@decartai/sdk 0.0.41 → 0.0.43

package/dist/index.d.ts

@@ -4,6 +4,7 @@ import { ProcessClient } from "./process/client.js";
 import { JobStatus, JobStatusResponse, JobSubmitResponse, QueueJobResult, QueueSubmitAndPollOptions, QueueSubmitOptions } from "./queue/types.js";
 import { QueueClient } from "./queue/client.js";
 import { DecartSDKError, ERROR_CODES } from "./utils/errors.js";
+import { SetInput } from "./realtime/methods.js";
 import { AvatarOptions, RealTimeClient, RealTimeClientConnectOptions, RealTimeClientInitialState } from "./realtime/client.js";
 import { ModelState } from "./shared/types.js";
 import { CreateTokenResponse, TokensClient } from "./tokens/client.js";
@@ -117,4 +118,4 @@ declare const createDecartClient: (options?: DecartClientOptions) => {
   tokens: TokensClient;
 };
 //#endregion
-export { type AvatarOptions, 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 RealTimeModels, type TokensClient, type VideoModelDefinition, type VideoModels, createDecartClient, isImageModel, isRealtimeModel, isVideoModel, models };
+export { type AvatarOptions, 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 RealTimeModels, type SetInput, type TokensClient, type VideoModelDefinition, type VideoModels, createDecartClient, isImageModel, isRealtimeModel, isVideoModel, models };

package/dist/process/types.d.ts

@@ -158,12 +158,6 @@ interface ProcessInputs {
    * @default true
    */
   enhance_prompt?: boolean;
-  /**
-   * The number of inference steps.
-   *
-   * @default 50
-   */
-  num_inference_steps?: number;
 }
 /**
  * ProcessInputs combined with model-specific inputs.

package/dist/realtime/client.d.ts

@@ -1,4 +1,5 @@
 import { DecartSDKError } from "../utils/errors.js";
+import { SetInput } from "./methods.js";
 import { z } from "zod";
 
 //#region src/realtime/client.d.ts
@@ -43,6 +44,7 @@ type Events = {
   error: DecartSDKError;
 };
 type RealTimeClient = {
+  set: (input: SetInput) => Promise<void>;
   setPrompt: (prompt: string, {
     enhance
   }?: {

package/dist/realtime/client.js

@@ -20,6 +20,18 @@ async function blobToBase64(blob) {
 		reader.readAsDataURL(blob);
 	});
 }
+async function imageToBase64(image) {
+	if (typeof image === "string") {
+		let url = null;
+		try {
+			url = new URL(image);
+		} catch {}
+		if (url?.protocol === "data:") return image.split(",")[1];
+		if (url?.protocol === "http:" || url?.protocol === "https:") return blobToBase64(await (await fetch(image)).blob());
+		return image;
+	}
+	return blobToBase64(image);
+}
 const realTimeClientInitialStateSchema = modelStateSchema;
 const createAsyncFunctionSchema = (schema) => z.custom((fn) => schema.implementAsync(fn));
 const avatarOptionsSchema = z.object({ avatarImage: z.union([
@@ -83,12 +95,13 @@ const createRealTimeClient = (opts) => {
 			initialPrompt
 		});
 		await webrtcManager.connect(inputStream);
-		const methods = realtimeMethods(webrtcManager);
+		const methods = realtimeMethods(webrtcManager, imageToBase64);
 		if (!isAvatarLive && options.initialState?.prompt) {
 			const { text, enhance } = options.initialState.prompt;
 			methods.setPrompt(text, { enhance });
 		}
 		const client = {
+			set: methods.set,
 			setPrompt: methods.setPrompt,
 			isConnected: () => webrtcManager.isConnected(),
 			getConnectionState: () => webrtcManager.getConnectionState(),
@@ -101,17 +114,8 @@ const createRealTimeClient = (opts) => {
 			sessionId,
 			setImage: async (image, options$1) => {
 				if (image === null) return webrtcManager.setImage(null, options$1);
-				let imageBase64;
-				if (typeof image === "string") {
-					let url = null;
-					try {
-						url = new URL(image);
-					} catch {}
-					if (url?.protocol === "data:") imageBase64 = image.split(",")[1];
-					else if (url?.protocol === "http:" || url?.protocol === "https:") imageBase64 = await blobToBase64(await (await fetch(image)).blob());
-					else imageBase64 = image;
-				} else imageBase64 = await blobToBase64(image);
-				return webrtcManager.setImage(imageBase64, options$1);
+				const base64 = await imageToBase64(image);
+				return webrtcManager.setImage(base64, options$1);
 			}
 		};
 		if (isAvatarLive && audioStreamManager) {

package/dist/realtime/methods.d.ts

@@ -0,0 +1,11 @@
+import { z } from "zod";
+
+//#region src/realtime/methods.d.ts
+declare const setInputSchema: z.ZodObject<{
+  prompt: z.ZodOptional<z.ZodString>;
+  enhance: z.ZodDefault<z.ZodOptional<z.ZodBoolean>>;
+  image: z.ZodOptional<z.ZodUnion<readonly [z.ZodCustom<Blob, Blob>, z.ZodCustom<File, File>, z.ZodString, z.ZodNull]>>;
+}, z.core.$strip>;
+type SetInput = z.input<typeof setInputSchema>;
+//#endregion
+export { SetInput };

package/dist/realtime/methods.js

@@ -2,7 +2,30 @@ import { z } from "zod";
 
 //#region src/realtime/methods.ts
 const PROMPT_TIMEOUT_MS = 15 * 1e3;
-const realtimeMethods = (webrtcManager) => {
+const UPDATE_TIMEOUT_MS = 30 * 1e3;
+const setInputSchema = z.object({
+	prompt: z.string().min(1).optional(),
+	enhance: z.boolean().optional().default(true),
+	image: z.union([
+		z.instanceof(Blob),
+		z.instanceof(File),
+		z.string(),
+		z.null()
+	]).optional()
+}).refine((data) => data.prompt !== void 0 || data.image !== void 0, { message: "At least one of 'prompt' or 'image' must be provided" });
+const realtimeMethods = (webrtcManager, imageToBase64) => {
+	const set = async (input) => {
+		const parsed = setInputSchema.safeParse(input);
+		if (!parsed.success) throw parsed.error;
+		const { prompt, enhance, image } = parsed.data;
+		let imageBase64 = null;
+		if (image !== void 0 && image !== null) imageBase64 = await imageToBase64(image);
+		await webrtcManager.setImage(imageBase64, {
+			prompt,
+			enhance,
+			timeout: UPDATE_TIMEOUT_MS
+		});
+	};
 	const setPrompt = async (prompt, { enhance } = {}) => {
 		const parsedInput = z.object({
 			prompt: z.string().min(1),
@@ -37,7 +60,10 @@ const realtimeMethods = (webrtcManager) => {
 			if (timeoutId) clearTimeout(timeoutId);
 		}
 	};
-	return { setPrompt };
+	return {
+		set,
+		setPrompt
+	};
 };
 
 //#endregion

package/dist/realtime/webrtc-connection.js

@@ -3,7 +3,7 @@ import mitt from "mitt";
 
 //#region src/realtime/webrtc-connection.ts
 const ICE_SERVERS = [{ urls: "stun:stun.l.google.com:19302" }];
-const AVATAR_SETUP_TIMEOUT_MS = 15e3;
+const AVATAR_SETUP_TIMEOUT_MS = 3e4;
 var WebRTCConnection = class {
 	pc = null;
 	ws = null;
@@ -14,7 +14,7 @@ var WebRTCConnection = class {
 	constructor(callbacks = {}) {
 		this.callbacks = callbacks;
 	}
-	async connect(url, localStream, timeout = 6e4, integration) {
+	async connect(url, localStream, timeout, integration) {
 		const deadline = Date.now() + timeout;
 		this.localStream = localStream;
 		const userAgent = encodeURIComponent(buildUserAgent(integration));

package/dist/realtime/webrtc-manager.js

@@ -29,7 +29,7 @@ var WebRTCManager = class {
 	}
 	async connect(localStream) {
 		return pRetry(async () => {
-			await this.connection.connect(this.config.webrtcUrl, localStream, 6e4, this.config.integration);
+			await this.connection.connect(this.config.webrtcUrl, localStream, 6e4 * 5, this.config.integration);
 			return true;
 		}, {
 			retries: 5,

package/dist/shared/model.d.ts

@@ -69,7 +69,6 @@ declare const modelInputSchemas: {
     seed: z.ZodOptional<z.ZodNumber>;
     resolution: z.ZodDefault<z.ZodOptional<z.ZodLiteral<"720p">>>;
     enhance_prompt: z.ZodOptional<z.ZodBoolean>;
-    num_inference_steps: z.ZodOptional<z.ZodNumber>;
   }, z.core.$strip>;
   readonly "lucy-fast-v2v": z.ZodObject<{
     prompt: z.ZodString;

package/dist/shared/model.js

@@ -96,8 +96,7 @@ const modelInputSchemas = {
 		reference_image: fileInputSchema.optional().describe("Optional reference image to guide what to add to the video (File, Blob, ReadableStream, URL, or string URL)"),
 		seed: z.number().optional().describe("The seed to use for the generation"),
 		resolution: proV2vResolutionSchema,
-		enhance_prompt: z.boolean().optional().describe("Whether to enhance the prompt"),
-		num_inference_steps: z.number().optional().describe("The number of inference steps")
+		enhance_prompt: z.boolean().optional().describe("Whether to enhance the prompt")
 	}),
 	"lucy-fast-v2v": z.object({
 		prompt: z.string().min(1).max(1e3).describe("The prompt to use for the generation"),

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@decartai/sdk",
-  "version": "0.0.41",
+  "version": "0.0.43",
   "description": "Decart's JavaScript SDK",
   "type": "module",
   "license": "MIT",