@decartai/sdk 0.1.2 → 0.1.4

package/dist/realtime/media-channel.js

@@ -1,7 +1,7 @@
 import { createConsoleLogger } from "../utils/logger.js";
 import { REALTIME_CONFIG } from "./config-realtime.js";
 import mitt from "mitt";
-import { Room, RoomEvent, Track, TrackEvent } from "livekit-client";
+import { Room, RoomEvent, Track } from "livekit-client";
 //#region src/realtime/media-channel.ts
 function getDefaultVideoPublishOptions(videoCodec) {
 	const resolvedCodec = videoCodec ?? REALTIME_CONFIG.livekit.defaultVideoCodec;
@@ -40,16 +40,13 @@ var MediaChannel = class {
 		room.on(RoomEvent.TrackSubscribed, (track, _pub, participant) => {
 			if (!participant.identity.startsWith(REALTIME_CONFIG.livekit.inferenceServerIdentityPrefix)) return;
 			if (track.kind !== Track.Kind.Video && track.kind !== Track.Kind.Audio) return;
-			track.attach();
 			const mediaStreamTrack = track.mediaStreamTrack;
 			if (mediaStreamTrack) {
-				this.remoteStream ??= new MediaStream();
-				if (!this.remoteStream.getTracks().includes(mediaStreamTrack)) this.remoteStream.addTrack(mediaStreamTrack);
+				const tracks = this.remoteStream?.getTracks() ?? [];
+				if (!tracks.includes(mediaStreamTrack)) tracks.push(mediaStreamTrack);
+				this.remoteStream = new MediaStream(tracks);
 				this.events.emit("remoteStream", this.remoteStream);
 			}
-			track.on(TrackEvent.VideoPlaybackStarted, () => {
-				this.events.emit("firstFrame");
-			});
 		});
 		room.on(RoomEvent.Disconnected, (reason) => {
 			this.logger.warn("livekit: room disconnected", { reason });

package/dist/realtime/signaling-channel.js

@@ -262,6 +262,9 @@ var SignalingChannel = class {
 					queueSize: msg.queue_size
 				});
 				break;
+			case "generation_started":
+				this.events.emit("generationStarted");
+				break;
 			case "generation_tick":
 				this.events.emit("generationTick", { seconds: msg.seconds });
 				break;

package/dist/realtime/stream-session.js

@@ -165,7 +165,11 @@ var StreamSession = class {
 			this.queue = qp;
 			this.events.emit("queuePosition", qp);
 		});
-		this.signaling.on("generationTick", (e) => this.events.emit("generationTick", e));
+		this.signaling.on("generationStarted", () => this.markGenerating());
+		this.signaling.on("generationTick", (e) => {
+			this.markGenerating();
+			this.events.emit("generationTick", e);
+		});
 		this.signaling.on("generationEnded", (e) => this.events.emit("generationEnded", e));
 		this.signaling.on("serverError", (err) => this.events.emit("error", err));
 		this.signaling.on("closed", (info) => this.handleConnectionLoss({
@@ -173,11 +177,11 @@ var StreamSession = class {
 			...info
 		}));
 	}
+	markGenerating() {
+		if (this.state === "connected") this.setState("generating");
+	}
 	wireMediaEvents() {
 		this.media.on("remoteStream", (stream) => this.events.emit("remoteStream", stream));
-		this.media.on("firstFrame", () => {
-			if (this.state === "connected") this.setState("generating");
-		});
 		this.media.on("disconnected", (info) => this.handleConnectionLoss({
 			source: "media",
 			reason: info.reason

package/dist/realtime/subscribe-client.js

@@ -81,11 +81,11 @@ const createRealTimeSubscribeClient = (opts) => {
 			activeRoom.on(RoomEvent.TrackSubscribed, (track, _pub, participant) => {
 				if (!participant.identity.startsWith(REALTIME_CONFIG.livekit.inferenceServerIdentityPrefix)) return;
 				if (track.kind !== Track.Kind.Video && track.kind !== Track.Kind.Audio) return;
-				track.attach();
 				const mediaStreamTrack = track.mediaStreamTrack;
 				if (!mediaStreamTrack) return;
-				remoteStream ??= new MediaStream();
-				if (!remoteStream.getTracks().includes(mediaStreamTrack)) remoteStream.addTrack(mediaStreamTrack);
+				const tracks = remoteStream?.getTracks() ?? [];
+				if (!tracks.includes(mediaStreamTrack)) tracks.push(mediaStreamTrack);
+				remoteStream = new MediaStream(tracks);
 				options.onRemoteStream(remoteStream);
 			});
 			activeRoom.on(RoomEvent.ConnectionStateChanged, (state) => {

package/dist/shared/model.js

@@ -145,7 +145,7 @@ const proResolutionSchema = () => {
 */
 const v2vResolutionSchema = z.literal("720p").optional().describe("The resolution to use for the generation").default("720p");
 const videoEditSchema = z.object({
-	prompt: z.string().min(1).max(1e3).describe("The prompt to use for the generation"),
+	prompt: z.string().min(1).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"),
@@ -161,7 +161,7 @@ const imageEditSchema = z.object({
 	enhance_prompt: z.boolean().optional().describe("Whether to enhance the prompt")
 });
 const restyleSchema = z.object({
-	prompt: z.string().min(1).max(1e3).optional().describe("Text prompt for the video editing"),
+	prompt: z.string().min(1).optional().describe("Text prompt for the video editing"),
 	reference_image: fileInputSchema.optional().describe("Reference image to transform into a prompt (File, Blob, ReadableStream, URL, or string URL)"),
 	data: fileInputSchema.describe("Video file to process (File, Blob, ReadableStream, URL, or string URL)"),
 	seed: z.number().optional().describe("Seed for the video generation"),
@@ -169,7 +169,7 @@ const restyleSchema = z.object({
 	enhance_prompt: z.boolean().optional().describe("Whether to enhance the prompt (only valid with text prompt, defaults to true on backend)")
 }).refine((data) => data.prompt !== void 0 !== (data.reference_image !== void 0), { message: "Must provide either 'prompt' or 'reference_image', but not both" }).refine((data) => !(data.reference_image !== void 0 && data.enhance_prompt !== void 0), { message: "'enhance_prompt' is only valid when using 'prompt', not 'reference_image'" });
 const videoEdit2Schema = z.object({
-	prompt: z.string().max(1e3).describe("Text prompt for the video editing. Send an empty string if you want no text prompt."),
+	prompt: z.string().describe("Text prompt for the video editing. Send an empty string if you want no text prompt."),
 	reference_image: fileInputSchema.optional().describe("Optional reference image to guide the edit (File, Blob, ReadableStream, URL, or string URL)"),
 	data: fileInputSchema.describe("Video file to process (File, Blob, ReadableStream, URL, or string URL)"),
 	seed: z.number().optional().describe("The seed to use for the generation"),

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@decartai/sdk",
-  "version": "0.1.2",
+  "version": "0.1.4",
   "description": "Decart's JavaScript SDK",
   "type": "module",
   "license": "MIT",