@cartesia/cartesia-js 3.0.0-b5 → 3.0.0-b6

package/src/backcompat/tts-wrapper.ts

@@ -0,0 +1,324 @@
+import WebSocket from "ws";
+import { Cartesia } from "../client";
+import { type RequestOptions as InternalRequestOptions } from "../internal/request-options";
+import { BackCompatRequestOptions } from "./types";
+
+// Define compatible interfaces to match the old SDK types for WebSocket
+export interface BackCompatWebSocketOptions {
+	container?: "raw" | "wav" | "mp3";
+	encoding?: "pcm_f32le" | "pcm_s16le" | "pcm_alaw" | "pcm_mulaw";
+	sampleRate: number;
+}
+
+export type BackCompatTtsRequestVoiceSpecifier =
+	| { mode: "id"; id: string }
+	| { mode: "embedding"; embedding: number[] };
+
+export interface BackCompatGenerationConfig {
+	volume?: number;
+	speed?: number;
+	emotion?: string[]; // Simplified from strict union for backcompat flexibility
+}
+
+export interface BackCompatWebSocketTtsRequest {
+	modelId: string;
+	transcript: string;
+	voice: BackCompatTtsRequestVoiceSpecifier;
+	generationConfig?: BackCompatGenerationConfig;
+	outputFormat?: {
+		container?: "raw" | "wav" | "mp3";
+		encoding?: "pcm_f32le" | "pcm_s16le" | "pcm_alaw" | "pcm_mulaw";
+		sampleRate?: number;
+		bitRate?: number;
+	};
+	contextId?: string; // Backcompat might pass this in request?
+	// Add other fields as needed
+	continue?: boolean;
+	duration?: number;
+	addTimestamps?: boolean;
+	addPhonemeTimestamps?: boolean;
+}
+
+// Helper for generating UUIDs. Not cryptographically secure.
+function uuidv4() {
+	return 'xxxxxxxx-xxxx-4xxx-yxxx-xxxxxxxxxxxx'.replace(/[xy]/g, function(c) {
+		var r = Math.random() * 16 | 0, v = c == 'x' ? r : (r & 0x3 | 0x8);
+		return v.toString(16);
+	});
+}
+
+class AudioSource {
+	private buffers: Buffer[] = [];
+	private waiter: ((val?: any) => void) | null = null;
+	public isDone = false;
+
+	push(data: Buffer) {
+		this.buffers.push(data);
+		if (this.waiter) {
+			this.waiter();
+			this.waiter = null;
+		}
+	}
+
+	markDone() {
+		this.isDone = true;
+		if (this.waiter) {
+			this.waiter();
+			this.waiter = null;
+		}
+	}
+
+	async read(outBuffer: Float32Array): Promise<number> {
+		if (this.buffers.length === 0 && !this.isDone) {
+			await new Promise<void>((resolve) => { this.waiter = resolve; });
+		}
+
+		if (this.buffers.length === 0 && this.isDone) {
+			return 0;
+		}
+
+		let totalFloatsRead = 0;
+		let outOffset = 0;
+		const maxFloats = outBuffer.length;
+
+		while (this.buffers.length > 0 && totalFloatsRead < maxFloats) {
+			const buf = this.buffers[0] as Buffer; // ts not smart enough to check loop condition
+			const floatsInBuf = buf.length / 4;
+			const floatsNeeded = maxFloats - totalFloatsRead;
+
+			const floatsToCopy = Math.min(floatsInBuf, floatsNeeded);
+			const bytesToCopy = floatsToCopy * 4;
+
+			// Copy to outBuffer.
+			// Create a view on the buffer to read floats.
+
+			// We need to ensure byteOffset is a multiple of 4.
+			// If not, we must copy the buffer to a new one.
+			let srcFloats: Float32Array;
+			if (buf.byteOffset % 4 === 0) {
+				srcFloats = new Float32Array(buf.buffer, buf.byteOffset, floatsInBuf);
+			} else {
+				const alignedBuf = new Uint8Array(buf);
+				srcFloats = new Float32Array(alignedBuf.buffer, alignedBuf.byteOffset, floatsInBuf);
+			}
+
+			outBuffer.set(srcFloats.subarray(0, floatsToCopy), outOffset);
+
+			totalFloatsRead += floatsToCopy;
+			outOffset += floatsToCopy;
+
+			if (floatsToCopy < floatsInBuf) {
+				// We didn't use the whole buffer. Update it.
+				this.buffers[0] = buf.subarray(bytesToCopy);
+			} else {
+				// We used the whole buffer. Remove it.
+				this.buffers.shift();
+			}
+		}
+
+		return totalFloatsRead;
+	}
+}
+
+export class WebSocketWrapper {
+	private client: Cartesia;
+	private config: BackCompatWebSocketOptions;
+	private socket: WebSocket | null = null;
+	private sources: Map<string, AudioSource> = new Map();
+	// Fallback source for messages without context_id or if we just want to capture everything (legacy behavior?)
+	// The original test didn't use context_id explicitly in send() but expected a response source.
+	// We'll map context_id to source.
+	private defaultSource: AudioSource | null = null;
+
+	constructor(client: Cartesia, config: BackCompatWebSocketOptions) {
+		this.client = client;
+		this.config = config;
+	}
+
+	async connect() {
+		const baseURL = this.client.baseURL;
+		// Construct WebSocket URL
+		// baseURL is like https://api.cartesia.ai
+		let urlStr = baseURL.replace(/^http/, "ws");
+		if (!urlStr.includes("/tts/websocket")) {
+			if (urlStr.endsWith("/")) {
+				urlStr += "tts/websocket";
+			} else {
+				urlStr += "/tts/websocket";
+			}
+		}
+
+		const url = new URL(urlStr);
+
+		const headers: any = {
+			"cartesia-version": "2025-04-16",
+		};
+		if (this.client.apiKey) {
+			headers["Authorization"] = `Bearer ${this.client.apiKey}`;
+		}
+
+		this.socket = new WebSocket(url.toString(), {
+			headers: headers,
+		});
+
+		return new Promise<void>((resolve, reject) => {
+			this.socket!.on("open", () => {
+				console.log("WebSocket connected.");
+				resolve();
+			});
+
+			this.socket!.on("error", (err) => {
+				console.error("WebSocket error:", err);
+				reject(err);
+			});
+
+			this.socket!.on("message", (data) => {
+				this.handleMessage(data);
+			});
+
+			this.socket!.on("close", () => {
+				console.log("WebSocket closed.");
+				this.sources.forEach((s) => { s.markDone(); });
+				if (this.defaultSource) this.defaultSource.markDone();
+			});
+		});
+	}
+
+	private handleMessage(data: WebSocket.Data) {
+		try {
+			const str = data.toString();
+			const msg = JSON.parse(str);
+
+			const contextId = msg.context_id;
+			let source = contextId ? this.sources.get(contextId) : this.defaultSource;
+
+			// If we received a message for a context we don't know about, and we have a default source, use it
+			if (!source && this.defaultSource) {
+				source = this.defaultSource;
+			}
+
+			if (msg.type === "chunk" && msg.data) {
+				const audioData = Buffer.from(msg.data, "base64");
+				if (source) source.push(audioData);
+			} else if (msg.type === "done") {
+				if (source) source.markDone();
+			} else if (msg.type === "error") {
+				console.error("Server error:", msg);
+				if (source) source.markDone(); // Fail the stream?
+			}
+		} catch (e) {
+			console.error("Error parsing message:", e);
+		}
+	}
+
+	async send(request: BackCompatWebSocketTtsRequest) {
+		if (!this.socket) {
+			throw new Error("WebSocket not connected");
+		}
+
+		// Ensure request has a context_id so we can route the response
+		const contextId = request.contextId || uuidv4();
+
+		const source = new AudioSource();
+		this.sources.set(contextId, source);
+		// Also set as default source if none exists, for compatibility with simple tests
+		if (!this.defaultSource) {
+			this.defaultSource = source;
+		}
+
+		// Construct payload
+		const payload: any = {
+			model_id: request.modelId,
+			transcript: request.transcript,
+			voice: request.voice,
+			context_id: contextId,
+		};
+
+		// Output Format
+		if (request.outputFormat) {
+			payload.output_format = {
+				container: request.outputFormat.container,
+				encoding: request.outputFormat.encoding,
+				sample_rate: request.outputFormat.sampleRate,
+				bit_rate: request.outputFormat.bitRate,
+			};
+		} else if (this.config) {
+			payload.output_format = {
+				container: this.config.container,
+				encoding: this.config.encoding,
+				sample_rate: this.config.sampleRate,
+			};
+		}
+
+		// Generation Config
+		if (request.generationConfig) {
+			payload.generation_config = request.generationConfig;
+		}
+
+		// Other fields
+		if (request.continue !== undefined) payload.continue = request.continue;
+		if (request.duration !== undefined) payload.duration = request.duration;
+		if (request.addTimestamps !== undefined) payload.add_timestamps = request.addTimestamps;
+		if (request.addPhonemeTimestamps !== undefined) payload.add_phoneme_timestamps = request.addPhonemeTimestamps;
+
+		this.socket.send(JSON.stringify(payload));
+
+		return {
+			source: source
+		};
+	}
+
+	disconnect() {
+		if (this.socket) {
+			this.socket.close();
+		}
+	}
+}
+
+export class TTSWrapper {
+	private client: Cartesia;
+
+	constructor(client: Cartesia) {
+		this.client = client;
+	}
+
+	websocket(config: BackCompatWebSocketOptions) {
+		return new WebSocketWrapper(this.client, config);
+	}
+
+	async bytes(request: any, requestOptions?: BackCompatRequestOptions) {
+		const params: any = {
+			model_id: request.modelId,
+			transcript: request.transcript,
+			voice: request.voice,
+			generation_config: request.generationConfig,
+			context_id: request.contextId,
+		};
+
+		if (request.outputFormat) {
+			params.output_format = {
+				...request.outputFormat,
+				sample_rate: request.outputFormat.sampleRate,
+				bit_rate: request.outputFormat.bitRate,
+			};
+			// Remove camelCase keys
+			delete params.output_format.sampleRate;
+			delete params.output_format.bitRate;
+		}
+
+		const options: any = {};
+		if (requestOptions) {
+			if (requestOptions.timeoutInSeconds) {
+				options.timeout = requestOptions.timeoutInSeconds * 1000;
+			}
+			if (requestOptions.maxRetries !== undefined) {
+				options.maxRetries = requestOptions.maxRetries;
+			}
+			options.headers = requestOptions.headers;
+			options.signal = requestOptions.abortSignal;
+		}
+
+		// @ts-ignore
+		return this.client.tts.synthesizeBytes(params, options);
+	}
+}

package/src/backcompat/types.ts

@@ -0,0 +1,19 @@
+export type Supplier<T> = T | Promise<T> | (() => T | Promise<T>);
+
+export interface CartesiaClientOptions {
+	environment?: Supplier<"https://api.cartesia.ai" | string>;
+	/** Specify a custom URL to connect the client to. */
+	baseUrl?: Supplier<string>;
+	apiKey?: Supplier<string | undefined>;
+	/** Override the Cartesia-Version header */
+	cartesiaVersion?: string;
+	fetcher?: any;
+}
+
+export interface BackCompatRequestOptions {
+	timeoutInSeconds?: number;
+	maxRetries?: number;
+	abortSignal?: AbortSignal;
+	cartesiaVersion?: string;
+	headers?: Record<string, string>;
+}

package/src/backcompat/voice-changer-wrapper.ts

@@ -0,0 +1,67 @@
+import * as fs from "fs";
+import { Cartesia } from "../client";
+import { type Uploadable } from "../core/uploads";
+import { type RequestOptions as InternalRequestOptions } from "../internal/request-options";
+import { Readable } from "stream";
+import { BackCompatRequestOptions } from "./types";
+
+export interface BackCompatVoiceChangerBytesRequest {
+	voiceId: string;
+	outputFormatContainer: "raw" | "wav" | "mp3";
+	outputFormatSampleRate: 8000 | 16000 | 22050 | 24000 | 44100 | 48000;
+	outputFormatEncoding?: "pcm_f32le" | "pcm_s16le" | "mulaw" | "alaw";
+	outputFormatBitRate?: 32000 | 64000 | 96000 | 128000 | 192000;
+}
+
+export class VoiceChangerWrapper {
+	private client: Cartesia;
+
+	constructor(client: Cartesia) {
+		this.client = client;
+	}
+
+	async bytes(
+		clip: File | fs.ReadStream | Blob,
+		request: BackCompatVoiceChangerBytesRequest,
+		requestOptions?: BackCompatRequestOptions
+	) {
+		const params: any = {
+			clip: clip as Uploadable,
+			"voice[id]": request.voiceId,
+			"output_format[container]": request.outputFormatContainer,
+			"output_format[sample_rate]": request.outputFormatSampleRate,
+		};
+
+		if (request.outputFormatEncoding) {
+			params["output_format[encoding]"] = request.outputFormatEncoding;
+		}
+		if (request.outputFormatBitRate) {
+			params["output_format[bit_rate]"] = request.outputFormatBitRate;
+		}
+
+		const options: InternalRequestOptions = {};
+		if (requestOptions) {
+			if (requestOptions.timeoutInSeconds) {
+				options.timeout = requestOptions.timeoutInSeconds * 1000;
+			}
+			if (requestOptions.maxRetries !== undefined) {
+				options.maxRetries = requestOptions.maxRetries;
+			}
+			options.headers = requestOptions.headers;
+			options.signal = requestOptions.abortSignal;
+		}
+
+		const response = await this.client.voiceChanger.changeVoiceBytes(params, {
+			...options,
+			__binaryResponse: true,
+		} as any);
+
+		// @ts-ignore
+		if (response.body) {
+			// @ts-ignore
+			return Readable.fromWeb(response.body);
+		}
+
+		return response;
+	}
+}

package/src/backcompat/voices-wrapper.ts

@@ -0,0 +1,158 @@
+import * as fs from "fs";
+import { Cartesia } from "../client";
+import { type Uploadable } from "../core/uploads";
+import {
+	type VoiceCloneParams,
+	type VoiceMetadata,
+	type SupportedLanguage,
+	type VoiceUpdateParams,
+	type Voice,
+	type VoiceLocalizeParams,
+} from "../resources/voices";
+import { type RequestOptions as InternalRequestOptions } from "../internal/request-options";
+import { BackCompatRequestOptions } from "./types";
+
+export interface BackCompatCloneVoiceRequest {
+	name: string;
+	description?: string;
+	language: string;
+	mode: "similarity" | "stability";
+	enhance?: boolean;
+	baseVoiceId?: string;
+}
+
+export interface BackCompatUpdateVoiceRequest {
+	name: string;
+	description: string;
+}
+
+export interface BackCompatLocalizeVoiceRequest {
+	voiceId: string;
+	name: string;
+	description: string;
+	language:
+		| "en"
+		| "de"
+		| "es"
+		| "fr"
+		| "ja"
+		| "pt"
+		| "zh"
+		| "hi"
+		| "it"
+		| "ko"
+		| "nl"
+		| "pl"
+		| "ru"
+		| "sv"
+		| "tr";
+	originalSpeakerGender: "male" | "female";
+	dialect?:
+		| "au"
+		| "in"
+		| "so"
+		| "uk"
+		| "us"
+		| "mx"
+		| "pe"
+		| "br"
+		| "eu"
+		| "ca";
+}
+
+export class VoicesWrapper {
+	private client: Cartesia;
+
+	constructor(client: Cartesia) {
+		this.client = client;
+	}
+
+	async clone(
+		clip: File | fs.ReadStream | Blob,
+		request: BackCompatCloneVoiceRequest,
+		requestOptions?: BackCompatRequestOptions
+	): Promise<VoiceMetadata> {
+		const params: VoiceCloneParams = {
+			clip: clip as Uploadable,
+			name: request.name,
+			language: request.language as SupportedLanguage,
+		};
+
+		if (request.description !== undefined) {
+			params.description = request.description;
+		}
+		if (request.baseVoiceId !== undefined) {
+			params.base_voice_id = request.baseVoiceId;
+		}
+
+		const options: InternalRequestOptions = {};
+		if (requestOptions) {
+			if (requestOptions.timeoutInSeconds) {
+				options.timeout = requestOptions.timeoutInSeconds * 1000;
+			}
+			if (requestOptions.maxRetries !== undefined) {
+				options.maxRetries = requestOptions.maxRetries;
+			}
+			options.headers = requestOptions.headers;
+			options.signal = requestOptions.abortSignal;
+		}
+
+		return this.client.voices.clone(params, options);
+	}
+
+	async update(
+		id: string,
+		request: BackCompatUpdateVoiceRequest,
+		requestOptions?: BackCompatRequestOptions
+	): Promise<Voice> {
+		const params: VoiceUpdateParams = {
+			name: request.name,
+			description: request.description,
+		};
+
+		const options: InternalRequestOptions = {};
+		if (requestOptions) {
+			if (requestOptions.timeoutInSeconds) {
+				options.timeout = requestOptions.timeoutInSeconds * 1000;
+			}
+			if (requestOptions.maxRetries !== undefined) {
+				options.maxRetries = requestOptions.maxRetries;
+			}
+			options.headers = requestOptions.headers;
+			options.signal = requestOptions.abortSignal;
+		}
+
+		return this.client.voices.update(id, params, options);
+	}
+
+	async localize(
+		request: BackCompatLocalizeVoiceRequest,
+		requestOptions?: BackCompatRequestOptions
+	): Promise<VoiceMetadata> {
+		const params: VoiceLocalizeParams = {
+			voice_id: request.voiceId,
+			name: request.name,
+			description: request.description,
+			language: request.language as any,
+			original_speaker_gender: request.originalSpeakerGender,
+		};
+
+		if (request.dialect) {
+			params.dialect = request.dialect as any; // Cast dialect as list might vary slightly or be strict
+		}
+
+		const options: InternalRequestOptions = {};
+		if (requestOptions) {
+			if (requestOptions.timeoutInSeconds) {
+				options.timeout = requestOptions.timeoutInSeconds * 1000;
+			}
+			if (requestOptions.maxRetries !== undefined) {
+				options.maxRetries = requestOptions.maxRetries;
+			}
+			options.headers = requestOptions.headers;
+			options.signal = requestOptions.abortSignal;
+		}
+
+		return this.client.voices.localize(params, options);
+	}
+}

package/src/version.ts

@@ -1 +1 @@
-export const VERSION = '3.0.0-b5'; // x-release-please-version
+export const VERSION = '3.0.0-b6'; // x-release-please-version

package/version.d.mts

@@ -1,2 +1,2 @@
-export declare const VERSION = "3.0.0-b5";
+export declare const VERSION = "3.0.0-b6";
 //# sourceMappingURL=version.d.mts.map

package/version.d.ts

@@ -1,2 +1,2 @@
-export declare const VERSION = "3.0.0-b5";
+export declare const VERSION = "3.0.0-b6";
 //# sourceMappingURL=version.d.ts.map

package/version.js

@@ -1,5 +1,5 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.VERSION = void 0;
-exports.VERSION = '3.0.0-b5'; // x-release-please-version
+exports.VERSION = '3.0.0-b6'; // x-release-please-version
 //# sourceMappingURL=version.js.map

package/version.mjs

@@ -1,2 +1,2 @@
-export const VERSION = '3.0.0-b5'; // x-release-please-version
+export const VERSION = '3.0.0-b6'; // x-release-please-version
 //# sourceMappingURL=version.mjs.map