@decartai/sdk 0.0.32 → 0.0.34

package/dist/process/types.d.ts

@@ -49,6 +49,27 @@ interface VideoModelInputs {
    */
   data?: FileInput;
 }
+/**
+ * Model-specific input documentation for lucy-pro-v2v.
+ */
+interface VideoEditInputs {
+  /**
+   * Text description to use for the video editing.
+   *
+   * See our [Prompt Engineering](https://docs.platform.decart.ai/models/video/video-generation#prompt-engineering) guide for how to write prompt for Decart video models effectively.
+   */
+  prompt: string;
+  /**
+   * Video file to process.
+   * Can be a File, Blob, ReadableStream, URL, or string URL.
+   */
+  data: FileInput;
+  /**
+   * Optional reference image to guide what to add to the video.
+   * Can be a File, Blob, ReadableStream, URL, or string URL.
+   */
+  reference_image?: FileInput;
+}
 /**
  * Model-specific input documentation for lucy-restyle-v2v.
  * Allows either prompt or reference_image (mutually exclusive).
@@ -85,7 +106,7 @@ interface PromptInput {
  * This allows different models to have field-specific documentation while maintaining type safety.
  * Specific models are checked first, then falls back to category-based selection.
  */
-type ModelSpecificInputs<T extends ModelDefinition> = T["name"] extends "lucy-pro-i2i" ? ImageEditingInputs : T["name"] extends "lucy-restyle-v2v" ? VideoRestyleInputs : T["name"] extends ImageModels ? ImageGenerationInputs : T["name"] extends VideoModels ? VideoModelInputs : PromptInput;
+type ModelSpecificInputs<T extends ModelDefinition> = T["name"] extends "lucy-pro-i2i" ? ImageEditingInputs : T["name"] extends "lucy-restyle-v2v" ? VideoRestyleInputs : T["name"] extends "lucy-pro-v2v" ? VideoEditInputs : T["name"] extends ImageModels ? ImageGenerationInputs : T["name"] extends VideoModels ? VideoModelInputs : PromptInput;
 interface ProcessInputs {
   /**
    * Random seed for reproducible results.

package/dist/realtime/client.js

@@ -56,6 +56,10 @@ const createRealTimeClient = (opts) => {
 			else imageBlob = avatar.avatarImage;
 			avatarImageBase64 = await blobToBase64(imageBlob);
 		}
+		const initialPrompt = isAvatarLive && options.initialState?.prompt ? {
+			text: options.initialState.prompt.text,
+			enhance: options.initialState.prompt.enhance
+		} : void 0;
 		const webrtcManager = new WebRTCManager({
 			webrtcUrl: `${`${baseUrl}${options.model.urlPath}`}?api_key=${apiKey}&model=${options.model.name}`,
 			apiKey,
@@ -75,15 +79,14 @@ const createRealTimeClient = (opts) => {
 			vp8MinBitrate: 300,
 			vp8StartBitrate: 600,
 			isAvatarLive,
-			avatarImageBase64
+			avatarImageBase64,
+			initialPrompt
 		});
 		await webrtcManager.connect(inputStream);
 		const methods = realtimeMethods(webrtcManager);
-		if (options.initialState) {
-			if (options.initialState.prompt) {
-				const { text, enhance } = options.initialState.prompt;
-				methods.setPrompt(text, { enhance });
-			}
+		if (!isAvatarLive && options.initialState?.prompt) {
+			const { text, enhance } = options.initialState.prompt;
+			methods.setPrompt(text, { enhance });
 		}
 		const client = {
 			setPrompt: methods.setPrompt,

package/dist/realtime/webrtc-connection.js

@@ -40,6 +40,7 @@ var WebRTCConnection = class {
 			this.ws.onclose = () => this.setState("disconnected");
 		});
 		if (this.callbacks.avatarImageBase64) await this.sendAvatarImage(this.callbacks.avatarImageBase64);
+		if (this.callbacks.initialPrompt) await this.sendInitialPrompt(this.callbacks.initialPrompt);
 		await this.setupNewPeerConnection();
 		return new Promise((resolve, reject) => {
 			this.connectionReject = reject;
@@ -71,6 +72,10 @@ var WebRTCConnection = class {
 				this.websocketMessagesEmitter.emit("setImageAck", msg);
 				return;
 			}
+			if (msg.type === "prompt_ack") {
+				this.websocketMessagesEmitter.emit("promptAck", msg);
+				return;
+			}
 			if (!this.pc) return;
 			switch (msg.type) {
 				case "ready": {
@@ -112,9 +117,6 @@ var WebRTCConnection = class {
 					if (turnConfig) await this.setupNewPeerConnection(turnConfig);
 					break;
 				}
-				case "prompt_ack":
-					this.websocketMessagesEmitter.emit("promptAck", msg);
-					break;
 			}
 		} catch (error) {
 			console.error("[WebRTC] Error:", error);
@@ -150,6 +152,31 @@ var WebRTCConnection = class {
 			});
 		});
 	}
+	/**
+	* Send the initial prompt to the server before WebRTC handshake.
+	*/
+	async sendInitialPrompt(prompt) {
+		return new Promise((resolve, reject) => {
+			const timeoutId = setTimeout(() => {
+				this.websocketMessagesEmitter.off("promptAck", listener);
+				reject(/* @__PURE__ */ new Error("Prompt send timed out"));
+			}, AVATAR_SETUP_TIMEOUT_MS);
+			const listener = (msg) => {
+				if (msg.prompt === prompt.text) {
+					clearTimeout(timeoutId);
+					this.websocketMessagesEmitter.off("promptAck", listener);
+					if (msg.success) resolve();
+					else reject(new Error(msg.error ?? "Failed to send prompt"));
+				}
+			};
+			this.websocketMessagesEmitter.on("promptAck", listener);
+			this.send({
+				type: "prompt",
+				prompt: prompt.text,
+				enhance_prompt: prompt.enhance ?? true
+			});
+		});
+	}
 	setState(state) {
 		if (this.state !== state) {
 			this.state = state;

package/dist/realtime/webrtc-manager.js

@@ -23,7 +23,8 @@ var WebRTCManager = class {
 			vp8MinBitrate: config.vp8MinBitrate,
 			vp8StartBitrate: config.vp8StartBitrate,
 			isAvatarLive: config.isAvatarLive,
-			avatarImageBase64: config.avatarImageBase64
+			avatarImageBase64: config.avatarImageBase64,
+			initialPrompt: config.initialPrompt
 		});
 	}
 	async connect(localStream) {

package/dist/shared/model.d.ts

@@ -49,6 +49,7 @@ declare const modelInputSchemas: {
   readonly "lucy-pro-v2v": z.ZodObject<{
     prompt: z.ZodString;
     data: z.ZodUnion<readonly [z.ZodCustom<File, File>, z.ZodCustom<Blob, Blob>, z.ZodCustom<ReadableStream<unknown>, ReadableStream<unknown>>, z.ZodCustom<URL, URL>, z.ZodURL]>;
+    reference_image: z.ZodOptional<z.ZodUnion<readonly [z.ZodCustom<File, File>, z.ZodCustom<Blob, Blob>, z.ZodCustom<ReadableStream<unknown>, ReadableStream<unknown>>, z.ZodCustom<URL, URL>, z.ZodURL]>>;
     seed: z.ZodOptional<z.ZodNumber>;
     resolution: z.ZodDefault<z.ZodOptional<z.ZodLiteral<"720p">>>;
     enhance_prompt: z.ZodOptional<z.ZodBoolean>;

package/dist/shared/model.js

@@ -87,6 +87,7 @@ const modelInputSchemas = {
 	"lucy-pro-v2v": z.object({
 		prompt: z.string().min(1).max(1e3).describe("The prompt to use for the generation"),
 		data: fileInputSchema.describe("The video data to use for generation (File, Blob, ReadableStream, URL, or string URL). Output video is limited to 5 seconds."),
+		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"),

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@decartai/sdk",
-  "version": "0.0.32",
+  "version": "0.0.34",
   "description": "Decart's JavaScript SDK",
   "type": "module",
   "license": "MIT",