@cartesia/cartesia-js 1.0.0-alpha.1 → 1.0.0-alpha.2

package/dist/tts/source.js

@@ -1,6 +1,6 @@
 import {
   Source
-} from "../chunk-JOHSCOLW.js";
+} from "../chunk-PQ6CIPFW.js";
 import "../chunk-WIFMLPT5.js";
 export {
   Source as default

package/dist/tts/utils.cjs

@@ -42,13 +42,19 @@ __export(utils_exports, {
 module.exports = __toCommonJS(utils_exports);
 var import_base64_js = __toESM(require("base64-js"), 1);
 function base64ToArray(b64) {
-  return filterSentinel(b64).reduce((acc, b) => {
-    const floats = new Float32Array(import_base64_js.default.toByteArray(b).buffer);
-    const newAcc = new Float32Array(acc.length + floats.length);
-    newAcc.set(acc, 0);
-    newAcc.set(floats, acc.length);
-    return newAcc;
-  }, new Float32Array(0));
+  const byteArrays = filterSentinel(b64).map((b) => import_base64_js.default.toByteArray(b));
+  const totalLength = byteArrays.reduce(
+    (acc, arr) => acc + arr.length / Float32Array.BYTES_PER_ELEMENT,
+    0
+  );
+  const result = new Float32Array(totalLength);
+  let offset = 0;
+  for (const arr of byteArrays) {
+    const floats = new Float32Array(arr.buffer);
+    result.set(floats, offset);
+    offset += floats.length;
+  }
+  return result;
 }
 function playAudioBuffer(floats, context, startAt, sampleRate) {
   const source = context.createBufferSource();

package/dist/tts/utils.js

@@ -7,7 +7,7 @@ import {
   isComplete,
   isSentinel,
   playAudioBuffer
-} from "../chunk-LYPTISWL.js";
+} from "../chunk-RO7TY474.js";
 import "../chunk-WIFMLPT5.js";
 export {
   base64ToArray,

package/dist/tts/websocket.cjs

@@ -100,7 +100,8 @@ var import_partysocket = require("partysocket");
 var import_cross_fetch = __toESM(require("cross-fetch"), 1);
 
 // src/lib/constants.ts
-var BASE_URL = "https://api.cartesia.ai/v0";
+var BASE_URL = "https://api.cartesia.ai/";
+var CARTESIA_VERSION = "2024-06-10";
 var constructApiUrl = (baseUrl, path, protocol) => {
   const normalizedPath = path.startsWith("/") ? path : `/${path}`;
   if (!protocol) {
@@ -125,7 +126,8 @@ var Client = class {
     const url = constructApiUrl(this.baseUrl, path);
     return (0, import_cross_fetch.default)(url.toString(), __spreadProps(__spreadValues({}, options), {
       headers: __spreadValues({
-        "X-API-KEY": this.apiKey
+        "X-API-Key": this.apiKey,
+        "Cartesia-Version": CARTESIA_VERSION
       }, options.headers)
     }));
   }
@@ -133,7 +135,7 @@ var Client = class {
 
 // src/tts/source.ts
 var import_emittery = __toESM(require("emittery"), 1);
-var _emitter, _buffer, _readIndex, _closed, _sampleRate;
+var _emitter, _buffer, _readIndex, _writeIndex, _closed, _sampleRate;
 var Source = class {
   /**
    * Create a new Source.
@@ -143,8 +145,9 @@ var Source = class {
    */
   constructor({ sampleRate }) {
     __privateAdd(this, _emitter, new import_emittery.default());
-    __privateAdd(this, _buffer, new Float32Array());
+    __privateAdd(this, _buffer, void 0);
     __privateAdd(this, _readIndex, 0);
+    __privateAdd(this, _writeIndex, 0);
     __privateAdd(this, _closed, false);
     __privateAdd(this, _sampleRate, void 0);
     this.on = __privateGet(this, _emitter).on.bind(__privateGet(this, _emitter));
@@ -152,6 +155,7 @@ var Source = class {
     this.events = __privateGet(this, _emitter).events.bind(__privateGet(this, _emitter));
     this.off = __privateGet(this, _emitter).off.bind(__privateGet(this, _emitter));
     __privateSet(this, _sampleRate, sampleRate);
+    __privateSet(this, _buffer, new Float32Array(1024));
   }
   get sampleRate() {
     return __privateGet(this, _sampleRate);
@@ -163,7 +167,18 @@ var Source = class {
    */
   enqueue(src) {
     return __async(this, null, function* () {
-      __privateSet(this, _buffer, new Float32Array([...__privateGet(this, _buffer), ...src]));
+      const requiredCapacity = __privateGet(this, _writeIndex) + src.length;
+      if (requiredCapacity > __privateGet(this, _buffer).length) {
+        let newCapacity = __privateGet(this, _buffer).length;
+        while (newCapacity < requiredCapacity) {
+          newCapacity *= 2;
+        }
+        const newBuffer = new Float32Array(newCapacity);
+        newBuffer.set(__privateGet(this, _buffer));
+        __privateSet(this, _buffer, newBuffer);
+      }
+      __privateGet(this, _buffer).set(src, __privateGet(this, _writeIndex));
+      __privateSet(this, _writeIndex, __privateGet(this, _writeIndex) + src.length);
       yield __privateGet(this, _emitter).emit("enqueue");
     });
   }
@@ -177,7 +192,7 @@ var Source = class {
   read(dst) {
     return __async(this, null, function* () {
       const targetReadIndex = __privateGet(this, _readIndex) + dst.length;
-      while (!__privateGet(this, _closed) && targetReadIndex > __privateGet(this, _buffer).length) {
+      while (!__privateGet(this, _closed) && targetReadIndex > __privateGet(this, _writeIndex)) {
         yield __privateGet(this, _emitter).emit("wait");
         yield Promise.race([
           __privateGet(this, _emitter).once("enqueue"),
@@ -185,8 +200,8 @@ var Source = class {
         ]);
         yield __privateGet(this, _emitter).emit("read");
       }
-      const read = Math.min(dst.length, __privateGet(this, _buffer).length - __privateGet(this, _readIndex));
-      dst.set(__privateGet(this, _buffer).slice(__privateGet(this, _readIndex), __privateGet(this, _readIndex) + read));
+      const read = Math.min(dst.length, __privateGet(this, _writeIndex) - __privateGet(this, _readIndex));
+      dst.set(__privateGet(this, _buffer).subarray(__privateGet(this, _readIndex), __privateGet(this, _readIndex) + read));
       __privateSet(this, _readIndex, __privateGet(this, _readIndex) + read);
       return read;
     });
@@ -224,19 +239,26 @@ var Source = class {
 _emitter = new WeakMap();
 _buffer = new WeakMap();
 _readIndex = new WeakMap();
+_writeIndex = new WeakMap();
 _closed = new WeakMap();
 _sampleRate = new WeakMap();
 
 // src/tts/utils.ts
 var import_base64_js = __toESM(require("base64-js"), 1);
 function base64ToArray(b64) {
-  return filterSentinel(b64).reduce((acc, b) => {
-    const floats = new Float32Array(import_base64_js.default.toByteArray(b).buffer);
-    const newAcc = new Float32Array(acc.length + floats.length);
-    newAcc.set(acc, 0);
-    newAcc.set(floats, acc.length);
-    return newAcc;
-  }, new Float32Array(0));
+  const byteArrays = filterSentinel(b64).map((b) => import_base64_js.default.toByteArray(b));
+  const totalLength = byteArrays.reduce(
+    (acc, arr) => acc + arr.length / Float32Array.BYTES_PER_ELEMENT,
+    0
+  );
+  const result = new Float32Array(totalLength);
+  let offset = 0;
+  for (const arr of byteArrays) {
+    const floats = new Float32Array(arr.buffer);
+    result.set(floats, offset);
+    offset += floats.length;
+  }
+  return result;
 }
 function createMessageHandlerForContextId(contextId, handler) {
   return (event) => {
@@ -386,7 +408,11 @@ var WebSocket = class extends Client {
    * @throws {Error} If the WebSocket fails to connect.
    */
   connect() {
-    const url = constructApiUrl(this.baseUrl, "/tts/websocket", "ws");
+    const url = constructApiUrl(
+      this.baseUrl,
+      `/tts/websocket?cartesia_version=${CARTESIA_VERSION}`,
+      "ws"
+    );
     url.searchParams.set("api_key", this.apiKey);
     const emitter = new import_emittery2.default();
     this.socket = new import_partysocket.WebSocket(url.toString());

package/dist/tts/websocket.js

@@ -1,10 +1,10 @@
 import {
   WebSocket
-} from "../chunk-NDNN326Q.js";
-import "../chunk-JOHSCOLW.js";
-import "../chunk-BCQ63627.js";
-import "../chunk-3GBZUGUD.js";
-import "../chunk-LYPTISWL.js";
+} from "../chunk-3F5E46FT.js";
+import "../chunk-PQ6CIPFW.js";
+import "../chunk-JGP5BIUV.js";
+import "../chunk-XHTDPLFR.js";
+import "../chunk-RO7TY474.js";
 import "../chunk-WIFMLPT5.js";
 export {
   WebSocket as default

package/dist/voices/index.cjs

@@ -75,7 +75,8 @@ module.exports = __toCommonJS(voices_exports);
 var import_cross_fetch = __toESM(require("cross-fetch"), 1);
 
 // src/lib/constants.ts
-var BASE_URL = "https://api.cartesia.ai/v0";
+var BASE_URL = "https://api.cartesia.ai/";
+var CARTESIA_VERSION = "2024-06-10";
 var constructApiUrl = (baseUrl, path, protocol) => {
   const normalizedPath = path.startsWith("/") ? path : `/${path}`;
   if (!protocol) {
@@ -100,7 +101,8 @@ var Client = class {
     const url = constructApiUrl(this.baseUrl, path);
     return (0, import_cross_fetch.default)(url.toString(), __spreadProps(__spreadValues({}, options), {
       headers: __spreadValues({
-        "X-API-KEY": this.apiKey
+        "X-API-Key": this.apiKey,
+        "Cartesia-Version": CARTESIA_VERSION
       }, options.headers)
     }));
   }

package/dist/voices/index.js

@@ -1,8 +1,8 @@
 import {
   Voices
-} from "../chunk-WBK6LLXX.js";
-import "../chunk-BCQ63627.js";
-import "../chunk-3GBZUGUD.js";
+} from "../chunk-WVTITUXX.js";
+import "../chunk-JGP5BIUV.js";
+import "../chunk-XHTDPLFR.js";
 import "../chunk-WIFMLPT5.js";
 export {
   Voices as default

package/package.json

@@ -4,7 +4,7 @@
 		"name": "Cartesia",
 		"url": "https://cartesia.ai"
 	},
-	"version": "1.0.0-alpha.1",
+	"version": "1.0.0-alpha.2",
 	"description": "Client for the Cartesia API.",
 	"type": "module",
 	"module": "./dist/index.js",

package/src/lib/client.ts

@@ -1,6 +1,6 @@
 import fetch from "cross-fetch";
 import type { ClientOptions } from "../types";
-import { BASE_URL, constructApiUrl } from "./constants";
+import { BASE_URL, CARTESIA_VERSION, constructApiUrl } from "./constants";
 
 export class Client {
 	apiKey: string;
@@ -22,7 +22,8 @@ export class Client {
 		return fetch(url.toString(), {
 			...options,
 			headers: {
-				"X-API-KEY": this.apiKey,
+				"X-API-Key": this.apiKey,
+				"Cartesia-Version": CARTESIA_VERSION,
 				...options.headers,
 			},
 		});

package/src/lib/constants.ts

@@ -1,4 +1,5 @@
-export const BASE_URL = "https://api.cartesia.ai/v0";
+export const BASE_URL = "https://api.cartesia.ai/";
+export const CARTESIA_VERSION = "2024-06-10";
 
 export const constructApiUrl = (
 	baseUrl: string,

package/src/react/index.ts

@@ -1,3 +1,4 @@
+import type { UnsubscribeFunction } from "emittery";
 import { useCallback, useEffect, useMemo, useRef, useState } from "react";
 import { Cartesia } from "../lib";
 import Player from "../tts/player";
@@ -91,11 +92,12 @@ export function useTTS({
 			if (!websocketReturn.current) {
 				return;
 			}
-			websocketReturn.current.on("message", (message) => {
+			const unsubscribe = websocketReturn.current.on("message", (message) => {
 				setMessages((messages) => [...messages, JSON.parse(message)]);
 			});
 			await websocketReturn.current.source.once("close");
 			setBufferStatus("buffered");
+			unsubscribe();
 		},
 		[websocket],
 	);
@@ -121,13 +123,21 @@ export function useTTS({
 				if (!connection) {
 					return;
 				}
+				const unsubscribes = <UnsubscribeFunction[]>[];
+				// The await ensures that the connection is open, so we already know that we are connected.
 				setIsConnected(true);
-				connection.on("open", () => {
-					setIsConnected(true);
-				});
-				const unsubscribe = connection.on("close", () => {
-					setIsConnected(false);
-				});
+				// If the WebSocket is the kind that automatically reconnects, we need an additional
+				// listener for the open event to update the connection status.
+				unsubscribes.push(
+					connection.on("open", () => {
+						setIsConnected(true);
+					}),
+				);
+				unsubscribes.push(
+					connection.on("close", () => {
+						setIsConnected(false);
+					}),
+				);
 				const intervalId = setInterval(() => {
 					if (baseUrl) {
 						pingServer(new URL(baseUrl).origin).then((ping) => {
@@ -144,7 +154,9 @@ export function useTTS({
 					}
 				}, PING_INTERVAL);
 				return () => {
-					unsubscribe();
+					for (const unsubscribe of unsubscribes) {
+						unsubscribe();
+					}
 					clearInterval(intervalId);
 					websocket?.disconnect();
 				};

package/src/tts/player.ts

@@ -54,7 +54,8 @@ export default class Player {
 			// If we've reached the end of the source, then read < buffer.length.
 			// In that case, we don't want to play the entire buffer, as that
 			// will cause repeated audio.
-			const playableAudio = buffer.slice(0, read);
+			// So we set the buffer to the correct length.
+			const playableAudio = buffer.subarray(0, read);
 			plays.push(this.#playBuffer(playableAudio, source.sampleRate));
 
 			if (read < buffer.length) {

package/src/tts/source.ts

@@ -3,8 +3,9 @@ import type { SourceEventData } from "../types";
 
 export default class Source {
 	#emitter = new Emittery<SourceEventData>();
-	#buffer = new Float32Array();
+	#buffer: Float32Array;
 	#readIndex = 0;
+	#writeIndex = 0;
 	#closed = false;
 	#sampleRate: number;
 
@@ -21,6 +22,7 @@ export default class Source {
 	 */
 	constructor({ sampleRate }: { sampleRate: number }) {
 		this.#sampleRate = sampleRate;
+		this.#buffer = new Float32Array(1024); // Initial size, can be adjusted
 	}
 
 	get sampleRate() {
@@ -33,8 +35,23 @@ export default class Source {
 	 * @param src The audio to append.
 	 */
 	async enqueue(src: Float32Array) {
+		const requiredCapacity = this.#writeIndex + src.length;
+
+		// Resize buffer if necessary
+		if (requiredCapacity > this.#buffer.length) {
+			let newCapacity = this.#buffer.length;
+			while (newCapacity < requiredCapacity) {
+				newCapacity *= 2; // Double the buffer size
+			}
+
+			const newBuffer = new Float32Array(newCapacity);
+			newBuffer.set(this.#buffer);
+			this.#buffer = newBuffer;
+		}
+
 		// Append the audio to the buffer.
-		this.#buffer = new Float32Array([...this.#buffer, ...src]);
+		this.#buffer.set(src, this.#writeIndex);
+		this.#writeIndex += src.length;
 		await this.#emitter.emit("enqueue");
 	}
 
@@ -49,7 +66,7 @@ export default class Source {
 		// Read the buffer into the provided buffer.
 		const targetReadIndex = this.#readIndex + dst.length;
 
-		while (!this.#closed && targetReadIndex > this.#buffer.length) {
+		while (!this.#closed && targetReadIndex > this.#writeIndex) {
 			// Wait for more audio to be enqueued.
 			await this.#emitter.emit("wait");
 			await Promise.race([
@@ -59,8 +76,8 @@ export default class Source {
 			await this.#emitter.emit("read");
 		}
 
-		const read = Math.min(dst.length, this.#buffer.length - this.#readIndex);
-		dst.set(this.#buffer.slice(this.#readIndex, this.#readIndex + read));
+		const read = Math.min(dst.length, this.#writeIndex - this.#readIndex);
+		dst.set(this.#buffer.subarray(this.#readIndex, this.#readIndex + read));
 		this.#readIndex += read;
 		return read;
 	}

package/src/tts/utils.ts

@@ -10,13 +10,21 @@ import type { Chunk, EmitteryCallbacks, Sentinel } from "../types";
  * @returns The audio buffer(s) as a Float32Array.
  */
 export function base64ToArray(b64: Chunk[]): Float32Array {
-	return filterSentinel(b64).reduce((acc, b) => {
-		const floats = new Float32Array(base64.toByteArray(b).buffer);
-		const newAcc = new Float32Array(acc.length + floats.length);
-		newAcc.set(acc, 0);
-		newAcc.set(floats, acc.length);
-		return newAcc;
-	}, new Float32Array(0));
+	const byteArrays = filterSentinel(b64).map((b) => base64.toByteArray(b));
+	const totalLength = byteArrays.reduce(
+		(acc, arr) => acc + arr.length / Float32Array.BYTES_PER_ELEMENT,
+		0,
+	);
+	const result = new Float32Array(totalLength);
+
+	let offset = 0;
+	for (const arr of byteArrays) {
+		const floats = new Float32Array(arr.buffer);
+		result.set(floats, offset);
+		offset += floats.length;
+	}
+
+	return result;
 }
 
 /**

package/src/tts/websocket.ts

@@ -2,7 +2,7 @@ import Emittery from "emittery";
 import { humanId } from "human-id";
 import { WebSocket as PartySocketWebSocket } from "partysocket";
 import { Client } from "../lib/client";
-import { constructApiUrl } from "../lib/constants";
+import { CARTESIA_VERSION, constructApiUrl } from "../lib/constants";
 import type {
 	ConnectionEventData,
 	EmitteryCallbacks,
@@ -150,7 +150,11 @@ export default class WebSocket extends Client {
 	 * @throws {Error} If the WebSocket fails to connect.
 	 */
 	connect() {
-		const url = constructApiUrl(this.baseUrl, "/tts/websocket", "ws");
+		const url = constructApiUrl(
+			this.baseUrl,
+			`/tts/websocket?cartesia_version=${CARTESIA_VERSION}`,
+			"ws",
+		);
 		url.searchParams.set("api_key", this.apiKey);
 		const emitter = new Emittery<ConnectionEventData>();
 		this.socket = new PartySocketWebSocket(url.toString());