Initial create a imagen sdk.

common/api-review/generative-ai.api.md

@@ -4,6 +4,8 @@
 
 ```ts
 
+import * as jspb from 'google-protobuf';
+
 // @public
 export interface BaseParams {
     // (undocumented)
@@ -620,29 +622,31 @@ export enum HarmProbability {
     NEGLIGIBLE = "NEGLIGIBLE"
 }
 
-// @public
-export class ImageGenerationModel {
-    constructor(apiKey: string, modelParams: ModelParams, _requestOptions?: RequestOptions);
-    // (undocumented)
-    apiKey: string;
-    generateImages(request: ImageGenerationRequest | string, requestOptions?: SingleRequestOptions): Promise<ImageGenerationResponse>;
-    // (undocumented)
-    model: string;
-}
-
 // @public
 export interface ImageGenerationRequest {
+    // (undocumented)
     aspectRatio?: "1:1" | "9:16" | "16:9" | "4:3" | "3:4";
+    // (undocumented)
     compressionQuality?: number;
+    // (undocumented)
     guidanceScale?: number;
+    // (undocumented)
     height?: number;
+    // (undocumented)
     language?: string;
+    // (undocumented)
     negativePrompt?: string;
+    // (undocumented)
     numberOfImages?: number;
+    // (undocumented)
     outputMimeType?: "image/png" | "image/jpeg";
+    // (undocumented)
     personGeneration?: "dont_allow" | "allow_adult";
+    // (undocumented)
     prompt: string;
+    // (undocumented)
     safetyFilterLevel?: "block_low_and_above" | "block_medium_and_above" | "block_only_high";
+    // (undocumented)
     width?: number;
 }
 
@@ -713,6 +717,18 @@ export type Part = TextPart | InlineDataPart | FunctionCallPart | FunctionRespon
 // @public
 export const POSSIBLE_ROLES: readonly ["user", "model", "function", "system"];
 
+// @public
+export interface PredictRequest {
+    instances?: any[];
+    model?: string;
+    parameters?: any;
+}
+
+// @public
+export interface PredictResponse {
+    predictions?: jspb.FieldValue[];
+}
+
 // @public
 export interface PromptFeedback {
     // (undocumented)

package.json

@@ -46,6 +46,7 @@
   "devDependencies": {
     "@changesets/cli": "^2.27.1",
     "@esm-bundle/chai": "^4.3.4-fix.0",
+    "@google-cloud/asset": "^5.7.0",
     "@microsoft/api-documenter": "^7.23.12",
     "@microsoft/api-extractor": "^7.38.2",
     "@rollup/plugin-json": "^6.0.1",
@@ -88,5 +89,9 @@
   "bugs": {
     "url": "https://github.com/google/generative-ai-js/issues"
   },
-  "homepage": "https://github.com/google/generative-ai-js#readme"
+  "homepage": "https://github.com/google/generative-ai-js#readme",
+  "dependencies": {
+    "@types/google-protobuf": "^3.15.12",
+    "google-protobuf": "^3.21.4"
+  }
 }

src/methods/generate-images.ts

@@ -1,27 +1,12 @@
-/**
- * @license
- * Copyright 2024 Google LLC
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License ats
- *
- *   http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
 import {
   ImageGenerationRequest,
   ImageGenerationResponse,
+  PredictResponse,
   SingleRequestOptions,
 } from "../../types";
 import { Task, makeModelRequest } from "../requests/request";
 import { convertFromImageGenerationRequest } from "../requests/request-helpers";
-import { ImageGenerationPredictResponse } from "../../types/predict";
+
 import { convertToImageGenerationResponse } from "../requests/response-helpers";
 
 export async function generateImages(
@@ -30,18 +15,14 @@ export async function generateImages(
   params: ImageGenerationRequest,
   requestOptions: SingleRequestOptions,
 ): Promise<ImageGenerationResponse> {
-  const predictRequest = convertFromImageGenerationRequest(model, params);
   const response = await makeModelRequest(
     model,
     Task.PREDICT,
     apiKey,
     /* stream */ false,
-    JSON.stringify(predictRequest),
+    JSON.stringify(convertFromImageGenerationRequest(model, params)),
     requestOptions,
   );
-  const responseJson: ImageGenerationPredictResponse = await response.json();
-  return convertToImageGenerationResponse(
-    predictRequest.parameters,
-    responseJson,
-  );
+  const responseJson: PredictResponse = await response.json();
+  return convertToImageGenerationResponse(responseJson);
 }

src/models/vision-model.ts

@@ -23,8 +23,6 @@ import {
   SingleRequestOptions,
 } from "../../types";
 import { generateImages } from "../methods/generate-images";
-import { formatGenerateImageInput } from "../requests/request-helpers";
-
 /**
  * Class for generative model APIs.
  * @public
@@ -45,19 +43,11 @@ export class ImageGenerationModel {
     }
   }
   /**
-   * Makes a single non-streaming call to the model
-   * and returns an object containing a single {@link ImageGenerationResponse}.
-   *
-   * Inside the response there will be generated pictures specified by
-   * numberOfImages in {@link ImageGenerationRequest}
-   *
-   * Fields set in the optional {@link SingleRequestOptions} parameter will
-   * take precedence over the {@link RequestOptions} values provided to
-   * {@link GoogleGenerativeAI.getImageGenerationModel }.
+   * Generates image based on the request.
    */
   async generateImages(
-    request: ImageGenerationRequest | string,
-    requestOptions: SingleRequestOptions = {},
+    request: ImageGenerationRequest,
+    requestOptions: SingleRequestOptions = {}
   ): Promise<ImageGenerationResponse> {
     const generativeModelRequestOptions: SingleRequestOptions = {
       ...this._requestOptions,
@@ -66,7 +56,7 @@ export class ImageGenerationModel {
     return generateImages(
       this.apiKey,
       this.model,
-      formatGenerateImageInput(request),
+      request,
       generativeModelRequestOptions,
     );
   }

src/requests/request-helpers.ts

@@ -21,12 +21,14 @@ import {
   EmbedContentRequest,
   GenerateContentRequest,
   ImageGenerationRequest,
+  ImageGenerationRequest,
   ModelParams,
   Part,
+  PredictRequest,
   _CountTokensRequestInternal,
   _GenerateContentRequestInternal,
 } from "../../types";
-import { PredictRequest } from "../../types/predict";
+import * as jspb from "google-protobuf";
 import {
   GoogleGenerativeAIError,
   GoogleGenerativeAIRequestInputError,
@@ -179,32 +181,28 @@ export function formatEmbedContentInput(
   return params;
 }
 
-export function formatGenerateImageInput(
-  params: ImageGenerationRequest | string,
-): ImageGenerationRequest {
-  let formattedRequest: ImageGenerationRequest;
-  if (typeof params === "string") {
-    formattedRequest = { prompt: params };
-  } else {
-    formattedRequest = params as ImageGenerationRequest;
-  }
-  return formattedRequest;
-}
-
 export function convertFromImageGenerationRequest(
-  model: string,
+  modelName: string,
   request: ImageGenerationRequest,
 ): PredictRequest {
   const instances = [{ prompt: request.prompt }];
-  const sampleImageSize = Math.max(request.width || 0, request.height || 0);
   const parameters = {
-    ...request,
-    sampleCount: request.numberOfImages || 1,
-    sampleImageSize: sampleImageSize === 0 ? undefined : sampleImageSize,
+    negativePrompt: request.negativePrompt,
+    numberOfImages: request.numberOfImages,
+    width: request.width,
+    height: request.height,
+    aspectRatio: request.aspectRatio,
+    guidanceScale: request.guidanceScale,
+    outputMimeType: request.outputMimeType,
+    compressionQuality: request.compressionQuality,
+    language: request.language,
+    safetyFilterLevel: request.safetyFilterLevel,
+    personGeneration: request.personGeneration,
   };
   return {
-    model,
-    instances,
-    parameters,
+    model: modelName,
+    instances: instances,
+    parameters: parameters,
   };
 }
+

src/requests/request.ts

@@ -39,7 +39,7 @@ export enum Task {
   COUNT_TOKENS = "countTokens",
   EMBED_CONTENT = "embedContent",
   BATCH_EMBED_CONTENTS = "batchEmbedContents",
-  PREDICT = "predict",
+  PREDICT = "predict"
 }
 
 export class RequestUrl {

src/requests/response-helpers.ts

@@ -22,6 +22,7 @@ import {
   GenerateContentCandidate,
   GenerateContentResponse,
   ImageGenerationResponse,
+  PredictResponse
 } from "../../types";
 import { GeneratedImage } from "../../types/generated-media";
 import {
@@ -214,15 +215,11 @@ export function formatBlockErrorMessage(
 }
 
 export function convertToImageGenerationResponse(
-  sharedParameters: PredictServiceValueType,
-  response: ImageGenerationPredictResponse,
+  response: PredictResponse
 ): ImageGenerationResponse {
-  const images: GeneratedImage[] = [];
-  for (const prediction of response!.predictions) {
-    images.push({
-      imageBytes: prediction.bytesBase64Encoded,
-      generationParameters: sharedParameters,
-    });
-  }
-  return { images };
+  console.log(response);
+  return {
+    images: []
+  };
 }
+

types/generated-media.ts

@@ -15,13 +15,29 @@
  * limitations under the License.
  */
 
-import { PredictServiceValueType } from "./predict";
-
+/**
+ * Generated media.
+ * @public
+ */
+export interface GeneratedMedia {
+  video?: GeneratedVideo;
+}
 
 /**
- * Generated images. It will be returned in response.
+ * Generated video.
  * @public
  */
+export interface GeneratedVideo {
+  /**
+   * Video bytes.
+   */
+  video?: string;
+  /**
+   * Specifies the dynamic retrieval configuration for the given source.
+   */
+  uri?: string;
+}
+
 export interface GeneratedImage {
   /**
    * Image bytes.
@@ -30,5 +46,5 @@ export interface GeneratedImage {
   /**
    * Generation parameters.
    */
-  generationParameters?: PredictServiceValueType; 
+  generationParameters?: { [key: string]: any }; 
 }