@huggingface/inference 3.10.0 → 3.12.0

package/src/tasks/nlp/textGenerationStream.ts

@@ -1,4 +1,5 @@
 import type { TextGenerationInput } from "@huggingface/tasks";
+import { resolveProvider } from "../../lib/getInferenceProviderMapping";
 import { getProviderHelper } from "../../lib/getProviderHelper";
 import type { BaseArgs, Options } from "../../types";
 import { innerStreamingRequest } from "../../utils/request";
@@ -90,7 +91,8 @@ export async function* textGenerationStream(
 	args: BaseArgs & TextGenerationInput,
 	options?: Options
 ): AsyncGenerator<TextGenerationStreamOutput> {
-	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "text-generation");
+	const provider = await resolveProvider(args.provider, args.model, args.endpointUrl);
+	const providerHelper = getProviderHelper(provider, "text-generation");
 	yield* innerStreamingRequest<TextGenerationStreamOutput>(args, providerHelper, {
 		...options,
 		task: "text-generation",

package/src/tasks/nlp/tokenClassification.ts

@@ -1,4 +1,5 @@
 import type { TokenClassificationInput, TokenClassificationOutput } from "@huggingface/tasks";
+import { resolveProvider } from "../../lib/getInferenceProviderMapping";
 import { getProviderHelper } from "../../lib/getProviderHelper";
 import type { BaseArgs, Options } from "../../types";
 import { innerRequest } from "../../utils/request";
@@ -12,7 +13,8 @@ export async function tokenClassification(
 	args: TokenClassificationArgs,
 	options?: Options
 ): Promise<TokenClassificationOutput> {
-	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "token-classification");
+	const provider = await resolveProvider(args.provider, args.model, args.endpointUrl);
+	const providerHelper = getProviderHelper(provider, "token-classification");
 	const { data: res } = await innerRequest<TokenClassificationOutput[number] | TokenClassificationOutput>(
 		args,
 		providerHelper,

package/src/tasks/nlp/translation.ts

@@ -1,4 +1,5 @@
 import type { TranslationInput, TranslationOutput } from "@huggingface/tasks";
+import { resolveProvider } from "../../lib/getInferenceProviderMapping";
 import { getProviderHelper } from "../../lib/getProviderHelper";
 import type { BaseArgs, Options } from "../../types";
 import { innerRequest } from "../../utils/request";
@@ -8,7 +9,8 @@ export type TranslationArgs = BaseArgs & TranslationInput;
  * This task is well known to translate text from one language to another. Recommended model: Helsinki-NLP/opus-mt-ru-en.
  */
 export async function translation(args: TranslationArgs, options?: Options): Promise<TranslationOutput> {
-	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "translation");
+	const provider = await resolveProvider(args.provider, args.model, args.endpointUrl);
+	const providerHelper = getProviderHelper(provider, "translation");
 	const { data: res } = await innerRequest<TranslationOutput>(args, providerHelper, {
 		...options,
 		task: "translation",

package/src/tasks/nlp/zeroShotClassification.ts

@@ -1,4 +1,5 @@
 import type { ZeroShotClassificationInput, ZeroShotClassificationOutput } from "@huggingface/tasks";
+import { resolveProvider } from "../../lib/getInferenceProviderMapping";
 import { getProviderHelper } from "../../lib/getProviderHelper";
 import type { BaseArgs, Options } from "../../types";
 import { innerRequest } from "../../utils/request";
@@ -12,7 +13,8 @@ export async function zeroShotClassification(
 	args: ZeroShotClassificationArgs,
 	options?: Options
 ): Promise<ZeroShotClassificationOutput> {
-	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "zero-shot-classification");
+	const provider = await resolveProvider(args.provider, args.model, args.endpointUrl);
+	const providerHelper = getProviderHelper(provider, "zero-shot-classification");
 	const { data: res } = await innerRequest<ZeroShotClassificationOutput[number] | ZeroShotClassificationOutput>(
 		args,
 		providerHelper,

package/src/tasks/tabular/tabularClassification.ts

@@ -1,3 +1,4 @@
+import { resolveProvider } from "../../lib/getInferenceProviderMapping";
 import { getProviderHelper } from "../../lib/getProviderHelper";
 import type { BaseArgs, Options } from "../../types";
 import { innerRequest } from "../../utils/request";
@@ -25,7 +26,8 @@ export async function tabularClassification(
 	args: TabularClassificationArgs,
 	options?: Options
 ): Promise<TabularClassificationOutput> {
-	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "tabular-classification");
+	const provider = await resolveProvider(args.provider, args.model, args.endpointUrl);
+	const providerHelper = getProviderHelper(provider, "tabular-classification");
 	const { data: res } = await innerRequest<TabularClassificationOutput>(args, providerHelper, {
 		...options,
 		task: "tabular-classification",

package/src/tasks/tabular/tabularRegression.ts

@@ -1,3 +1,4 @@
+import { resolveProvider } from "../../lib/getInferenceProviderMapping";
 import { getProviderHelper } from "../../lib/getProviderHelper";
 import type { BaseArgs, Options } from "../../types";
 import { innerRequest } from "../../utils/request";
@@ -25,7 +26,8 @@ export async function tabularRegression(
 	args: TabularRegressionArgs,
 	options?: Options
 ): Promise<TabularRegressionOutput> {
-	const providerHelper = getProviderHelper(args.provider ?? "hf-inference", "tabular-regression");
+	const provider = await resolveProvider(args.provider, args.model, args.endpointUrl);
+	const providerHelper = getProviderHelper(provider, "tabular-regression");
 	const { data: res } = await innerRequest<TabularRegressionOutput>(args, providerHelper, {
 		...options,
 		task: "tabular-regression",

package/src/types.ts

@@ -57,8 +57,12 @@ export const INFERENCE_PROVIDERS = [
 	"together",
 ] as const;
 
+export const PROVIDERS_OR_POLICIES = [...INFERENCE_PROVIDERS, "auto"] as const;
+
 export type InferenceProvider = (typeof INFERENCE_PROVIDERS)[number];
 
+export type InferenceProviderOrPolicy = (typeof PROVIDERS_OR_POLICIES)[number];
+
 export interface BaseArgs {
 	/**
 	 * The access token to use. Without it, you'll get rate-limited quickly.
@@ -80,18 +84,18 @@ export interface BaseArgs {
 	model?: ModelId;
 
 	/**
-	 * The URL of the endpoint to use. If not specified, will call huggingface.co/api/tasks to get the default endpoint for the task.
+	 * The URL of the endpoint to use.
 	 *
-	 * If specified, will use this URL instead of the default one.
+	 * If not specified, will call the default router.huggingface.co Inference Providers endpoint.
 	 */
 	endpointUrl?: string;
 
 	/**
 	 * Set an Inference provider to run this model on.
 	 *
-	 * Defaults to the first provider in your user settings that is compatible with this model.
+	 * Defaults to "auto" i.e. the first of the providers available for the model, sorted by the user's order in https://hf.co/settings/inference-providers.
 	 */
-	provider?: InferenceProvider;
+	provider?: InferenceProviderOrPolicy;
 }
 
 export type RequestArgs = BaseArgs &

package/src/utils/typedEntries.ts

@@ -0,0 +1,5 @@
+export function typedEntries<T extends { [s: string]: T[keyof T] } | ArrayLike<T[keyof T]>>(
+	obj: T
+): [keyof T, T[keyof T]][] {
+	return Object.entries(obj) as [keyof T, T[keyof T]][];
+}