@huggingface/tasks 0.15.8 → 0.16.0

package/src/snippets/js.ts

@@ -1,11 +1,11 @@
-import { openAIbaseUrl, type InferenceProvider } from "../inference-providers.js";
-import type { PipelineType } from "../pipelines.js";
+import { openAIbaseUrl, type SnippetInferenceProvider } from "../inference-providers.js";
+import type { PipelineType, WidgetType } from "../pipelines.js";
 import type { ChatCompletionInputMessage, GenerationParameters } from "../tasks/index.js";
 import { stringifyGenerationConfig, stringifyMessages } from "./common.js";
 import { getModelInputSnippet } from "./inputs.js";
 import type { InferenceSnippet, ModelDataMinimal } from "./types.js";
 
-const HFJS_METHODS: Record<string, string> = {
+const HFJS_METHODS: Partial<Record<WidgetType, string>> = {
 	"text-classification": "textClassification",
 	"token-classification": "tokenClassification",
 	"table-question-answering": "tableQuestionAnswering",
@@ -22,14 +22,14 @@ const HFJS_METHODS: Record<string, string> = {
 export const snippetBasic = (
 	model: ModelDataMinimal,
 	accessToken: string,
-	provider: InferenceProvider
+	provider: SnippetInferenceProvider
 ): InferenceSnippet[] => {
 	return [
 		...(model.pipeline_tag && model.pipeline_tag in HFJS_METHODS
 			? [
-					{
-						client: "huggingface.js",
-						content: `\
+				{
+					client: "huggingface.js",
+					content: `\
 import { HfInference } from "@huggingface/inference";
 
 const client = new HfInference("${accessToken || `{API_TOKEN}`}");
@@ -40,17 +40,17 @@ const output = await client.${HFJS_METHODS[model.pipeline_tag]}({
 	provider: "${provider}",
 });
 
-console.log(output)
+console.log(output);
 `,
-					},
-			  ]
+				},
+			]
 			: []),
 		{
 			client: "fetch",
 			content: `\
 async function query(data) {
 	const response = await fetch(
-		"https://api-inference.huggingface.co/models/${model.id}",
+		"https://router.huggingface.co/hf-inference/models/${model.id}",
 		{
 			headers: {
 				Authorization: "Bearer ${accessToken || `{API_TOKEN}`}",
@@ -74,7 +74,8 @@ query({"inputs": ${getModelInputSnippet(model)}}).then((response) => {
 export const snippetTextGeneration = (
 	model: ModelDataMinimal,
 	accessToken: string,
-	provider: InferenceProvider,
+	provider: SnippetInferenceProvider,
+	providerModelId?: string,
 	opts?: {
 		streaming?: boolean;
 		messages?: ChatCompletionInputMessage[];
@@ -137,9 +138,9 @@ const client = new OpenAI({
 let out = "";
 
 const stream = await client.chat.completions.create({
-	model: "${model.id}",
+	model: "${providerModelId ?? model.id}",
 	messages: ${messagesStr},
-	${configStr},
+	${configStr}
 	stream: true,
 });
 
@@ -167,7 +168,8 @@ const chatCompletion = await client.chatCompletion({
 	${configStr}
 });
 
-console.log(chatCompletion.choices[0].message);`,
+console.log(chatCompletion.choices[0].message);
+`,
 				},
 				{
 					client: "openai",
@@ -179,12 +181,13 @@ const client = new OpenAI({
 });
 
 const chatCompletion = await client.chat.completions.create({
-	model: "${model.id}",
+	model: "${providerModelId ?? model.id}",
 	messages: ${messagesStr},
 	${configStr}
 });
 
-console.log(chatCompletion.choices[0].message);`,
+console.log(chatCompletion.choices[0].message);
+`,
 				},
 			];
 		}
@@ -199,7 +202,7 @@ export const snippetZeroShotClassification = (model: ModelDataMinimal, accessTok
 			client: "fetch",
 			content: `async function query(data) {
 			const response = await fetch(
-				"https://api-inference.huggingface.co/models/${model.id}",
+				"https://router.huggingface.co/hf-inference/models/${model.id}",
 				{
 					headers: {
 						Authorization: "Bearer ${accessToken || `{API_TOKEN}`}",
@@ -214,8 +217,8 @@ export const snippetZeroShotClassification = (model: ModelDataMinimal, accessTok
 		}
 		
 		query({"inputs": ${getModelInputSnippet(
-			model
-		)}, "parameters": {"candidate_labels": ["refund", "legal", "faq"]}}).then((response) => {
+				model
+			)}, "parameters": {"candidate_labels": ["refund", "legal", "faq"]}}).then((response) => {
 			console.log(JSON.stringify(response));
 		});`,
 		},
@@ -225,7 +228,7 @@ export const snippetZeroShotClassification = (model: ModelDataMinimal, accessTok
 export const snippetTextToImage = (
 	model: ModelDataMinimal,
 	accessToken: string,
-	provider: InferenceProvider
+	provider: SnippetInferenceProvider
 ): InferenceSnippet[] => {
 	return [
 		{
@@ -246,11 +249,11 @@ const image = await client.textToImage({
 		},
 		...(provider === "hf-inference"
 			? [
-					{
-						client: "fetch",
-						content: `async function query(data) {
+				{
+					client: "fetch",
+					content: `async function query(data) {
 	const response = await fetch(
-		"https://api-inference.huggingface.co/models/${model.id}",
+		"https://router.huggingface.co/hf-inference/models/${model.id}",
 		{
 			headers: {
 				Authorization: "Bearer ${accessToken || `{API_TOKEN}`}",
@@ -266,8 +269,8 @@ const image = await client.textToImage({
 query({"inputs": ${getModelInputSnippet(model)}}).then((response) => {
 	// Use image
 });`,
-					},
-			  ]
+				},
+			]
 			: []),
 	];
 };
@@ -275,14 +278,14 @@ query({"inputs": ${getModelInputSnippet(model)}}).then((response) => {
 export const snippetTextToAudio = (
 	model: ModelDataMinimal,
 	accessToken: string,
-	provider: InferenceProvider
+	provider: SnippetInferenceProvider
 ): InferenceSnippet[] => {
 	if (provider !== "hf-inference") {
 		return [];
 	}
 	const commonSnippet = `async function query(data) {
 		const response = await fetch(
-			"https://api-inference.huggingface.co/models/${model.id}",
+			"https://router.huggingface.co/hf-inference/models/${model.id}",
 			{
 				headers: {
 					Authorization: "Bearer ${accessToken || `{API_TOKEN}`}",
@@ -329,7 +332,7 @@ export const snippetTextToAudio = (
 export const snippetAutomaticSpeechRecognition = (
 	model: ModelDataMinimal,
 	accessToken: string,
-	provider: InferenceProvider
+	provider: SnippetInferenceProvider
 ): InferenceSnippet[] => {
 	return [
 		{
@@ -357,7 +360,7 @@ console.log(output);
 export const snippetFile = (
 	model: ModelDataMinimal,
 	accessToken: string,
-	provider: InferenceProvider
+	provider: SnippetInferenceProvider
 ): InferenceSnippet[] => {
 	if (provider !== "hf-inference") {
 		return [];
@@ -368,7 +371,7 @@ export const snippetFile = (
 			content: `async function query(filename) {
 	const data = fs.readFileSync(filename);
 	const response = await fetch(
-		"https://api-inference.huggingface.co/models/${model.id}",
+		"https://router.huggingface.co/hf-inference/models/${model.id}",
 		{
 			headers: {
 				Authorization: "Bearer ${accessToken || `{API_TOKEN}`}",
@@ -395,7 +398,8 @@ export const jsSnippets: Partial<
 		(
 			model: ModelDataMinimal,
 			accessToken: string,
-			provider: InferenceProvider,
+			provider: SnippetInferenceProvider,
+			providerModelId?: string,
 			opts?: Record<string, unknown>
 		) => InferenceSnippet[]
 	>
@@ -429,10 +433,11 @@ export const jsSnippets: Partial<
 export function getJsInferenceSnippet(
 	model: ModelDataMinimal,
 	accessToken: string,
-	provider: InferenceProvider,
+	provider: SnippetInferenceProvider,
+	providerModelId?: string,
 	opts?: Record<string, unknown>
 ): InferenceSnippet[] {
 	return model.pipeline_tag && model.pipeline_tag in jsSnippets
-		? jsSnippets[model.pipeline_tag]?.(model, accessToken, provider, opts) ?? []
+		? jsSnippets[model.pipeline_tag]?.(model, accessToken, provider, providerModelId, opts) ?? []
 		: [];
 }

package/src/snippets/python.ts

@@ -1,11 +1,44 @@
-import { HF_HUB_INFERENCE_PROXY_TEMPLATE, openAIbaseUrl, type InferenceProvider } from "../inference-providers.js";
-import type { PipelineType } from "../pipelines.js";
+import {
+	openAIbaseUrl,
+	type SnippetInferenceProvider,
+} from "../inference-providers.js";
+import type { PipelineType, WidgetType } from "../pipelines.js";
 import type { ChatCompletionInputMessage, GenerationParameters } from "../tasks/index.js";
 import { stringifyGenerationConfig, stringifyMessages } from "./common.js";
 import { getModelInputSnippet } from "./inputs.js";
 import type { InferenceSnippet, ModelDataMinimal } from "./types.js";
 
-const snippetImportInferenceClient = (accessToken: string, provider: InferenceProvider): string =>
+const HFH_INFERENCE_CLIENT_METHODS: Partial<Record<WidgetType, string>> = {
+	"audio-classification": "audio_classification",
+	"audio-to-audio": "audio_to_audio",
+	"automatic-speech-recognition": "automatic_speech_recognition",
+	"text-to-speech": "text_to_speech",
+	"image-classification": "image_classification",
+	"image-segmentation": "image_segmentation",
+	"image-to-image": "image_to_image",
+	"image-to-text": "image_to_text",
+	"object-detection": "object_detection",
+	"text-to-image": "text_to_image",
+	"text-to-video": "text_to_video",
+	"zero-shot-image-classification": "zero_shot_image_classification",
+	"document-question-answering": "document_question_answering",
+	"visual-question-answering": "visual_question_answering",
+	"feature-extraction": "feature_extraction",
+	"fill-mask": "fill_mask",
+	"question-answering": "question_answering",
+	"sentence-similarity": "sentence_similarity",
+	summarization: "summarization",
+	"table-question-answering": "table_question_answering",
+	"text-classification": "text_classification",
+	"text-generation": "text_generation",
+	"token-classification": "token_classification",
+	translation: "translation",
+	"zero-shot-classification": "zero_shot_classification",
+	"tabular-classification": "tabular_classification",
+	"tabular-regression": "tabular_regression",
+};
+
+const snippetImportInferenceClient = (accessToken: string, provider: SnippetInferenceProvider): string =>
 	`\
 from huggingface_hub import InferenceClient
 
@@ -17,7 +50,8 @@ client = InferenceClient(
 export const snippetConversational = (
 	model: ModelDataMinimal,
 	accessToken: string,
-	provider: InferenceProvider,
+	provider: SnippetInferenceProvider,
+	providerModelId?: string,
 	opts?: {
 		streaming?: boolean;
 		messages?: ChatCompletionInputMessage[];
@@ -53,7 +87,7 @@ messages = ${messagesStr}
 stream = client.chat.completions.create(
 	model="${model.id}", 
 	messages=messages, 
-	${configStr},
+	${configStr}
 	stream=True
 )
 
@@ -73,9 +107,9 @@ client = OpenAI(
 messages = ${messagesStr}
 
 stream = client.chat.completions.create(
-    model="${model.id}", 
+    model="${providerModelId ?? model.id}", 
 	messages=messages, 
-	${configStr},
+	${configStr}
 	stream=True
 )
 
@@ -113,7 +147,7 @@ client = OpenAI(
 messages = ${messagesStr}
 
 completion = client.chat.completions.create(
-	model="${model.id}", 
+	model="${providerModelId ?? model.id}", 
 	messages=messages, 
 	${configStr}
 )
@@ -164,8 +198,30 @@ output = query({
 	];
 };
 
-export const snippetBasic = (model: ModelDataMinimal): InferenceSnippet[] => {
+export const snippetBasic = (
+	model: ModelDataMinimal,
+	accessToken: string,
+	provider: SnippetInferenceProvider
+): InferenceSnippet[] => {
 	return [
+		...(model.pipeline_tag && model.pipeline_tag in HFH_INFERENCE_CLIENT_METHODS
+			? [
+				{
+					client: "huggingface_hub",
+					content: `\
+${snippetImportInferenceClient(accessToken, provider)}
+
+result = client.${HFH_INFERENCE_CLIENT_METHODS[model.pipeline_tag]}(
+	model="${model.id}",
+	inputs=${getModelInputSnippet(model)},
+	provider="${provider}",
+)
+
+print(result)
+`,
+				},
+			]
+			: []),
 		{
 			client: "requests",
 			content: `\
@@ -199,7 +255,8 @@ output = query(${getModelInputSnippet(model)})`,
 export const snippetTextToImage = (
 	model: ModelDataMinimal,
 	accessToken: string,
-	provider: InferenceProvider
+	provider: SnippetInferenceProvider,
+	providerModelId?: string,
 ): InferenceSnippet[] => {
 	return [
 		{
@@ -215,28 +272,27 @@ image = client.text_to_image(
 		},
 		...(provider === "fal-ai"
 			? [
-					{
-						client: "fal-client",
-						content: `\
+				{
+					client: "fal-client",
+					content: `\
 import fal_client
 
 result = fal_client.subscribe(
-	# replace with correct id from fal.ai
-	"fal-ai/${model.id}",
+	"${providerModelId ?? model.id}",
 	arguments={
 		"prompt": ${getModelInputSnippet(model)},
 	},
 )
 print(result)
 `,
-					},
-			  ]
+				},
+			]
 			: []),
 		...(provider === "hf-inference"
 			? [
-					{
-						client: "requests",
-						content: `\
+				{
+					client: "requests",
+					content: `\
 def query(payload):
 	response = requests.post(API_URL, headers=headers, json=payload)
 	return response.content
@@ -249,8 +305,8 @@ image_bytes = query({
 import io
 from PIL import Image
 image = Image.open(io.BytesIO(image_bytes))`,
-					},
-			  ]
+				},
+			]
 			: []),
 	];
 };
@@ -337,7 +393,8 @@ export const pythonSnippets: Partial<
 		(
 			model: ModelDataMinimal,
 			accessToken: string,
-			provider: InferenceProvider,
+			provider: SnippetInferenceProvider,
+			providerModelId?: string,
 			opts?: Record<string, unknown>
 		) => InferenceSnippet[]
 	>
@@ -375,23 +432,19 @@ export const pythonSnippets: Partial<
 export function getPythonInferenceSnippet(
 	model: ModelDataMinimal,
 	accessToken: string,
-	provider: InferenceProvider,
+	provider: SnippetInferenceProvider,
+	providerModelId?: string,
 	opts?: Record<string, unknown>
 ): InferenceSnippet[] {
 	if (model.tags.includes("conversational")) {
 		// Conversational model detected, so we display a code snippet that features the Messages API
-		return snippetConversational(model, accessToken, provider, opts);
+		return snippetConversational(model, accessToken, provider, providerModelId, opts);
 	} else {
 		const snippets =
 			model.pipeline_tag && model.pipeline_tag in pythonSnippets
-				? pythonSnippets[model.pipeline_tag]?.(model, accessToken, provider) ?? []
+				? pythonSnippets[model.pipeline_tag]?.(model, accessToken, provider, providerModelId) ?? []
 				: [];
 
-		const baseUrl =
-			provider === "hf-inference"
-				? `https://api-inference.huggingface.co/models/${model.id}`
-				: HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", provider);
-
 		return snippets.map((snippet) => {
 			return {
 				...snippet,
@@ -400,7 +453,7 @@ export function getPythonInferenceSnippet(
 						? `\
 import requests
 
-API_URL = "${baseUrl}"
+API_URL = "${openAIbaseUrl(provider)}"
 headers = {"Authorization": ${accessToken ? `"Bearer ${accessToken}"` : `f"Bearer {API_TOKEN}"`}}
 
 ${snippet.content}`

package/src/tasks/audio-classification/about.md

@@ -33,7 +33,7 @@ import json
 import requests
 
 headers = {"Authorization": f"Bearer {API_TOKEN}"}
-API_URL = "https://api-inference.huggingface.co/models/superb/hubert-large-superb-er"
+API_URL = "https://router.huggingface.co/hf-inference/models/superb/hubert-large-superb-er"
 
 def query(filename):
     with open(filename, "rb") as f:

package/src/tasks/audio-to-audio/about.md

@@ -19,7 +19,7 @@ import json
 import requests
 
 headers = {"Authorization": f"Bearer {API_TOKEN}"}
-API_URL = "https://api-inference.huggingface.co/models/speechbrain/mtl-mimic-voicebank"
+API_URL = "https://router.huggingface.co/hf-inference/models/speechbrain/mtl-mimic-voicebank"
 
 def query(filename):
     with open(filename, "rb") as f:

package/src/tasks/automatic-speech-recognition/about.md

@@ -25,7 +25,7 @@ import json
 import requests
 
 headers = {"Authorization": f"Bearer {API_TOKEN}"}
-API_URL = "https://api-inference.huggingface.co/models/openai/whisper-large-v3"
+API_URL = "https://router.huggingface.co/hf-inference/models/openai/whisper-large-v3"
 
 def query(filename):
     with open(filename, "rb") as f:

package/src/tasks/feature-extraction/spec/input.json

@@ -8,11 +8,18 @@
 	"properties": {
 		"inputs": {
 			"title": "FeatureExtractionInputs",
+			"description": "The text or list of texts to embed.",
 			"oneOf": [
-				{ "type": "string" },
-				{ "type": "array", "items": { "type": "string" } }
-			],
-			"description": "The text or list of texts to embed."
+				{
+					"type": "string"
+				},
+				{
+					"type": "array",
+					"items": {
+						"type": "string"
+					}
+				}
+			]
 		},
 		"normalize": {
 			"type": "boolean",

package/src/tasks/image-text-to-text/about.md

@@ -76,7 +76,7 @@ outputs[0]["generated_text"]
 You can also use the Inference API to test image-text-to-text models. You need to use a [Hugging Face token](https://huggingface.co/settings/tokens) for authentication.
 
 ```bash
-curl https://api-inference.huggingface.co/models/meta-llama/Llama-3.2-11B-Vision-Instruct \
+curl https://router.huggingface.co/hf-inference/models/meta-llama/Llama-3.2-11B-Vision-Instruct \
 	-X POST \
 	-d '{"messages": [{"role": "user","content": [{"type": "image"}, {"type": "text", "text": "Can you describe the image?"}]}]}' \
 	-H "Content-Type: application/json" \

package/src/tasks/index.ts

@@ -174,6 +174,7 @@ export const TASKS_MODEL_LIBRARIES: Record<PipelineType, ModelLibraryKey[]> = {
 	"text-to-3d": ["diffusers"],
 	"image-to-3d": ["diffusers"],
 	"any-to-any": ["transformers"],
+	"visual-document-retrieval": ["transformers"],
 };
 
 /**
@@ -202,6 +203,7 @@ export const TASKS_DATA: Record<PipelineType, TaskData | undefined> = {
 	"automatic-speech-recognition": getData("automatic-speech-recognition", automaticSpeechRecognition),
 	"depth-estimation": getData("depth-estimation", depthEstimation),
 	"document-question-answering": getData("document-question-answering", documentQuestionAnswering),
+	"visual-document-retrieval": getData("visual-document-retrieval", placeholder),
 	"feature-extraction": getData("feature-extraction", featureExtraction),
 	"fill-mask": getData("fill-mask", fillMask),
 	"graph-ml": undefined,

package/src/tasks/sentence-similarity/about.md

@@ -22,7 +22,7 @@ You can infer with Passage Ranking models using [Inference Endpoints](https://hu
 import json
 import requests
 
-API_URL = "https://api-inference.huggingface.co/models/sentence-transformers/msmarco-distilbert-base-tas-b"
+API_URL = "https://router.huggingface.co/hf-inference/models/sentence-transformers/msmarco-distilbert-base-tas-b"
 headers = {"Authorization": f"Bearer {api_token}"}
 
 def query(payload):
@@ -51,7 +51,7 @@ Semantic Textual Similarity is the task of evaluating how similar two texts are
 import json
 import requests
 
-API_URL = "https://api-inference.huggingface.co/models/sentence-transformers/all-MiniLM-L6-v2"
+API_URL = "https://router.huggingface.co/hf-inference/models/sentence-transformers/all-MiniLM-L6-v2"
 headers = {"Authorization": f"Bearer {api_token}"}
 
 def query(payload):

package/src/tasks/text-generation/data.ts

@@ -61,9 +61,7 @@ const taskData: TaskDataCustom = {
 		},
 	],
 	models: [
-		{	description: "A text-generation model trained to follow instructions.",
-			id: "google/gemma-2-2b-it",
-		},
+		{ description: "A text-generation model trained to follow instructions.", id: "google/gemma-2-2b-it" },
 		{
 			description: "Smaller variant of one of the most powerful models.",
 			id: "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",

package/src/tasks/text-to-speech/about.md

@@ -19,7 +19,7 @@ import json
 import requests
 
 headers = {"Authorization": f"Bearer {API_TOKEN}"}
-API_URL = "https://api-inference.huggingface.co/models/microsoft/speecht5_tts"
+API_URL = "https://router.huggingface.co/hf-inference/models/microsoft/speecht5_tts"
 
 def query(payload):
 	response = requests.post(API_URL, headers=headers, json=payload)