@huggingface/inference 4.13.9 → 4.13.10

package/src/providers/nscale.ts

@@ -15,10 +15,10 @@
  * Thanks!
  */
 import type { TextToImageInput } from "@huggingface/tasks";
-import type { BodyParams } from "../types.js";
+import type { BodyParams, OutputType } from "../types.js";
 import { omit } from "../utils/omit.js";
 import { BaseConversationalTask, TaskProviderHelper, type TextToImageTaskHelper } from "./providerHelper.js";
-import { InferenceClientProviderOutputError } from "../errors.js";
+import { InferenceClientInputError, InferenceClientProviderOutputError } from "../errors.js";
 
 const NSCALE_API_BASE_URL = "https://inference.api.nscale.com";
 
@@ -40,6 +40,11 @@ export class NscaleTextToImageTask extends TaskProviderHelper implements TextToI
 	}
 
 	preparePayload(params: BodyParams<TextToImageInput>): Record<string, unknown> {
+		if (params.outputType === "url") {
+			throw new InferenceClientInputError(
+				"nscale provider does not support URL output. Use outputType 'blob', 'dataUrl' or 'json' instead."
+			);
+		}
 		return {
 			...omit(params.args, ["inputs", "parameters"]),
 			...params.args.parameters,
@@ -57,7 +62,7 @@ export class NscaleTextToImageTask extends TaskProviderHelper implements TextToI
 		response: NscaleCloudBase64ImageGeneration,
 		url?: string,
 		headers?: HeadersInit,
-		outputType?: "url" | "blob" | "json"
+		outputType?: OutputType
 	): Promise<string | Blob | Record<string, unknown>> {
 		if (
 			typeof response === "object" &&
@@ -71,7 +76,7 @@ export class NscaleTextToImageTask extends TaskProviderHelper implements TextToI
 				return { ...response };
 			}
 			const base64Data = response.data[0].b64_json;
-			if (outputType === "url") {
+			if (outputType === "dataUrl") {
 				return `data:image/jpeg;base64,${base64Data}`;
 			}
 			return fetch(`data:image/jpeg;base64,${base64Data}`).then((res) => res.blob());

package/src/providers/providerHelper.ts

@@ -51,7 +51,15 @@ import type {
 import { HF_ROUTER_URL } from "../config.js";
 import { InferenceClientProviderOutputError, InferenceClientRoutingError } from "../errors.js";
 import type { AudioToAudioOutput } from "../tasks/audio/audioToAudio.js";
-import type { BaseArgs, BodyParams, HeaderParams, InferenceProvider, RequestArgs, UrlParams } from "../types.js";
+import type {
+	BaseArgs,
+	BodyParams,
+	HeaderParams,
+	InferenceProvider,
+	OutputType,
+	RequestArgs,
+	UrlParams,
+} from "../types.js";
 import { toArray } from "../utils/toArray.js";
 import type { ImageToImageArgs } from "../tasks/cv/imageToImage.js";
 import type { AutomaticSpeechRecognitionArgs } from "../tasks/audio/automaticSpeechRecognition.js";
@@ -78,7 +86,7 @@ export abstract class TaskProviderHelper {
 		response: unknown,
 		url?: string,
 		headers?: HeadersInit,
-		outputType?: "url" | "blob"
+		outputType?: OutputType
 	): Promise<unknown>;
 
 	/**
@@ -115,6 +123,9 @@ export abstract class TaskProviderHelper {
 	makeUrl(params: UrlParams): string {
 		const baseUrl = this.makeBaseUrl(params);
 		const route = this.makeRoute(params).replace(/^\/+/, "");
+		if (params.urlTransform) {
+			return params.urlTransform(`${baseUrl}/${route}`);
+		}
 		return `${baseUrl}/${route}`;
 	}
 
@@ -141,7 +152,7 @@ export interface TextToImageTaskHelper {
 		response: unknown,
 		url?: string,
 		headers?: HeadersInit,
-		outputType?: "url" | "blob" | "json"
+		outputType?: OutputType
 	): Promise<string | Blob | Record<string, unknown>>;
 	preparePayload(params: BodyParams<TextToImageInput & BaseArgs>): Record<string, unknown>;
 }

package/src/providers/replicate.ts

@@ -16,7 +16,8 @@
  */
 import { InferenceClientProviderOutputError } from "../errors.js";
 import { isUrl } from "../lib/isUrl.js";
-import type { BodyParams, HeaderParams, RequestArgs, UrlParams } from "../types.js";
+import type { BodyParams, HeaderParams, OutputType, RequestArgs, UrlParams } from "../types.js";
+import { dataUrlFromBlob } from "../utils/dataUrlFromBlob.js";
 import { omit } from "../utils/omit.js";
 import {
 	TaskProviderHelper,
@@ -91,7 +92,7 @@ export class ReplicateTextToImageTask extends ReplicateTask implements TextToIma
 		res: ReplicateOutput | Blob,
 		url?: string,
 		headers?: Record<string, string>,
-		outputType?: "url" | "blob" | "json"
+		outputType?: OutputType
 	): Promise<string | Blob | Record<string, unknown>> {
 		void url;
 		void headers;
@@ -105,7 +106,8 @@ export class ReplicateTextToImageTask extends ReplicateTask implements TextToIma
 				return res.output;
 			}
 			const urlResponse = await fetch(res.output);
-			return await urlResponse.blob();
+			const blob = await urlResponse.blob();
+			return outputType === "dataUrl" ? dataUrlFromBlob(blob) : blob;
 		}
 
 		// Handle array output
@@ -123,7 +125,8 @@ export class ReplicateTextToImageTask extends ReplicateTask implements TextToIma
 				return res.output[0];
 			}
 			const urlResponse = await fetch(res.output[0]);
-			return await urlResponse.blob();
+			const blob = await urlResponse.blob();
+			return outputType === "dataUrl" ? dataUrlFromBlob(blob) : blob;
 		}
 
 		throw new InferenceClientProviderOutputError("Received malformed response from Replicate text-to-image API");

package/src/providers/together.ts

@@ -15,7 +15,7 @@
  * Thanks!
  */
 import type { ChatCompletionOutput, TextGenerationOutput, TextGenerationOutputFinishReason } from "@huggingface/tasks";
-import type { BodyParams } from "../types.js";
+import type { BodyParams, OutputType } from "../types.js";
 import { omit } from "../utils/omit.js";
 import {
 	BaseConversationalTask,
@@ -38,9 +38,10 @@ interface TogetherTextCompletionOutput extends Omit<ChatCompletionOutput, "choic
 	}>;
 }
 
-interface TogetherBase64ImageGeneration {
+interface TogetherImageGeneration {
 	data: Array<{
-		b64_json: string;
+		b64_json?: string;
+		url?: string;
 	}>;
 }
 
@@ -109,33 +110,38 @@ export class TogetherTextToImageTask extends TaskProviderHelper implements TextT
 			...omit(params.args, ["inputs", "parameters"]),
 			...(params.args.parameters as Record<string, unknown>),
 			prompt: params.args.inputs,
-			response_format: "base64",
+			response_format: params.outputType === "url" ? "url" : "base64",
 			model: params.model,
 		};
 	}
 
 	async getResponse(
-		response: TogetherBase64ImageGeneration,
+		response: TogetherImageGeneration,
 		url?: string,
 		headers?: HeadersInit,
-		outputType?: "url" | "blob" | "json"
+		outputType?: OutputType
 	): Promise<string | Blob | Record<string, unknown>> {
 		if (
 			typeof response === "object" &&
 			"data" in response &&
 			Array.isArray(response.data) &&
-			response.data.length > 0 &&
-			"b64_json" in response.data[0] &&
-			typeof response.data[0].b64_json === "string"
+			response.data.length > 0
 		) {
 			if (outputType === "json") {
 				return { ...response };
 			}
-			const base64Data = response.data[0].b64_json;
-			if (outputType === "url") {
-				return `data:image/jpeg;base64,${base64Data}`;
+
+			if ("url" in response.data[0] && typeof response.data[0].url === "string") {
+				return response.data[0].url;
+			}
+
+			if ("b64_json" in response.data[0] && typeof response.data[0].b64_json === "string") {
+				const base64Data = response.data[0].b64_json;
+				if (outputType === "dataUrl") {
+					return `data:image/jpeg;base64,${base64Data}`;
+				}
+				return fetch(`data:image/jpeg;base64,${base64Data}`).then((res) => res.blob());
 			}
-			return fetch(`data:image/jpeg;base64,${base64Data}`).then((res) => res.blob());
 		}
 
 		throw new InferenceClientProviderOutputError("Received malformed response from Together text-to-image API");

package/src/providers/wavespeed.ts

@@ -1,8 +1,11 @@
 import type { TextToImageArgs } from "../tasks/cv/textToImage.js";
 import type { ImageToImageArgs } from "../tasks/cv/imageToImage.js";
+import type { ImageTextToImageArgs } from "../tasks/cv/imageTextToImage.js";
 import type { TextToVideoArgs } from "../tasks/cv/textToVideo.js";
 import type { ImageToVideoArgs } from "../tasks/cv/imageToVideo.js";
-import type { BodyParams, RequestArgs, UrlParams } from "../types.js";
+import type { BodyParams, OutputType, RequestArgs, UrlParams } from "../types.js";
+import type { ImageTextToVideoArgs } from "../tasks/cv/imageTextToVideo.js";
+import { dataUrlFromBlob } from "../utils/dataUrlFromBlob.js";
 import { delay } from "../utils/delay.js";
 import { omit } from "../utils/omit.js";
 import { base64FromBytes } from "../utils/base64FromBytes.js";
@@ -11,6 +14,8 @@ import type {
 	TextToVideoTaskHelper,
 	ImageToImageTaskHelper,
 	ImageToVideoTaskHelper,
+	ImageTextToImageTaskHelper,
+	ImageTextToVideoTaskHelper,
 } from "./providerHelper.js";
 import { TaskProviderHelper } from "./providerHelper.js";
 import {
@@ -93,7 +98,14 @@ abstract class WavespeedAITask extends TaskProviderHelper {
 	}
 
 	preparePayload(
-		params: BodyParams<ImageToImageArgs | TextToImageArgs | TextToVideoArgs | ImageToVideoArgs>
+		params: BodyParams<
+			| ImageToImageArgs
+			| ImageTextToImageArgs
+			| ImageTextToVideoArgs
+			| TextToImageArgs
+			| TextToVideoArgs
+			| ImageToVideoArgs
+		>
 	): Record<string, unknown> {
 		const payload: Record<string, unknown> = {
 			...omit(params.args, ["inputs", "parameters"]),
@@ -116,7 +128,7 @@ abstract class WavespeedAITask extends TaskProviderHelper {
 		response: WaveSpeedAISubmitTaskResponse,
 		url?: string,
 		headers?: Record<string, string>,
-		outputType?: "url" | "blob" | "json"
+		outputType?: OutputType
 	): Promise<string | Blob | Record<string, unknown>> {
 		if (!url || !headers) {
 			throw new InferenceClientInputError("Headers are required for WaveSpeed AI API calls");
@@ -179,7 +191,8 @@ abstract class WavespeedAITask extends TaskProviderHelper {
 							}
 						);
 					}
-					return await mediaResponse.blob();
+					const blob = await mediaResponse.blob();
+					return outputType === "dataUrl" ? dataUrlFromBlob(blob) : blob;
 				}
 				case "failed": {
 					throw new InferenceClientProviderOutputError(taskResult.error || "Task failed");
@@ -256,3 +269,34 @@ export class WavespeedAIImageToVideoTask extends WavespeedAITask implements Imag
 		return super.getResponse(response, url, headers) as Promise<Blob>;
 	}
 }
+
+// 1x1 fully transparent PNG for use when no input image is provided
+const TRANSPARENT_1PX_PNG_BASE64 =
+	"iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mNkYAAAAAYAAjCB0C8AAAAASUVORK5CYII=";
+
+function getTransparentPngBlob(): Blob {
+	const bytes = Uint8Array.from(Buffer.from(TRANSPARENT_1PX_PNG_BASE64, "base64"));
+	return new Blob([bytes], { type: "image/png" });
+}
+
+export class WavespeedAIImageTextToImageTask extends WavespeedAIImageToImageTask implements ImageTextToImageTaskHelper {
+	constructor() {
+		super();
+	}
+
+	override async preparePayloadAsync(args: ImageTextToImageArgs): Promise<RequestArgs> {
+		const inputs = args.inputs ?? getTransparentPngBlob();
+		return super.preparePayloadAsync({ ...args, inputs } as ImageToImageArgs);
+	}
+}
+
+export class WavespeedAIImageTextToVideoTask extends WavespeedAIImageToVideoTask implements ImageTextToVideoTaskHelper {
+	constructor() {
+		super();
+	}
+
+	override async preparePayloadAsync(args: ImageTextToVideoArgs): Promise<RequestArgs> {
+		const inputs = args.inputs ?? getTransparentPngBlob();
+		return super.preparePayloadAsync({ ...args, inputs } as ImageToVideoArgs);
+	}
+}

package/src/providers/zai-org.ts

@@ -20,7 +20,8 @@ import {
 	InferenceClientProviderOutputError,
 } from "../errors.js";
 import { isUrl } from "../lib/isUrl.js";
-import type { BodyParams, HeaderParams } from "../types.js";
+import type { BodyParams, HeaderParams, OutputType } from "../types.js";
+import { dataUrlFromBlob } from "../utils/dataUrlFromBlob.js";
 import { delay } from "../utils/delay.js";
 import { omit } from "../utils/omit.js";
 import { BaseConversationalTask, TaskProviderHelper, type TextToImageTaskHelper } from "./providerHelper.js";
@@ -96,7 +97,7 @@ export class ZaiTextToImageTask extends TaskProviderHelper implements TextToImag
 		response: ZaiTextToImageResponse,
 		url?: string,
 		headers?: Record<string, string>,
-		outputType?: "url" | "blob" | "json"
+		outputType?: OutputType
 	): Promise<string | Blob | Record<string, unknown>> {
 		if (!url || !headers) {
 			throw new InferenceClientInputError(`URL and headers are required for 'text-to-image' task`);
@@ -179,7 +180,8 @@ export class ZaiTextToImageTask extends TaskProviderHelper implements TextToImag
 				}
 
 				const imageResponse = await fetch(imageUrl);
-				return await imageResponse.blob();
+				const blob = await imageResponse.blob();
+				return outputType === "dataUrl" ? dataUrlFromBlob(blob) : blob;
 			}
 		}
 

package/src/tasks/cv/textToImage.ts

@@ -2,13 +2,13 @@ import type { TextToImageInput } from "@huggingface/tasks";
 import { resolveProvider } from "../../lib/getInferenceProviderMapping.js";
 import { getProviderHelper } from "../../lib/getProviderHelper.js";
 import { makeRequestOptions } from "../../lib/makeRequestOptions.js";
-import type { BaseArgs, Options } from "../../types.js";
+import type { BaseArgs, Options, OutputType } from "../../types.js";
 import { innerRequest } from "../../utils/request.js";
 
 export type TextToImageArgs = BaseArgs & TextToImageInput;
 
 interface TextToImageOptions extends Options {
-	outputType?: "url" | "blob" | "json";
+	outputType?: OutputType;
 }
 
 /**
@@ -19,6 +19,10 @@ export async function textToImage(
 	args: TextToImageArgs,
 	options?: TextToImageOptions & { outputType: "url" }
 ): Promise<string>;
+export async function textToImage(
+	args: TextToImageArgs,
+	options?: TextToImageOptions & { outputType: "dataUrl" }
+): Promise<string>;
 export async function textToImage(
 	args: TextToImageArgs,
 	options?: TextToImageOptions & { outputType?: undefined | "blob" }

package/src/types.ts

@@ -163,6 +163,7 @@ export type RequestArgs = BaseArgs &
 		| ChatCompletionInput
 	) & {
 		parameters?: Record<string, unknown>;
+		urlTransform?: (url: string) => string;
 	};
 
 export type AuthMethod = "none" | "hf-token" | "credentials-include" | "provider-key";
@@ -176,11 +177,15 @@ export interface UrlParams {
 	authMethod: AuthMethod;
 	model: string;
 	task?: InferenceTask;
+	urlTransform?: (url: string) => string;
 }
 
+export type OutputType = "url" | "dataUrl" | "blob" | "json";
+
 export interface BodyParams<T extends Record<string, unknown> = Record<string, unknown>> {
 	args: T;
 	model: string;
 	mapping?: InferenceProviderMappingEntry | undefined;
 	task?: InferenceTask;
+	outputType?: OutputType;
 }

package/src/utils/dataUrlFromBlob.ts

@@ -0,0 +1,7 @@
+import { base64FromBytes } from "./base64FromBytes.js";
+
+export async function dataUrlFromBlob(blob: Blob, mimeType = "image/jpeg"): Promise<string> {
+	const buffer = await blob.arrayBuffer();
+	const base64 = base64FromBytes(new Uint8Array(buffer));
+	return `data:${mimeType};base64,${base64}`;
+}