Add 3d inference

.gitignore

@@ -2,4 +2,5 @@ node_modules/
 dist/
 
 *.env
-/tests/script.ts
+/tests/script.ts
+.npmrc

Runware/Runware-base.ts

@@ -45,6 +45,8 @@ import {
   IRequestAudio,
   IAudio,
   MediaUUID,
+  IRequestThreeD,
+  IThreeDImage,
 } from "./types";
 import {
   BASE_RUNWARE_URLS,
@@ -101,7 +103,7 @@ export class RunwareBase {
 
 
   private getUniqueUUID(item: MediaUUID): string | undefined {
-    return item.mediaUUID || item.audioUUID  || item.imageUUID  || item.videoUUID;
+    return item.mediaUUID || item.audioUUID  || item.imageUUID  || item.videoUUID || item.outputs?.files?.map((file) => file.uuid).join("-");
   }
 
   /**
@@ -1000,9 +1002,53 @@ export class RunwareBase {
     }
   };
 
+
+  threeDInference = async (
+    payload: IRequestThreeD
+  ): Promise<IThreeDImage[] | IThreeDImage> => {
+    const { skipResponse, deliveryMethod = "sync", ...rest } = payload;
+
+    try {
+      const requestMethod =
+        deliveryMethod === "sync"
+          ? this.baseSyncRequest
+          : this.baseSingleRequest;
+
+      const request = await requestMethod<IThreeDImage>({
+        payload: {
+          ...rest,
+          numberResults: rest.numberResults || 1,
+          taskType: ETaskType.THREE_D_INFERENCE,
+          deliveryMethod: deliveryMethod,
+        },
+        groupKey: LISTEN_TO_MEDIA_KEY.REQUEST_IMAGES,
+        debugKey: "three-d-inference",
+        skipResponse,
+      });
+
+
+      if (skipResponse) {
+        return request;
+      }
+
+
+      const taskUUID = request?.taskUUID;
+      if (deliveryMethod === "async") {
+        return this.pollForAsyncResults<IThreeDImage>({
+          taskUUID,
+          numberResults: payload?.numberResults,
+        });
+      }
+
+      // If not async, just return the initial result
+      return request;
+    } catch (e) {
+      throw e;
+    }
+  };
+
   getResponse = async <T>(payload: IAsyncResults): Promise<T[]> => {
     const taskUUID = payload.taskUUID;
-    // const mock = getRandomTaskResponses({ count: 2, taskUUID });
     return this.baseSingleRequest({
       payload: {
         ...payload,

Runware/Runware-server.ts

@@ -4,7 +4,6 @@ import WebSocket from "ws";
 
 import { RunwareBase } from "./Runware-base";
 import { ETaskType, RunwareBaseType, SdkType } from "./types";
-import { delay } from "./utils";
 
 // let allImages: IImage[] = [];
 
@@ -103,8 +102,6 @@ export class RunwareServer extends RunwareBase {
       if (!data) return;
       const m = JSON.parse(data);
 
-      // console.log("response", JSON.stringify(m, null, 4));
-
       this._listeners.forEach((lis) => {
         const result = lis.listener(m);
         if (result) {

Runware/types.ts

@@ -16,6 +16,7 @@ export enum ETaskType {
   VIDEO_INFERENCE = "videoInference",
   CAPTION = "caption",
   AUDIO_INFERENCE = "audioInference",
+  THREE_D_INFERENCE = "3dInference",
   GET_RESPONSE = "getResponse",
   PHOTO_MAKER = "photoMaker",
   IMAGE_CONTROL_NET_PRE_PROCESS = "imageControlNetPreProcess",
@@ -39,7 +40,8 @@ export type RunwareBaseType = {
 export type IOutputType = "base64Data" | "dataURI" | "URL";
 export type IOutputFormat = "JPG" | "PNG" | "WEBP";
 export type IVideoOutputFormat = "MP4" | "WEBM" | "MOV";
-export type IAudioOutputFormat = "MP3"
+export type IAudioOutputFormat = "MP3";
+export type IThreeDOutputFormat = "GLB" | "PLY";
 
 export interface IAdditionalResponsePayload {
   includePayload?: boolean;
@@ -77,6 +79,25 @@ export interface IVideoToImage {
   videoURL?: string;
 }
 
+interface TOutputFile {
+  uuid: string;
+  url: string;
+}
+
+interface TOutputFiles {
+  files: TOutputFile[];
+}
+
+export interface IThreeDImage {
+  taskType: ETaskType;
+  taskUUID: string;
+  status: string;
+  NSFWContent?: boolean;
+  cost?: number;
+  seed: number;
+  outputs?: TOutputFiles;
+}
+
 export interface IControlNetImage {
   taskUUID: string;
   inputImageUUID: string;
@@ -279,12 +300,12 @@ export interface IImageToText {
 
 export interface IRemoveImageBackground extends IRequestImageToText {
   outputType?: IOutputType;
-  outputFormat?: IOutputFormat| "MP4" | "WEBM" | "MOV";
+  outputFormat?: IOutputFormat | "MP4" | "WEBM" | "MOV";
   model: string;
   inputs?: {
     video?: InputsValue;
     image?: InputsValue;
-  }
+  };
   settings?: {
     rgba?: number[];
     postProcessMask?: boolean;
@@ -302,7 +323,6 @@ export interface IRemoveImageBackground extends IRequestImageToText {
   deliveryMethod?: string;
 }
 
-
 type InputsValue = string | Record<string, unknown>;
 
 export interface IRequestVideo extends IRequestImageToText {
@@ -379,7 +399,7 @@ export interface IRequestAudio {
     [key: string]: unknown;
   };
   deliveryMethod?: string;
-  
+
   taskUUID?: string;
   customTaskUUID?: string;
 
@@ -389,6 +409,31 @@ export interface IRequestAudio {
   [key: string]: unknown;
 }
 
+export interface IRequestThreeD {
+  model: string;
+  numberResults?: number;
+  outputType?: IOutputType;
+  outputFormat?: IThreeDOutputFormat;
+  uploadEndpoint?: string;
+  includeCost?: boolean;
+  positivePrompt?: string;
+  onPartialResponse?: (results: IThreeDImage[], error?: IError) => void;
+
+  inputs?: {
+    image?: InputsValue;
+    mask?: InputsValue;
+  } & {
+    [key: string]: unknown;
+  };
+  deliveryMethod?: string;
+  taskUUID?: string;
+  customTaskUUID?: string;
+
+  skipResponse?: boolean;
+  retry?: number;
+  [key: string]: unknown;
+}
+
 export interface IAsyncResults {
   taskUUID: string;
   onPartialImages?: (images: IImage[], error?: IError) => void;
@@ -434,7 +479,7 @@ export interface IUpscaleGan extends IAdditionalResponsePayload {
     image?: InputsValue;
   } & {
     [key: string]: unknown;
-  }
+  };
   model?: string;
 
   customTaskUUID?: string;
@@ -449,7 +494,7 @@ export type ReconnectingWebsocketProps = {
   addEventListener: (
     type: string,
     listener: EventListener,
-    options: any
+    options: any,
   ) => void;
   send: (data: any) => void;
 } & WebSocket;
@@ -600,7 +645,7 @@ export type TAddModelBaseType = {
   retry?: number;
   onUploadStream?: (
     response?: IAddModelResponse,
-    error?: IErrorResponse
+    error?: IErrorResponse,
   ) => void;
 };
 
@@ -855,7 +900,7 @@ export type TImageMaskingResponse = {
       y_min: number;
       x_max: number;
       y_max: number;
-    }
+    },
   ];
   maskImageURL: string;
   cost: number;
@@ -876,4 +921,5 @@ export type MediaUUID = {
   audioUUID?: string;
   imageUUID?: string;
   videoUUID?: string;
-}
+  outputs?: TOutputFiles;
+};

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@runware/sdk-js",
-  "version": "1.2.3",
+  "version": "1.2.4-beta.2",
   "description": "The SDK is used to run image inference with the Runware API, powered by the RunWare inference platform. It can be used to generate imaged with text-to-image and image-to-image. It also allows the use of an existing gallery of models or selecting any model or LoRA from the CivitAI gallery. The API also supports upscaling, background removal, inpainting and outpainting, and a series of other ControlNet models.",
   "main": "dist/index.js",
   "module": "dist/index.js",

readme.md

@@ -739,6 +739,10 @@ export type TImageMaskingResponse = {
 
 ## Changelog
 
+### - v1.2.4
+
+- Add 3d inference
+
 ### - v1.2.3
 
 - Handle async delivery method for image inference
@@ -774,7 +778,6 @@ Breaking change: `imageUUID` can now be undefined when using removeBackground/re
 **Added or Changed**
 
 - Added method aliases so task types match those of official API. It is recommended to use these new aliases going forward:
-
   - imageInference > requestImages
   - controlNetPreprocess > controlNetPreProcess
   - caption > requestImageToText