK,UAAU,KAAK,0BAA0B;GAC9C,KAAK,MAAM;EACb;EACA,IAAI,OAAO;GACT,MAAM,MAAM;GACZ,MAAM,YAAY;EACpB;EACA,SAAS,KAAK;EACd,iBAAiB;EACjB,MAAM,MAAM;CACd;CACA,SAAS,KAAK,SAAuB;EACnC,IAAI,UAAU,YAAY,UAAU,UAAU;EAC9C,QAAQ;EACR,QAAQ;EACR,KAAK,SAAS,KAAK;EACnB,KAAK,SAAS,IAAI,oBAAoB,SAAS,OAAO,CAAC;CACzD;CACA,SAAS,KAAK,OAAsC;EAClD,IAAI,UAAU,eAAe,SAAS,eAAe,QACnD,MAAM,IAAI,yBACR,kCACA,OACF;EACF,QAAQ,KAAK,KAAK,UAAU,KAAK,CAAC;CACpC;CACA,SAAS,WACP,QACA,MACA,MACA,OACA,QAAQ,OACF;EACN,IAAI,OAAO,WAAW,YAAY,OAAO,SAAS,UAAU;EAC5D,MAAM,QAAQ,MAAM,IAAI,MAAM;EAE9B,IAAI,OAAO,SAAU,SAAS,MAAM,SAAS,MAAO;EACpD,MAAM,OAAO;GACX;GACA;GACA,MAAM,SAAS,OAAO,QAAQ,MAAM,OAAO;GAC3C;EACF;EACA,MAAM,IAAI,QAAQ,IAAI;EACtB,KAAK,cAAc,IAAI;CACzB;CACA,SAAS,QAAQ,MAAqB;EACpC,IAAI;EACJ,IAAI;GACF,MAAM,SACJ,OAAO,SAAS,WAAW,KAAK,MAAM,IAAI,IAAI;GAChD,IAAI,CAAC,UAAU,OAAO,WAAW,YAAY,MAAM,QAAQ,MAAM,GAC/D;GACF,QAAQ;EACV,QAAQ;GACN;EACF;EACA,QAAQ,MAAM,MAAd;GACE,KAAK;IACH,WAAW,MAAM,SAAS,QAAQ,MAAM,OAAO,OAAO,IAAI;IAC1D;GACF,KAAK;IACH,WAAW,MAAM,SAAS,QAAQ,MAAM,YAAY,IAAI;IACxD;GACF,KAAK;IACH,KACE,SACA,IAAI,oBAAoB,SAAS,4BAA4B,CAC/D;IACA;GACF,KAAK;IACH,WAAW,MAAM,SAAS,aAAa,MAAM,OAAO,OAAO,IAAI;IAC/D;GACF,KAAK;IACH,WAAW,MAAM,SAAS,aAAa,MAAM,YAAY,IAAI;IAC7D;GACF,KAAK;IACH,WAAW,MAAM,SAAS,aAAa,MAAM,OAAO,OAAO,IAAI;IAC/D;GACF,KAAK;IACH,WAAW,MAAM,SAAS,aAAa,MAAM,MAAM,IAAI;IACvD;GACF,KAAK;IACH,KAAK,eAAe,IAAI;IACxB;GACF,KAAK;IACH,KAAK,eAAe,KAAK;IACzB;GACF,KAAK;IACH,SAAS,CAAC,WAAW;IACrB;GACF,KAAK;GACL,KAAK;IACH,SAAS,KAAK;IACd;GACF,KAAK,oBAAoB;IACvB,MAAM,WAAW,OAAO,MAAM,QAAQ;IACtC,aAAa,OAAO,SAAS,OAAO,WAAW,SAAS,KAAK,KAAA;IAC7D,iBAAiB;IACjB,KAAK,YAAY;KAAE,QAAQ;KAAM;IAAW,CAAC;IAC7C;GACF;GACA,KAAK,iBAAiB;IACpB,MAAM,WAAW,OAAO,MAAM,QAAQ;IACtC,iBAAiB;IACjB,KAAK,YAAY;KAAE,QAAQ;KAAO;IAAW,CAAC;IAC9C,IAAI,SAAS,SAAS,OAAO,SAAS,UAAU,UAC9C,KAAK,SAAS,OAAO,SAAS,KAAK,CAAC;IACtC,IAAI,SAAS,WAAW,YAAY,SAAS,WAAW,cACtD,KACE,SACA,IAAI,oBACF,SACA,SAAS,WAAW,eAChB,8BACA,uBACN,CACF;IACF;GACF;GACA,KAAK;IACH,IACE,OAAO,MAAM,YAAY,YACzB,OAAO,MAAM,SAAS,YACtB,OAAO,MAAM,cAAc,UAE3B,KAAK,YAAY;KACf,QAAQ,MAAM;KACd,MAAM,MAAM;KACZ,WAAW,MAAM;IACnB,CAAC;IACH;GACF,KAAK;IACH,KACE,SACA,IAAI,oBACF,SACA,iDACF,CACF;IACA;EACJ;CACF;CAEA,MAAM,UAAwB;EAC5B,IAAI,QAAQ;GACV,OAAO;EACT;EACA,UAAU;GACR,IAAI,SAAS,OAAO;GACpB,IAAI,UAAU,aAAa,OAAO,QAAQ,QAAQ;GAClD,IAAI,UAAU,QACZ,OAAO,QAAQ,OACb,IAAI,yBACF,qDACA,OACF,CACF;GACF,MAAM,YAAY,QAAQ,oBAAoB;GAC9C,IAAI,CAAC,OAAO,SAAS,SAAS,KAAK,aAAa,GAC9C,OAAO,QAAQ,OACb,IAAI,yBACF,qCACA,OACF,CACF;GACF,OAAO,YAAY;GACnB,MAAM,SAAS,YAAY,IAAI,CAC7B,MAAM,QACN,GAAI,QAAQ,SAAS,CAAC,QAAQ,MAAM,IAAI,CAAC,CAC3C,CAAC;GACD,WAAW,YAAY;IACrB,MAAM,UAAU,iBACR,KAAK,4BAA4B,GACvC,SACF;IACA,MAAM,gBAAgB;KACpB,IAAI,UAAU,YAAY,UAAU,UAClC,KAAK,4BAA4B;IACrC;IACA,OAAO,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;IACxD,UAAU,WAAW,OAAO,oBAAoB,SAAS,OAAO,CAAC;IACjE,IAAI;KACF,OAAO,eAAe;KAStB,MAAM,SAAS,MAAM,UAPD,QAAQ,cAAc,CAAC,CAAC,MAAM,WAAW;MAC3D,IAAI,OAAO,SACT,OAAO,UAAU,CAAC,CAAC,SAAS,UAAU;OACpC,MAAM,KAAK;MACb,CAAC;MACH,OAAO;KACT,CAC+B,GAAa,MAAM;KAClD,OAAO,eAAe;KACtB,aAAa;KACb,OAAO,eAAe,CAAC,CAAC,SAAS,UAAU;MACzC,MAAM,UAAU,CAAC;MACjB,MAAM,cACJ,KAAK,wDAAwD;MAC/D,MAAM,iBAAiB,SAAS,KAAK;MACrC,UAAU,WAAW,MAAM,oBAAoB,SAAS,KAAK,CAAC;MAC9D,IAAI,MAAM,eAAe,SACvB,MAAM,IAAI,yBACR,+BACA,OACF;KACJ,CAAC;KACD,IAAI,CAAC,OAAO,eAAe,CAAC,CAAC,QAC3B,MAAM,IAAI,yBACR,yCACA,OACF;KACF,OAAO,QAAQ,uBAAuB,KAAK,IAAI,kBAAkB;KACjE,QAAQ,QAAQ,SAAS,IAAI,MAAM;KACnC,MAAM,WAAW;KACjB,MAAM,QAAQ;KACd,KAAK,WAAW,UAAU;MACxB,IAAI,CAAC,SAAS,OAAO,SAAS;MAC9B,MAAM,YACJ,MAAM,QAAQ,MAAM,IAAI,YAAY,CAAC,MAAM,KAAK,CAAC;MACnD,MACG,KAAK,CAAC,CACN,YACC,KACE,SACA,IAAI,oBACF,SACA,qCACF,CACF,CACF;KACJ;KACA,KAAK,gCAAgC;MACnC,IACE,MAAM,oBAAoB,YAC1B,MAAM,oBAAoB,gBAE1B,KAAK,4CAA4C;KACrD;KACA,KAAK,MAAM,SAAS,OAAO,eAAe,GACxC,KAAK,SAAS,OAAO,MAAM;KAC7B,UAAU,KAAK,kBAAkB,YAAY;KAC7C,QAAQ,aAAa,UAAU,QAAQ,MAAM,IAAI;KAEjD,MAAM,SAAS,IAAI,SAAe,SAAS,WAAW;MACpD,MAAM,eACJ,OACE,IAAI,oBAAoB,SAAS,4BAA4B,CAC/D;MACF,OAAO,iBAAiB,SAAS,QAAQ,EAAE,MAAM,KAAK,CAAC;MACvD,UAAU,WAAW,OAAO,oBAAoB,SAAS,MAAM,CAAC;MAChE,IAAI,SAAS;OACX,QAAQ,eAAe;QACrB,OAAO,oBAAoB,SAAS,MAAM;QAC1C,QAAQ;OACV;OACA,QAAQ,gBAAgB;QACtB,KAAK,sBAAsB;QAC3B,OACE,IAAI,oBAAoB,SAAS,sBAAsB,CACzD;OACF;OACA,QAAQ,gBAAgB;QACtB,KAAK,sBAAsB;QAC3B,OACE,IAAI,oBAAoB,SAAS,sBAAsB,CACzD;OACF;MACF;KACF,CAAC;KAED,OAAY,YAAY,KAAA,CAAS;KACjC,MAAM,QAAQ,MAAM,UAAU,KAAK,YAAY,GAAG,MAAM;KACxD,MAAM,UAAU,KAAK,oBAAoB,KAAK,GAAG,MAAM;KACvD,OAAO,eAAe;KACtB,IAAI,CAAC,MAAM,KACT,MAAM,IAAI,yBACR,uCACA,OACF;KACF,MAAM,SAAS,MAAM,UACnB,QAAQ,UAAU,MAAM,KAAK,MAAM,GACnC,MACF;KACA,OAAO,eAAe;KACtB,MAAM,UACJ,KAAK,qBAAqB;MAAE,MAAM;MAAU,KAAK;KAAO,CAAC,GACzD,MACF;KACA,MAAM;KACN,OAAO,eAAe;KACtB,OAAO,WAAW;IACpB,SAAS,OAAO;KACd,KAAK,iCAAiC;KACtC,MAAM;IACR,UAAU;KACR,aAAa,OAAO;IACtB;GACF,EAAA,CAAG,CAAC,CAAC,cAAc;IACjB,UAAU,KAAA;GACZ,CAAC;GACD,OAAO;EACT;EACA,SAAS,MAAM;GACb,MAAM,UAAU,KAAK,KAAK;GAC1B,IAAI,CAAC,WAAW,QAAQ,SAAS,MAC/B,MAAM,IAAI,yBACR,wCACA,OACF;GACF,MAAM,KAAK,MAAM,OAAO,WAAW,CAAC,CAAC,WAAW,KAAK,EAAE,CAAC,CAAC,MAAM,GAAG,EAAE;GACpE,KAAK;IACH,MAAM;IACN,MAAM;KACJ;KACA,MAAM;KACN,MAAM;KACN,SAAS,CAAC;MAAE,MAAM;MAAc,MAAM;KAAQ,CAAC;IACjD;GACF,CAAC;GACD,WAAW,IAAI,QAAQ,SAAS,IAAI;GACpC,QAAQ,QAAQ;GAChB,OAAO;EACT;EACA,YAAY,UAAU,MAAM;GAC1B,KAAK,EAAE,MAAM,4BAA4B,CAAC;GAC1C,IAAI,SAAS,QAAQ,QAAQ;EAC/B;EACA,QAAQ,cAAc;GACpB,KAAK;IACH,MAAM;IACN,GAAI,eAAe,EAAE,UAAU,EAAE,aAAa,EAAE,IAAI,CAAC;GACvD,CAAC;EACH;EACA,YAAY;GACV,IAAI,UAAU,aAAa;GAC3B,IAAI,gBACF,KAAK;IACH,MAAM;IACN,GAAI,aAAa,EAAE,aAAa,WAAW,IAAI,CAAC;GAClD,CAAC;GACH,KAAK,EAAE,MAAM,4BAA4B,CAAC;GAC1C,SAAS,KAAK;EAChB;EACA,YAAY,OAAO;GACjB,WAAW;GACX,YAAY,eAAe,CAAC,CAAC,SAAS,UAAU;IAC9C,MAAM,UAAU,CAAC;GACnB,CAAC;GACD,IAAI,SAAS,UAAU,aACrB,KAAK,EAAE,MAAM,2BAA2B,CAAC;EAC7C;EACA,eAAe,OAAO;GACpB,cAAc;GACd,IAAI,OAAO,MAAM,QAAQ;GACzB,IAAI,OAAO,SAAS,KAAK;EAC3B;EACA,iBAAiB,QAAQ,QAAQ;GAC/B,KAAK;IACH,MAAM;IACN,MAAM;KACJ,MAAM;KACN,SAAS;KACT,QAAQ,KAAK,UAAU,MAAM;IAC/B;GACF,CAAC;EACH;EACA,QAAQ;GACN,IAAI,UAAU,UAAU;GACxB,QAAQ;GACR,QAAQ;GACR,KAAK,SAAS,KAAK;GACnB,UAAU,MAAM;EAClB;EACA,GAAG,OAAO,UAAU;GAClB,MAAM,MAAM,UAAU,IAAI,KAAK,qBAAK,IAAI,IAAI;GAC5C,IAAI,IAAI,QAAkC;GAC1C,UAAU,IAAI,OAAO,GAAG;GACxB,aAAa;IACX,IAAI,OAAO,QAAkC;GAC/C;EACF;CACF;CACA,OAAO;AACT;AAEA,SAAS,OAAO,OAAyC;CACvD,OAAO,SAAS,OAAO,UAAU,YAAY,CAAC,MAAM,QAAQ,KAAK,IAC5D,QACD,CAAC;AACP;AAEA,SAAS,UAAa,SAAqB,QAAiC;CAC1E,IAAI,OAAO,SAAS,OAAO,QAAQ,OAAO,OAAO,MAAM;CACvD,OAAO,IAAI,SAAS,SAAS,WAAW;EACtC,MAAM,eAAe,OAAO,OAAO,MAAM;EACzC,OAAO,iBAAiB,SAAS,QAAQ,EAAE,MAAM,KAAK,CAAC;EACvD,QACG,KAAK,SAAS,MAAM,CAAC,CACrB,cAAc,OAAO,oBAAoB,SAAS,MAAM,CAAC;CAC9D,CAAC;AACH"}

package/dist/index.js

@@ -1,4 +1,6 @@
-import { C as SpeechProviderError, S as SpeechError, _ as mimeTypeToAudioExtension, a as TRANSCRIBER_ENV_KEYS, b as InvalidSpeechAdapterError, c as hasTranscriberEnv, d as parseOptionalInteger, f as readEnv, g as DEFAULT_MAX_AUDIO_BYTES, h as reportSpeechUsage, i as registeredOptionalTranscribers, l as mergeHeaderInits, m as audioSecondsFromProviderUsage, n as optionalTranscriberEntry, o as defaultEnv, p as resolveTranscriberConfig, s as hasAnyEnv, t as getOptionalTranscriberFactory, u as parseHeadersEnv, v as normalizeAudioInput, x as SpeechConfigurationError, y as normalizeMimeType } from "./chunks/registry-DfID41Ew.js";
+import { i as SpeechProviderError, n as SpeechConfigurationError, r as SpeechError, t as InvalidSpeechAdapterError } from "./chunks/errors-CQBX3OPS.js";
+import { i as toWebSocketUrl, n as messageDataToText, r as resolveWebSocketFactory, t as isBrowserRuntime } from "./chunks/websocket-CR9uXAuS.js";
+import { _ as mimeTypeToAudioExtension, a as TRANSCRIBER_ENV_KEYS, c as hasTranscriberEnv, d as parseOptionalInteger, f as readEnv, g as DEFAULT_MAX_AUDIO_BYTES, h as reportSpeechUsage, i as registeredOptionalTranscribers, l as mergeHeaderInits, m as audioSecondsFromProviderUsage, n as optionalTranscriberEntry, o as defaultEnv, p as resolveTranscriberConfig, s as hasAnyEnv, t as getOptionalTranscriberFactory, u as parseHeadersEnv, v as normalizeAudioInput, y as normalizeMimeType } from "./chunks/registry-C_5BakmM.js";
 //#region src/shared/retry.ts
 /**
 * Retry with exponential backoff for speech provider calls. Retries only
@@ -375,46 +377,6 @@ function parseHeaderNumber(response, headerName) {
 	return Number.isFinite(parsed) ? parsed : void 0;
 }
 //#endregion
-//#region src/shared/websocket.ts
-/**
-* Minimal WebSocket surface used by streaming adapters. Browser, Node 22+
-* (`globalThis.WebSocket`), Deno, Bun, and the `ws` package all satisfy it.
-*/
-/** True in browsers and web workers, where API keys must never be used. */
-function isBrowserRuntime() {
-	const scope = globalThis;
-	if (scope.process?.versions?.node) return false;
-	return typeof scope.document !== "undefined" || typeof scope.WorkerGlobalScope !== "undefined";
-}
-/**
-* Resolves a socket factory from explicit options, falling back to
-* `globalThis.WebSocket`. Headers are only passed when present, so standard
-* constructors that cannot send them are never handed an options object.
-*/
-function resolveWebSocketFactory(adapter, options) {
-	if (options.createWebSocket) return options.createWebSocket;
-	const Ctor = options.WebSocket ?? globalThis.WebSocket;
-	if (!Ctor) throw new SpeechConfigurationError("No WebSocket implementation available; pass `WebSocket` or `createWebSocket`", adapter);
-	return (url, init) => Object.keys(init.headers).length > 0 ? new Ctor(url, {
-		protocols: init.protocols,
-		headers: init.headers
-	}) : new Ctor(url, init.protocols);
-}
-/** Converts `http(s)://` to `ws(s)://`; leaves `ws(s)://` unchanged. */
-function toWebSocketUrl(url, adapter) {
-	const parsed = new URL(url);
-	if (parsed.protocol === "http:") parsed.protocol = "ws:";
-	else if (parsed.protocol === "https:") parsed.protocol = "wss:";
-	else if (parsed.protocol !== "ws:" && parsed.protocol !== "wss:") throw new SpeechConfigurationError(`Unsupported streaming URL protocol: ${parsed.protocol}`, adapter);
-	return parsed.toString();
-}
-/** Converts a message payload to text; returns `undefined` for unsupported data. */
-function messageDataToText(data) {
-	if (typeof data === "string") return data;
-	if (data instanceof ArrayBuffer) return new TextDecoder().decode(data);
-	if (ArrayBuffer.isView(data)) return new TextDecoder().decode(data);
-}
-//#endregion
 //#region src/shared/realtime-auth.ts
 /**
 * Credential handling shared by realtime (WebSocket) adapters.