@decartai/sdk 0.0.46 → 0.0.48

package/dist/index.d.ts

@@ -7,7 +7,7 @@ import { DecartSDKError, ERROR_CODES } from "./utils/errors.js";
 import { ConnectionState } from "./realtime/types.js";
 import { SetInput } from "./realtime/methods.js";
 import { RealTimeSubscribeClient, SubscribeEvents, SubscribeOptions } from "./realtime/subscribe-client.js";
-import { AvatarOptions, Events, RealTimeClient, RealTimeClientConnectOptions, RealTimeClientInitialState } from "./realtime/client.js";
+import { Events, RealTimeClient, RealTimeClientConnectOptions, RealTimeClientInitialState } from "./realtime/client.js";
 import { ModelState } from "./shared/types.js";
 import { CreateTokenResponse, TokensClient } from "./tokens/client.js";
 
@@ -121,4 +121,4 @@ declare const createDecartClient: (options?: DecartClientOptions) => {
   tokens: TokensClient;
 };
 //#endregion
-export { type AvatarOptions, type ConnectionState, type CreateTokenResponse, DecartClientOptions, type DecartSDKError, ERROR_CODES, type FileInput, type ImageModelDefinition, type ImageModels, type JobStatus, type JobStatusResponse, type JobSubmitResponse, type Model, type ModelDefinition, type ModelState, type ProcessClient, type ProcessOptions, type QueueClient, type QueueJobResult, type QueueSubmitAndPollOptions, type QueueSubmitOptions, type ReactNativeFile, type RealTimeClient, type RealTimeClientConnectOptions, type RealTimeClientInitialState, type Events as RealTimeEvents, type RealTimeModels, type RealTimeSubscribeClient, type SetInput, type SubscribeEvents, type SubscribeOptions, type TokensClient, type VideoModelDefinition, type VideoModels, createDecartClient, isImageModel, isRealtimeModel, isVideoModel, models };
+export { type ConnectionState, type CreateTokenResponse, DecartClientOptions, type DecartSDKError, ERROR_CODES, type FileInput, type ImageModelDefinition, type ImageModels, type JobStatus, type JobStatusResponse, type JobSubmitResponse, type Model, type ModelDefinition, type ModelState, type ProcessClient, type ProcessOptions, type QueueClient, type QueueJobResult, type QueueSubmitAndPollOptions, type QueueSubmitOptions, type ReactNativeFile, type RealTimeClient, type RealTimeClientConnectOptions, type RealTimeClientInitialState, type Events as RealTimeEvents, type RealTimeModels, type RealTimeSubscribeClient, type SetInput, type SubscribeEvents, type SubscribeOptions, type TokensClient, type VideoModelDefinition, type VideoModels, createDecartClient, isImageModel, isRealtimeModel, isVideoModel, models };

package/dist/queue/polling.js

@@ -19,12 +19,17 @@ async function pollUntilComplete({ checkStatus, getContent, onStatusChange, sign
 		if (signal?.aborted) throw new Error("Polling aborted");
 		const status = await checkStatus();
 		if (onStatusChange) onStatusChange(status);
-		if (status.status === "completed") return {
-			status: "completed",
-			data: await getContent()
-		};
+		if (status.status === "completed") {
+			const data = await getContent();
+			return {
+				status: "completed",
+				job_id: status.job_id,
+				data
+			};
+		}
 		if (status.status === "failed") return {
 			status: "failed",
+			job_id: status.job_id,
 			error: "Job failed"
 		};
 		await sleep(POLLING_DEFAULTS.interval);

package/dist/queue/types.d.ts

@@ -26,9 +26,11 @@ type JobStatusResponse = {
  */
 type QueueJobResult = {
   status: "completed";
+  job_id: string;
   data: Blob;
 } | {
   status: "failed";
+  job_id: string;
   error: string;
 };
 /**

package/dist/realtime/client.d.ts

@@ -10,13 +10,10 @@ declare const realTimeClientInitialStateSchema: z.ZodObject<{
     text: z.ZodString;
     enhance: z.ZodDefault<z.ZodOptional<z.ZodBoolean>>;
   }, z.core.$strip>>;
+  image: z.ZodOptional<z.ZodUnion<readonly [z.ZodCustom<Blob, Blob>, z.ZodCustom<File, File>, z.ZodString]>>;
 }, z.core.$strip>;
 type OnRemoteStreamFn = (stream: MediaStream) => void;
 type RealTimeClientInitialState = z.infer<typeof realTimeClientInitialStateSchema>;
-declare const avatarOptionsSchema: z.ZodObject<{
-  avatarImage: z.ZodUnion<readonly [z.ZodCustom<Blob, Blob>, z.ZodCustom<File, File>, z.ZodString]>;
-}, z.core.$strip>;
-type AvatarOptions = z.infer<typeof avatarOptionsSchema>;
 declare const realTimeClientConnectOptionsSchema: z.ZodObject<{
   model: z.ZodObject<{
     name: z.ZodUnion<readonly [z.ZodUnion<readonly [z.ZodLiteral<"mirage">, z.ZodLiteral<"mirage_v2">, z.ZodLiteral<"lucy_v2v_720p_rt">, z.ZodLiteral<"lucy_2_rt">, z.ZodLiteral<"live_avatar">]>, z.ZodUnion<readonly [z.ZodLiteral<"lucy-dev-i2v">, z.ZodLiteral<"lucy-fast-v2v">, z.ZodLiteral<"lucy-pro-t2v">, z.ZodLiteral<"lucy-pro-i2v">, z.ZodLiteral<"lucy-pro-v2v">, z.ZodLiteral<"lucy-pro-flf2v">, z.ZodLiteral<"lucy-motion">, z.ZodLiteral<"lucy-restyle-v2v">]>, z.ZodUnion<readonly [z.ZodLiteral<"lucy-pro-t2i">, z.ZodLiteral<"lucy-pro-i2i">]>]>;
@@ -33,16 +30,17 @@ declare const realTimeClientConnectOptionsSchema: z.ZodObject<{
       text: z.ZodString;
       enhance: z.ZodDefault<z.ZodOptional<z.ZodBoolean>>;
     }, z.core.$strip>>;
+    image: z.ZodOptional<z.ZodUnion<readonly [z.ZodCustom<Blob, Blob>, z.ZodCustom<File, File>, z.ZodString]>>;
   }, z.core.$strip>>;
   customizeOffer: z.ZodOptional<z.ZodCustom<z.core.$InferInnerFunctionTypeAsync<z.core.$ZodFunctionArgs, z.core.$ZodFunctionOut>, z.core.$InferInnerFunctionTypeAsync<z.core.$ZodFunctionArgs, z.core.$ZodFunctionOut>>>;
-  avatar: z.ZodOptional<z.ZodObject<{
-    avatarImage: z.ZodUnion<readonly [z.ZodCustom<Blob, Blob>, z.ZodCustom<File, File>, z.ZodString]>;
-  }, z.core.$strip>>;
 }, z.core.$strip>;
 type RealTimeClientConnectOptions = z.infer<typeof realTimeClientConnectOptionsSchema>;
 type Events = {
   connectionChange: ConnectionState;
   error: DecartSDKError;
+  generationTick: {
+    seconds: number;
+  };
 };
 type RealTimeClient = {
   set: (input: SetInput) => Promise<void>;
@@ -66,4 +64,4 @@ type RealTimeClient = {
   playAudio?: (audio: Blob | File | ArrayBuffer) => Promise<void>;
 };
 //#endregion
-export { AvatarOptions, Events, RealTimeClient, RealTimeClientConnectOptions, RealTimeClientInitialState };
+export { Events, RealTimeClient, RealTimeClientConnectOptions, RealTimeClientInitialState };

package/dist/realtime/client.js

@@ -51,17 +51,11 @@ async function imageToBase64(image) {
 }
 const realTimeClientInitialStateSchema = modelStateSchema;
 const createAsyncFunctionSchema = (schema) => z.custom((fn) => schema.implementAsync(fn));
-const avatarOptionsSchema = z.object({ avatarImage: z.union([
-	z.instanceof(Blob),
-	z.instanceof(File),
-	z.string()
-]) });
 const realTimeClientConnectOptionsSchema = z.object({
 	model: modelDefinitionSchema,
 	onRemoteStream: z.custom((val) => typeof val === "function", { message: "onRemoteStream must be a function" }),
 	initialState: realTimeClientInitialStateSchema.optional(),
-	customizeOffer: createAsyncFunctionSchema(z.function()).optional(),
-	avatar: avatarOptionsSchema.optional()
+	customizeOffer: createAsyncFunctionSchema(z.function()).optional()
 });
 const createRealTimeClient = (opts) => {
 	const { baseUrl, apiKey, integration } = opts;
@@ -69,7 +63,7 @@ const createRealTimeClient = (opts) => {
 		const parsedOptions = realTimeClientConnectOptionsSchema.safeParse(options);
 		if (!parsedOptions.success) throw parsedOptions.error;
 		const isAvatarLive = options.model.name === "live_avatar";
-		const { onRemoteStream, initialState, avatar } = parsedOptions.data;
+		const { onRemoteStream, initialState } = parsedOptions.data;
 		let audioStreamManager;
 		let inputStream;
 		if (isAvatarLive && !stream) {
@@ -78,13 +72,8 @@ const createRealTimeClient = (opts) => {
 		} else inputStream = stream ?? new MediaStream();
 		let webrtcManager;
 		try {
-			let avatarImageBase64;
-			if (isAvatarLive && avatar?.avatarImage) if (typeof avatar.avatarImage === "string") {
-				const response = await fetch(avatar.avatarImage);
-				if (!response.ok) throw new Error(`Failed to fetch image: ${response.status} ${response.statusText}`);
-				avatarImageBase64 = await blobToBase64(await response.blob());
-			} else avatarImageBase64 = await blobToBase64(avatar.avatarImage);
-			const initialPrompt = isAvatarLive && initialState?.prompt ? {
+			const initialImage = initialState?.image ? await imageToBase64(initialState.image) : void 0;
+			const initialPrompt = initialState?.prompt ? {
 				text: initialState.prompt.text,
 				enhance: initialState.prompt.enhance
 			} : void 0;
@@ -104,8 +93,8 @@ const createRealTimeClient = (opts) => {
 				customizeOffer: options.customizeOffer,
 				vp8MinBitrate: 300,
 				vp8StartBitrate: 600,
-				isAvatarLive,
-				avatarImageBase64,
+				modelName: options.model.name,
+				initialImage,
 				initialPrompt
 			});
 			const manager = webrtcManager;
@@ -116,12 +105,12 @@ const createRealTimeClient = (opts) => {
 				sessionId = msg.session_id;
 			};
 			manager.getWebsocketMessageEmitter().on("sessionId", sessionIdListener);
+			const tickListener = (msg) => {
+				emitOrBuffer("generationTick", { seconds: msg.seconds });
+			};
+			manager.getWebsocketMessageEmitter().on("generationTick", tickListener);
 			await manager.connect(inputStream);
 			const methods = realtimeMethods(manager, imageToBase64);
-			if (!isAvatarLive && initialState?.prompt) {
-				const { text, enhance } = initialState.prompt;
-				await methods.setPrompt(text, { enhance });
-			}
 			const client = {
 				set: methods.set,
 				setPrompt: methods.setPrompt,

package/dist/realtime/webrtc-connection.js

@@ -53,8 +53,11 @@ var WebRTCConnection = class {
 					rejectConnect(/* @__PURE__ */ new Error("WebSocket closed"));
 				};
 			}), connectAbort]);
-			if (this.callbacks.avatarImageBase64) await Promise.race([this.sendAvatarImage(this.callbacks.avatarImageBase64), connectAbort]);
-			if (this.callbacks.initialPrompt) await Promise.race([this.sendInitialPrompt(this.callbacks.initialPrompt), connectAbort]);
+			if (this.callbacks.initialImage) await Promise.race([this.setImageBase64(this.callbacks.initialImage, {
+				prompt: this.callbacks.initialPrompt?.text,
+				enhance: this.callbacks.initialPrompt?.enhance
+			}), connectAbort]);
+			else if (this.callbacks.initialPrompt) await Promise.race([this.sendInitialPrompt(this.callbacks.initialPrompt), connectAbort]);
 			await this.setupNewPeerConnection();
 			await Promise.race([new Promise((resolve, reject) => {
 				const checkConnection = setInterval(() => {
@@ -98,6 +101,11 @@ var WebRTCConnection = class {
 				this.setState("generating");
 				return;
 			}
+			if (msg.type === "generation_tick") {
+				this.websocketMessagesEmitter.emit("generationTick", msg);
+				return;
+			}
+			if (msg.type === "generation_ended") return;
 			if (msg.type === "session_id") {
 				this.websocketMessagesEmitter.emit("sessionId", msg);
 				return;
@@ -158,15 +166,6 @@ var WebRTCConnection = class {
 		console.warn("[WebRTC] Message dropped: WebSocket is not open");
 		return false;
 	}
-	async sendAvatarImage(imageBase64) {
-		return this.setImageBase64(imageBase64);
-	}
-	/**
-	* Send an image to the server (e.g., as a reference for inference).
-	* Can be called after connection is established.
-	* Pass null to clear the reference image or use a placeholder.
-	* Optionally include a prompt to send with the image.
-	*/
 	async setImageBase64(imageBase64, options) {
 		return new Promise((resolve, reject) => {
 			const timeoutId = setTimeout(() => {
@@ -244,7 +243,7 @@ var WebRTCConnection = class {
 		this.pc = new RTCPeerConnection({ iceServers });
 		this.setState("connecting");
 		if (this.localStream) {
-			if (this.callbacks.isAvatarLive) this.pc.addTransceiver("video", { direction: "recvonly" });
+			if (this.callbacks.modelName === "live_avatar") this.pc.addTransceiver("video", { direction: "recvonly" });
 			this.localStream.getTracks().forEach((track) => {
 				if (this.pc && this.localStream) this.pc.addTrack(track, this.localStream);
 			});

package/dist/realtime/webrtc-manager.js

@@ -36,8 +36,8 @@ var WebRTCManager = class {
 			customizeOffer: config.customizeOffer,
 			vp8MinBitrate: config.vp8MinBitrate,
 			vp8StartBitrate: config.vp8StartBitrate,
-			isAvatarLive: config.isAvatarLive,
-			avatarImageBase64: config.avatarImageBase64,
+			modelName: config.modelName,
+			initialImage: config.initialImage,
 			initialPrompt: config.initialPrompt
 		});
 	}

package/dist/shared/types.d.ts

@@ -6,6 +6,7 @@ declare const modelStateSchema: z.ZodObject<{
     text: z.ZodString;
     enhance: z.ZodDefault<z.ZodOptional<z.ZodBoolean>>;
   }, z.core.$strip>>;
+  image: z.ZodOptional<z.ZodUnion<readonly [z.ZodCustom<Blob, Blob>, z.ZodCustom<File, File>, z.ZodString]>>;
 }, z.core.$strip>;
 type ModelState = z.infer<typeof modelStateSchema>;
 //#endregion

package/dist/shared/types.js

@@ -1,10 +1,17 @@
 import { z } from "zod";
 
 //#region src/shared/types.ts
-const modelStateSchema = z.object({ prompt: z.object({
-	text: z.string().min(1),
-	enhance: z.boolean().optional().default(true)
-}).optional() });
+const modelStateSchema = z.object({
+	prompt: z.object({
+		text: z.string().min(1),
+		enhance: z.boolean().optional().default(true)
+	}).optional(),
+	image: z.union([
+		z.instanceof(Blob),
+		z.instanceof(File),
+		z.string()
+	]).optional()
+});
 
 //#endregion
 export { modelStateSchema };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@decartai/sdk",
-  "version": "0.0.46",
+  "version": "0.0.48",
   "description": "Decart's JavaScript SDK",
   "type": "module",
   "license": "MIT",
@@ -32,13 +32,15 @@
     "@biomejs/biome": "2.3.8",
     "@types/bun": "^1.3.3",
     "@types/node": "^22.15.17",
+    "@vitest/browser-playwright": "^4",
     "bumpp": "^10.1.0",
     "msw": "^2.11.3",
     "pkg-pr-new": "^0.0.56",
+    "playwright": "^1.58.2",
     "tsdown": "^0.14.1",
     "typescript": "^5.8.3",
     "vite": "^7.1.2",
-    "vitest": "^3.1.3"
+    "vitest": "^4.0.18"
   },
   "dependencies": {
     "mitt": "^3.0.1",
@@ -51,6 +53,7 @@
     "dev:example": "vite dev --port 3000",
     "test": "vitest unit",
     "test:e2e": "vitest e2e",
+    "test:e2e:realtime": "vitest --config vitest.config.e2e-realtime.ts",
     "typecheck": "tsc --noEmit",
     "format": "biome format --write",
     "format:check": "biome check",