@huggingface/tasks - npm Package Compare versions

Comparing version 0.15.8 to 0.15.9

dist/commonjs/inference-providers.d.ts

@@ -1,4 +0,4 @@
-export declare const INFERENCE_PROVIDERS: readonly ["hf-inference", "fal-ai", "replicate", "sambanova", "together"];
-export type InferenceProvider = (typeof INFERENCE_PROVIDERS)[number];
-export declare const HF_HUB_INFERENCE_PROXY_TEMPLATE = "https://huggingface.co/api/inference-proxy/{{PROVIDER}}";
+declare const INFERENCE_PROVIDERS: readonly ["fal-ai", "fireworks-ai", "hf-inference", "hyperbolic", "replicate", "sambanova", "together"];
+export type SnippetInferenceProvider = (typeof INFERENCE_PROVIDERS)[number] | string;
+export declare const HF_HUB_INFERENCE_PROXY_TEMPLATE = "https://router.huggingface.co/{{PROVIDER}}";
 /**
@@ -9,3 +9,4 @@ * URL to set as baseUrl in the OpenAI SDK.
  */
-export declare function openAIbaseUrl(provider: InferenceProvider): string;
+export declare function openAIbaseUrl(provider: SnippetInferenceProvider): string;
+export {};
 //# sourceMappingURL=inference-providers.d.ts.map

dist/commonjs/inference-providers.js

 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.HF_HUB_INFERENCE_PROXY_TEMPLATE = exports.INFERENCE_PROVIDERS = void 0;
+exports.HF_HUB_INFERENCE_PROXY_TEMPLATE = void 0;
 exports.openAIbaseUrl = openAIbaseUrl;
-exports.INFERENCE_PROVIDERS = ["hf-inference", "fal-ai", "replicate", "sambanova", "together"];
-exports.HF_HUB_INFERENCE_PROXY_TEMPLATE = `https://huggingface.co/api/inference-proxy/{{PROVIDER}}`;
+/// This list is for illustration purposes only.
+/// in the `tasks` sub-package, we do not need actual strong typing of the inference providers.
+const INFERENCE_PROVIDERS = [
+    "fal-ai",
+    "fireworks-ai",
+    "hf-inference",
+    "hyperbolic",
+    "replicate",
+    "sambanova",
+    "together",
+];
+exports.HF_HUB_INFERENCE_PROXY_TEMPLATE = `https://router.huggingface.co/{{PROVIDER}}`;
 /**
@@ -13,5 +23,4 @@ * URL to set as baseUrl in the OpenAI SDK.
 function openAIbaseUrl(provider) {
-    return provider === "hf-inference"
-        ? "https://api-inference.huggingface.co/v1/"
-        : exports.HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", provider);
+    const url = exports.HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", provider);
+    return provider === "hf-inference" ? `${url}/v1` : url;
 }

dist/commonjs/library-to-tasks.d.ts

@@ -5,3 +5,3 @@ import type { ModelLibraryKey } from "./model-libraries.js";
  * Mapping from library name to its supported tasks.
- * Inference API (serverless) should be disabled for all other (library, task) pairs beyond this mapping.
+ * HF-Inference API (serverless) should be disabled for all other (library, task) pairs beyond this mapping.
  * This mapping is partially generated automatically by "python-api-export-tasks" action in
@@ -8,0 +8,0 @@ * huggingface/api-inference-community repo upon merge. For transformers, the mapping is manually

dist/commonjs/library-to-tasks.js

@@ -6,3 +6,3 @@ "use strict";
  * Mapping from library name to its supported tasks.
- * Inference API (serverless) should be disabled for all other (library, task) pairs beyond this mapping.
+ * HF-Inference API (serverless) should be disabled for all other (library, task) pairs beyond this mapping.
  * This mapping is partially generated automatically by "python-api-export-tasks" action in
@@ -9,0 +9,0 @@ * huggingface/api-inference-community repo upon merge. For transformers, the mapping is manually

dist/commonjs/model-libraries-snippets.d.ts

@@ -6,2 +6,3 @@ import type { ModelData } from "./model-data.js";
 export declare const audioseal: (model: ModelData) => string[];
+export declare const ben2: (model: ModelData) => string[];
 export declare const bertopic: (model: ModelData) => string[];
@@ -8,0 +9,0 @@ export declare const bm25s: (model: ModelData) => string[];

dist/commonjs/model-libraries-snippets.js

 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.terratorch = exports.speechbrain = exports.stanza = exports.span_marker = exports.spacy = exports.setfit = exports.sentenceTransformers = exports.sampleFactory = exports.sam2 = exports.fastai = exports.stable_audio_tools = exports.sklearn = exports.seed_story = exports.saelens = exports.timm = exports.tensorflowtts = exports.relik = exports.pyannote_audio = exports.pyannote_audio_pipeline = exports.paddlenlp = exports.open_clip = exports.mesh_anything = exports.mars5_tts = exports.mamba_ssm = exports.tf_keras = exports.llama_cpp_python = exports.keras_hub = exports.keras = exports.htrflow = exports.gliner = exports.flair = exports.fairseq = exports.espnet = exports.espnetASR = exports.espnetTTS = exports.edsnlp = exports.cartesia_mlx = exports.cartesia_pytorch = exports.diffusionkit = exports.diffusers = exports.derm_foundation = exports.depth_pro = exports.depth_anything_v2 = exports.cxr_foundation = exports.bm25s = exports.bertopic = exports.audioseal = exports.asteroid = exports.allennlp = exports.adapters = void 0;
-exports.hezar = exports.threedtopia_xl = exports.whisperkit = exports.audiocraft = exports.anemoi = exports.pythae = exports.pxia = exports.nemo = exports.model2vec = exports.mlxim = exports.mlx = exports.swarmformer = exports.birefnet = exports.ultralytics = exports.chattts = exports.voicecraft = exports.vfimamba = exports.sana = exports.sentis = exports.mlAgents = exports.stableBaselines3 = exports.fasttext = exports.peft = exports.transformersJS = exports.transformers = void 0;
+exports.speechbrain = exports.stanza = exports.span_marker = exports.spacy = exports.setfit = exports.sentenceTransformers = exports.sampleFactory = exports.sam2 = exports.fastai = exports.stable_audio_tools = exports.sklearn = exports.seed_story = exports.saelens = exports.timm = exports.tensorflowtts = exports.relik = exports.pyannote_audio = exports.pyannote_audio_pipeline = exports.paddlenlp = exports.open_clip = exports.mesh_anything = exports.mars5_tts = exports.mamba_ssm = exports.tf_keras = exports.llama_cpp_python = exports.keras_hub = exports.keras = exports.htrflow = exports.gliner = exports.flair = exports.fairseq = exports.espnet = exports.espnetASR = exports.espnetTTS = exports.edsnlp = exports.cartesia_mlx = exports.cartesia_pytorch = exports.diffusionkit = exports.diffusers = exports.derm_foundation = exports.depth_pro = exports.depth_anything_v2 = exports.cxr_foundation = exports.bm25s = exports.bertopic = exports.ben2 = exports.audioseal = exports.asteroid = exports.allennlp = exports.adapters = void 0;
+exports.hezar = exports.threedtopia_xl = exports.whisperkit = exports.audiocraft = exports.anemoi = exports.pythae = exports.pxia = exports.nemo = exports.model2vec = exports.mlxim = exports.mlx = exports.swarmformer = exports.birefnet = exports.ultralytics = exports.chattts = exports.voicecraft = exports.vfimamba = exports.sana = exports.sentis = exports.mlAgents = exports.stableBaselines3 = exports.fasttext = exports.peft = exports.transformersJS = exports.transformers = exports.terratorch = void 0;
 const library_to_tasks_js_1 = require("./library-to-tasks.js");
@@ -77,2 +77,16 @@ const inputs_js_1 = require("./snippets/inputs.js");
 }
+const ben2 = (model) => [
+    `import requests
+from PIL import Image
+from ben2 import AutoModel
+
+url = "https://huggingface.co/datasets/mishig/sample_images/resolve/main/teapot.jpg"
+image = Image.open(requests.get(url, stream=True).raw)
+
+model = AutoModel.from_pretrained("${model.id}")
+model.to("cuda").eval()
+foreground = model.inference(image)
+`,
+];
+exports.ben2 = ben2;
 const bertopic = (model) => [
@@ -79,0 +93,0 @@ `from bertopic import BERTopic

dist/commonjs/model-libraries.d.ts

@@ -108,2 +108,9 @@ import type { ModelData } from "./model-data.js";
     };
+    ben2: {
+        prettyLabel: string;
+        repoName: string;
+        repoUrl: string;
+        snippets: (model: ModelData) => string[];
+        filter: false;
+    };
     bertopic: {
@@ -898,3 +905,3 @@ prettyLabel: string;
 export declare const ALL_MODEL_LIBRARY_KEYS: ModelLibraryKey[];
-export declare const ALL_DISPLAY_MODEL_LIBRARY_KEYS: ("adapter-transformers" | "allennlp" | "anemoi" | "asteroid" | "audiocraft" | "audioseal" | "bertopic" | "big_vision" | "birder" | "birefnet" | "bm25s" | "champ" | "chat_tts" | "colpali" | "cosmos" | "cxr-foundation" | "deepforest" | "depth-anything-v2" | "depth-pro" | "derm-foundation" | "diffree" | "diffusers" | "diffusionkit" | "doctr" | "cartesia_pytorch" | "cartesia_mlx" | "clipscope" | "cosyvoice" | "cotracker" | "edsnlp" | "elm" | "espnet" | "fairseq" | "fastai" | "fasttext" | "flair" | "gemma.cpp" | "gliner" | "glyph-byt5" | "grok" | "hallo" | "hezar" | "htrflow" | "hunyuan-dit" | "imstoucan" | "keras" | "tf-keras" | "keras-hub" | "k2" | "liveportrait" | "llama-cpp-python" | "mini-omni2" | "mindspore" | "mamba-ssm" | "mars5-tts" | "mesh-anything" | "mitie" | "ml-agents" | "mlx" | "mlx-image" | "mlc-llm" | "model2vec" | "moshi" | "nemo" | "open-oasis" | "open_clip" | "paddlenlp" | "peft" | "pxia" | "pyannote-audio" | "py-feat" | "pythae" | "recurrentgemma" | "relik" | "refiners" | "reverb" | "saelens" | "sam2" | "sample-factory" | "sapiens" | "sentence-transformers" | "setfit" | "sklearn" | "spacy" | "span-marker" | "speechbrain" | "ssr-speech" | "stable-audio-tools" | "diffusion-single-file" | "seed-story" | "soloaudio" | "stable-baselines3" | "stanza" | "swarmformer" | "f5-tts" | "genmo" | "tensorflowtts" | "tabpfn" | "terratorch" | "tic-clip" | "timesfm" | "timm" | "transformers" | "transformers.js" | "trellis" | "ultralytics" | "unity-sentis" | "sana" | "vfi-mamba" | "voicecraft" | "whisperkit" | "yolov10" | "3dtopia-xl")[];
+export declare const ALL_DISPLAY_MODEL_LIBRARY_KEYS: ("adapter-transformers" | "allennlp" | "anemoi" | "asteroid" | "audiocraft" | "audioseal" | "ben2" | "bertopic" | "big_vision" | "birder" | "birefnet" | "bm25s" | "champ" | "chat_tts" | "colpali" | "cosmos" | "cxr-foundation" | "deepforest" | "depth-anything-v2" | "depth-pro" | "derm-foundation" | "diffree" | "diffusers" | "diffusionkit" | "doctr" | "cartesia_pytorch" | "cartesia_mlx" | "clipscope" | "cosyvoice" | "cotracker" | "edsnlp" | "elm" | "espnet" | "fairseq" | "fastai" | "fasttext" | "flair" | "gemma.cpp" | "gliner" | "glyph-byt5" | "grok" | "hallo" | "hezar" | "htrflow" | "hunyuan-dit" | "imstoucan" | "keras" | "tf-keras" | "keras-hub" | "k2" | "liveportrait" | "llama-cpp-python" | "mini-omni2" | "mindspore" | "mamba-ssm" | "mars5-tts" | "mesh-anything" | "mitie" | "ml-agents" | "mlx" | "mlx-image" | "mlc-llm" | "model2vec" | "moshi" | "nemo" | "open-oasis" | "open_clip" | "paddlenlp" | "peft" | "pxia" | "pyannote-audio" | "py-feat" | "pythae" | "recurrentgemma" | "relik" | "refiners" | "reverb" | "saelens" | "sam2" | "sample-factory" | "sapiens" | "sentence-transformers" | "setfit" | "sklearn" | "spacy" | "span-marker" | "speechbrain" | "ssr-speech" | "stable-audio-tools" | "diffusion-single-file" | "seed-story" | "soloaudio" | "stable-baselines3" | "stanza" | "swarmformer" | "f5-tts" | "genmo" | "tensorflowtts" | "tabpfn" | "terratorch" | "tic-clip" | "timesfm" | "timm" | "transformers" | "transformers.js" | "trellis" | "ultralytics" | "unity-sentis" | "sana" | "vfi-mamba" | "voicecraft" | "whisperkit" | "yolov10" | "3dtopia-xl")[];
 //# sourceMappingURL=model-libraries.d.ts.map

dist/commonjs/model-libraries.js

@@ -94,2 +94,9 @@ "use strict";
     },
+    ben2: {
+        prettyLabel: "BEN2",
+        repoName: "BEN2",
+        repoUrl: "https://github.com/PramaLLC/BEN2",
+        snippets: snippets.ben2,
+        filter: false,
+    },
     bertopic: {
@@ -96,0 +103,0 @@ prettyLabel: "BERTopic",

dist/commonjs/pipelines.d.ts

@@ -412,2 +412,8 @@ export declare const MODALITIES: readonly ["multimodal", "nlp", "cv", "audio", "tabular", "rl", "other"];
     };
+    "visual-document-retrieval": {
+        name: string;
+        modality: "multimodal";
+        color: "yellow";
+        hideInDatasets: true;
+    };
     "any-to-any": {
@@ -431,3 +437,3 @@ name: string;
 export declare const SUBTASK_TYPES: string[];
-export declare const PIPELINE_TYPES_SET: Set<"other" | "text-classification" | "token-classification" | "table-question-answering" | "question-answering" | "zero-shot-classification" | "translation" | "summarization" | "feature-extraction" | "text-generation" | "text2text-generation" | "fill-mask" | "sentence-similarity" | "text-to-speech" | "text-to-audio" | "automatic-speech-recognition" | "audio-to-audio" | "audio-classification" | "audio-text-to-text" | "voice-activity-detection" | "depth-estimation" | "image-classification" | "object-detection" | "image-segmentation" | "text-to-image" | "image-to-text" | "image-to-image" | "image-to-video" | "unconditional-image-generation" | "video-classification" | "reinforcement-learning" | "robotics" | "tabular-classification" | "tabular-regression" | "tabular-to-text" | "table-to-text" | "multiple-choice" | "text-retrieval" | "time-series-forecasting" | "text-to-video" | "image-text-to-text" | "visual-question-answering" | "document-question-answering" | "zero-shot-image-classification" | "graph-ml" | "mask-generation" | "zero-shot-object-detection" | "text-to-3d" | "image-to-3d" | "image-feature-extraction" | "video-text-to-text" | "keypoint-detection" | "any-to-any">;
+export declare const PIPELINE_TYPES_SET: Set<"other" | "text-classification" | "token-classification" | "table-question-answering" | "question-answering" | "zero-shot-classification" | "translation" | "summarization" | "feature-extraction" | "text-generation" | "text2text-generation" | "fill-mask" | "sentence-similarity" | "text-to-speech" | "text-to-audio" | "automatic-speech-recognition" | "audio-to-audio" | "audio-classification" | "audio-text-to-text" | "voice-activity-detection" | "depth-estimation" | "image-classification" | "object-detection" | "image-segmentation" | "text-to-image" | "image-to-text" | "image-to-image" | "image-to-video" | "unconditional-image-generation" | "video-classification" | "reinforcement-learning" | "robotics" | "tabular-classification" | "tabular-regression" | "tabular-to-text" | "table-to-text" | "multiple-choice" | "text-retrieval" | "time-series-forecasting" | "text-to-video" | "image-text-to-text" | "visual-question-answering" | "document-question-answering" | "zero-shot-image-classification" | "graph-ml" | "mask-generation" | "zero-shot-object-detection" | "text-to-3d" | "image-to-3d" | "image-feature-extraction" | "video-text-to-text" | "keypoint-detection" | "visual-document-retrieval" | "any-to-any">;
 //# sourceMappingURL=pipelines.d.ts.map

dist/commonjs/pipelines.js

@@ -633,2 +633,8 @@ "use strict";
     },
+    "visual-document-retrieval": {
+        name: "Visual Document Retrieval",
+        modality: "multimodal",
+        color: "yellow",
+        hideInDatasets: true,
+    },
     "any-to-any": {
@@ -635,0 +641,0 @@ name: "Any-to-Any",

dist/commonjs/snippets/common.js

@@ -21,4 +21,4 @@ "use strict";
     return Object.entries(config)
-        .map(([key, val]) => `${quote}${key}${quote}${opts.attributeValueConnector}${val}`)
-        .join(`,${opts.indent}`);
+        .map(([key, val]) => `${quote}${key}${quote}${opts.attributeValueConnector}${val},`)
+        .join(`${opts.indent}`);
 }

dist/commonjs/snippets/curl.d.ts

@@ -1,7 +0,7 @@
-import { type InferenceProvider } from "../inference-providers.js";
+import { type SnippetInferenceProvider } from "../inference-providers.js";
 import type { PipelineType } from "../pipelines.js";
 import type { ChatCompletionInputMessage, GenerationParameters } from "../tasks/index.js";
 import type { InferenceSnippet, ModelDataMinimal } from "./types.js";
-export declare const snippetBasic: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
-export declare const snippetTextGeneration: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: {
+export declare const snippetBasic: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
+export declare const snippetTextGeneration: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: {
     streaming?: boolean;
@@ -13,6 +13,6 @@ messages?: ChatCompletionInputMessage[];
 }) => InferenceSnippet[];
-export declare const snippetZeroShotClassification: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
-export declare const snippetFile: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
-export declare const curlSnippets: Partial<Record<PipelineType, (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: Record<string, unknown>) => InferenceSnippet[]>>;
-export declare function getCurlInferenceSnippet(model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: Record<string, unknown>): InferenceSnippet[];
+export declare const snippetZeroShotClassification: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
+export declare const snippetFile: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
+export declare const curlSnippets: Partial<Record<PipelineType, (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: Record<string, unknown>) => InferenceSnippet[]>>;
+export declare function getCurlInferenceSnippet(model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: Record<string, unknown>): InferenceSnippet[];
 //# sourceMappingURL=curl.d.ts.map

dist/commonjs/snippets/curl.js

@@ -16,3 +16,3 @@ "use strict";
             content: `\
-curl https://api-inference.huggingface.co/models/${model.id} \\
+curl https://router.huggingface.co/hf-inference/models/${model.id} \\
 	-X POST \\
@@ -29,3 +29,3 @@ -d '{"inputs": ${(0, inputs_js_1.getModelInputSnippet)(model, true)}}' \\
         const baseUrl = provider === "hf-inference"
-            ? `https://api-inference.huggingface.co/models/${model.id}/v1/chat/completions`
+            ? `https://router.huggingface.co/hf-inference/models/${model.id}/v1/chat/completions`
             : inference_providers_js_1.HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", provider) + "/v1/chat/completions";
@@ -58,3 +58,3 @@ // Conversational model detected, so we display a code snippet that features the Messages API
                     attributeValueConnector: ": ",
-                })},
+                })}
     "stream": ${!!streaming}
@@ -77,3 +77,3 @@ }'`,
             client: "curl",
-            content: `curl https://api-inference.huggingface.co/models/${model.id} \\
+            content: `curl https://router.huggingface.co/hf-inference/models/${model.id} \\
 	-X POST \\
@@ -94,3 +94,3 @@ -d '{"inputs": ${(0, inputs_js_1.getModelInputSnippet)(model, true)}, "parameters": {"candidate_labels": ["refund", "legal", "faq"]}}' \\
             client: "curl",
-            content: `curl https://api-inference.huggingface.co/models/${model.id} \\
+            content: `curl https://router.huggingface.co/hf-inference/models/${model.id} \\
 	-X POST \\
@@ -97,0 +97,0 @@ --data-binary '@${(0, inputs_js_1.getModelInputSnippet)(model, true, true)}' \\

dist/commonjs/snippets/js.d.ts

@@ -1,7 +0,7 @@
-import { type InferenceProvider } from "../inference-providers.js";
+import { type SnippetInferenceProvider } from "../inference-providers.js";
 import type { PipelineType } from "../pipelines.js";
 import type { ChatCompletionInputMessage, GenerationParameters } from "../tasks/index.js";
 import type { InferenceSnippet, ModelDataMinimal } from "./types.js";
-export declare const snippetBasic: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
-export declare const snippetTextGeneration: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: {
+export declare const snippetBasic: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
+export declare const snippetTextGeneration: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: {
     streaming?: boolean;
@@ -14,8 +14,8 @@ messages?: ChatCompletionInputMessage[];
 export declare const snippetZeroShotClassification: (model: ModelDataMinimal, accessToken: string) => InferenceSnippet[];
-export declare const snippetTextToImage: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
-export declare const snippetTextToAudio: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
-export declare const snippetAutomaticSpeechRecognition: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
-export declare const snippetFile: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
-export declare const jsSnippets: Partial<Record<PipelineType, (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: Record<string, unknown>) => InferenceSnippet[]>>;
-export declare function getJsInferenceSnippet(model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: Record<string, unknown>): InferenceSnippet[];
+export declare const snippetTextToImage: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
+export declare const snippetTextToAudio: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
+export declare const snippetAutomaticSpeechRecognition: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
+export declare const snippetFile: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
+export declare const jsSnippets: Partial<Record<PipelineType, (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: Record<string, unknown>) => InferenceSnippet[]>>;
+export declare function getJsInferenceSnippet(model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: Record<string, unknown>): InferenceSnippet[];
 //# sourceMappingURL=js.d.ts.map

dist/commonjs/snippets/js.js

@@ -38,3 +38,3 @@ "use strict";
 
-console.log(output)
+console.log(output);
 `,
@@ -49,3 +49,3 @@ },
 	const response = await fetch(
-		"https://api-inference.huggingface.co/models/${model.id}",
+		"https://router.huggingface.co/hf-inference/models/${model.id}",
 		{
@@ -126,3 +126,3 @@ headers: {
 	messages: ${messagesStr},
-	${configStr},
+	${configStr}
 	stream: true,
@@ -156,3 +156,4 @@ });
 
-console.log(chatCompletion.choices[0].message);`,
+console.log(chatCompletion.choices[0].message);
+`,
                 },
@@ -174,3 +175,4 @@ {
 
-console.log(chatCompletion.choices[0].message);`,
+console.log(chatCompletion.choices[0].message);
+`,
                 },
@@ -191,3 +193,3 @@ ];
 			const response = await fetch(
-				"https://api-inference.huggingface.co/models/${model.id}",
+				"https://router.huggingface.co/hf-inference/models/${model.id}",
 				{
@@ -237,3 +239,3 @@ headers: {
 	const response = await fetch(
-		"https://api-inference.huggingface.co/models/${model.id}",
+		"https://router.huggingface.co/hf-inference/models/${model.id}",
 		{
@@ -266,3 +268,3 @@ headers: {
 		const response = await fetch(
-			"https://api-inference.huggingface.co/models/${model.id}",
+			"https://router.huggingface.co/hf-inference/models/${model.id}",
 			{
@@ -344,3 +346,3 @@ headers: {
 	const response = await fetch(
-		"https://api-inference.huggingface.co/models/${model.id}",
+		"https://router.huggingface.co/hf-inference/models/${model.id}",
 		{
@@ -347,0 +349,0 @@ headers: {

dist/commonjs/snippets/python.d.ts

@@ -1,6 +0,6 @@
-import { type InferenceProvider } from "../inference-providers.js";
+import { type SnippetInferenceProvider } from "../inference-providers.js";
 import type { PipelineType } from "../pipelines.js";
 import type { ChatCompletionInputMessage, GenerationParameters } from "../tasks/index.js";
 import type { InferenceSnippet, ModelDataMinimal } from "./types.js";
-export declare const snippetConversational: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: {
+export declare const snippetConversational: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: {
     streaming?: boolean;
@@ -14,10 +14,10 @@ messages?: ChatCompletionInputMessage[];
 export declare const snippetZeroShotImageClassification: (model: ModelDataMinimal) => InferenceSnippet[];
-export declare const snippetBasic: (model: ModelDataMinimal) => InferenceSnippet[];
+export declare const snippetBasic: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
 export declare const snippetFile: (model: ModelDataMinimal) => InferenceSnippet[];
-export declare const snippetTextToImage: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
+export declare const snippetTextToImage: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
 export declare const snippetTabular: (model: ModelDataMinimal) => InferenceSnippet[];
 export declare const snippetTextToAudio: (model: ModelDataMinimal) => InferenceSnippet[];
 export declare const snippetDocumentQuestionAnswering: (model: ModelDataMinimal) => InferenceSnippet[];
-export declare const pythonSnippets: Partial<Record<PipelineType, (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: Record<string, unknown>) => InferenceSnippet[]>>;
-export declare function getPythonInferenceSnippet(model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: Record<string, unknown>): InferenceSnippet[];
+export declare const pythonSnippets: Partial<Record<PipelineType, (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: Record<string, unknown>) => InferenceSnippet[]>>;
+export declare function getPythonInferenceSnippet(model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: Record<string, unknown>): InferenceSnippet[];
 //# sourceMappingURL=python.d.ts.map

dist/commonjs/snippets/python.js

@@ -8,2 +8,31 @@ "use strict";
 const inputs_js_1 = require("./inputs.js");
+const HFH_INFERENCE_CLIENT_METHODS = {
+    "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, provider) => `\
@@ -42,3 +71,3 @@ from huggingface_hub import InferenceClient
 	messages=messages, 
-	${configStr},
+	${configStr}
 	stream=True
@@ -65,3 +94,3 @@ )
 	messages=messages, 
-	${configStr},
+	${configStr}
 	stream=True
@@ -156,4 +185,22 @@ )
 exports.snippetZeroShotImageClassification = snippetZeroShotImageClassification;
-const snippetBasic = (model) => {
+const snippetBasic = (model, accessToken, provider) => {
     return [
+        ...(model.pipeline_tag && model.pipeline_tag in HFH_INFERENCE_CLIENT_METHODS
+            ? [
+                {
+                    client: "huggingface_hub",
+                    content: `\
+${snippetImportInferenceClient(accessToken, provider)}
+
+result = client.${HFH_INFERENCE_CLIENT_METHODS[model.pipeline_tag]}(
+	model="${model.id}",
+	inputs=${(0, inputs_js_1.getModelInputSnippet)(model)},
+	provider="${provider}",
+)
+
+print(result)
+`,
+                },
+            ]
+            : []),
         {
@@ -360,5 +407,2 @@ client: "requests",
             : [];
-        const baseUrl = provider === "hf-inference"
-            ? `https://api-inference.huggingface.co/models/${model.id}`
-            : inference_providers_js_1.HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", provider);
         return snippets.map((snippet) => {
@@ -371,3 +415,3 @@ return {
 
-API_URL = "${baseUrl}"
+API_URL = "${(0, inference_providers_js_1.openAIbaseUrl)(provider)}"
 headers = {"Authorization": ${accessToken ? `"Bearer ${accessToken}"` : `f"Bearer {API_TOKEN}"`}}
@@ -374,0 +418,0 @@

dist/commonjs/tasks/index.js

@@ -113,2 +113,3 @@ "use strict";
     "any-to-any": ["transformers"],
+    "visual-document-retrieval": ["transformers"],
 };
@@ -140,2 +141,3 @@ /**
     "document-question-answering": getData("document-question-answering", data_js_4.default),
+    "visual-document-retrieval": getData("visual-document-retrieval", data_js_16.default),
     "feature-extraction": getData("feature-extraction", data_js_5.default),
@@ -142,0 +144,0 @@ "fill-mask": getData("fill-mask", data_js_6.default),

dist/commonjs/tasks/text-generation/data.js

@@ -61,5 +61,3 @@ "use strict";
     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" },
         {
@@ -66,0 +64,0 @@ description: "Smaller variant of one of the most powerful models.",

dist/esm/inference-providers.d.ts

@@ -1,4 +0,4 @@
-export declare const INFERENCE_PROVIDERS: readonly ["hf-inference", "fal-ai", "replicate", "sambanova", "together"];
-export type InferenceProvider = (typeof INFERENCE_PROVIDERS)[number];
-export declare const HF_HUB_INFERENCE_PROXY_TEMPLATE = "https://huggingface.co/api/inference-proxy/{{PROVIDER}}";
+declare const INFERENCE_PROVIDERS: readonly ["fal-ai", "fireworks-ai", "hf-inference", "hyperbolic", "replicate", "sambanova", "together"];
+export type SnippetInferenceProvider = (typeof INFERENCE_PROVIDERS)[number] | string;
+export declare const HF_HUB_INFERENCE_PROXY_TEMPLATE = "https://router.huggingface.co/{{PROVIDER}}";
 /**
@@ -9,3 +9,4 @@ * URL to set as baseUrl in the OpenAI SDK.
  */
-export declare function openAIbaseUrl(provider: InferenceProvider): string;
+export declare function openAIbaseUrl(provider: SnippetInferenceProvider): string;
+export {};
 //# sourceMappingURL=inference-providers.d.ts.map

dist/esm/inference-providers.js

@@ -1,3 +0,13 @@
-export const INFERENCE_PROVIDERS = ["hf-inference", "fal-ai", "replicate", "sambanova", "together"];
-export const HF_HUB_INFERENCE_PROXY_TEMPLATE = `https://huggingface.co/api/inference-proxy/{{PROVIDER}}`;
+/// This list is for illustration purposes only.
+/// in the `tasks` sub-package, we do not need actual strong typing of the inference providers.
+const INFERENCE_PROVIDERS = [
+    "fal-ai",
+    "fireworks-ai",
+    "hf-inference",
+    "hyperbolic",
+    "replicate",
+    "sambanova",
+    "together",
+];
+export const HF_HUB_INFERENCE_PROXY_TEMPLATE = `https://router.huggingface.co/{{PROVIDER}}`;
 /**
@@ -9,5 +19,4 @@ * URL to set as baseUrl in the OpenAI SDK.
 export function openAIbaseUrl(provider) {
-    return provider === "hf-inference"
-        ? "https://api-inference.huggingface.co/v1/"
-        : HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", provider);
+    const url = HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", provider);
+    return provider === "hf-inference" ? `${url}/v1` : url;
 }

dist/esm/library-to-tasks.d.ts

@@ -5,3 +5,3 @@ import type { ModelLibraryKey } from "./model-libraries.js";
  * Mapping from library name to its supported tasks.
- * Inference API (serverless) should be disabled for all other (library, task) pairs beyond this mapping.
+ * HF-Inference API (serverless) should be disabled for all other (library, task) pairs beyond this mapping.
  * This mapping is partially generated automatically by "python-api-export-tasks" action in
@@ -8,0 +8,0 @@ * huggingface/api-inference-community repo upon merge. For transformers, the mapping is manually

dist/esm/library-to-tasks.js

 /**
  * Mapping from library name to its supported tasks.
- * Inference API (serverless) should be disabled for all other (library, task) pairs beyond this mapping.
+ * HF-Inference API (serverless) should be disabled for all other (library, task) pairs beyond this mapping.
  * This mapping is partially generated automatically by "python-api-export-tasks" action in
@@ -5,0 +5,0 @@ * huggingface/api-inference-community repo upon merge. For transformers, the mapping is manually

dist/esm/model-libraries-snippets.d.ts

@@ -6,2 +6,3 @@ import type { ModelData } from "./model-data.js";
 export declare const audioseal: (model: ModelData) => string[];
+export declare const ben2: (model: ModelData) => string[];
 export declare const bertopic: (model: ModelData) => string[];
@@ -8,0 +9,0 @@ export declare const bm25s: (model: ModelData) => string[];

dist/esm/model-libraries-snippets.js

@@ -69,2 +69,15 @@ import { LIBRARY_TASK_MAPPING } from "./library-to-tasks.js";
 }
+export const ben2 = (model) => [
+    `import requests
+from PIL import Image
+from ben2 import AutoModel
+
+url = "https://huggingface.co/datasets/mishig/sample_images/resolve/main/teapot.jpg"
+image = Image.open(requests.get(url, stream=True).raw)
+
+model = AutoModel.from_pretrained("${model.id}")
+model.to("cuda").eval()
+foreground = model.inference(image)
+`,
+];
 export const bertopic = (model) => [
@@ -71,0 +84,0 @@ `from bertopic import BERTopic

dist/esm/model-libraries.d.ts

@@ -108,2 +108,9 @@ import type { ModelData } from "./model-data.js";
     };
+    ben2: {
+        prettyLabel: string;
+        repoName: string;
+        repoUrl: string;
+        snippets: (model: ModelData) => string[];
+        filter: false;
+    };
     bertopic: {
@@ -898,3 +905,3 @@ prettyLabel: string;
 export declare const ALL_MODEL_LIBRARY_KEYS: ModelLibraryKey[];
-export declare const ALL_DISPLAY_MODEL_LIBRARY_KEYS: ("adapter-transformers" | "allennlp" | "anemoi" | "asteroid" | "audiocraft" | "audioseal" | "bertopic" | "big_vision" | "birder" | "birefnet" | "bm25s" | "champ" | "chat_tts" | "colpali" | "cosmos" | "cxr-foundation" | "deepforest" | "depth-anything-v2" | "depth-pro" | "derm-foundation" | "diffree" | "diffusers" | "diffusionkit" | "doctr" | "cartesia_pytorch" | "cartesia_mlx" | "clipscope" | "cosyvoice" | "cotracker" | "edsnlp" | "elm" | "espnet" | "fairseq" | "fastai" | "fasttext" | "flair" | "gemma.cpp" | "gliner" | "glyph-byt5" | "grok" | "hallo" | "hezar" | "htrflow" | "hunyuan-dit" | "imstoucan" | "keras" | "tf-keras" | "keras-hub" | "k2" | "liveportrait" | "llama-cpp-python" | "mini-omni2" | "mindspore" | "mamba-ssm" | "mars5-tts" | "mesh-anything" | "mitie" | "ml-agents" | "mlx" | "mlx-image" | "mlc-llm" | "model2vec" | "moshi" | "nemo" | "open-oasis" | "open_clip" | "paddlenlp" | "peft" | "pxia" | "pyannote-audio" | "py-feat" | "pythae" | "recurrentgemma" | "relik" | "refiners" | "reverb" | "saelens" | "sam2" | "sample-factory" | "sapiens" | "sentence-transformers" | "setfit" | "sklearn" | "spacy" | "span-marker" | "speechbrain" | "ssr-speech" | "stable-audio-tools" | "diffusion-single-file" | "seed-story" | "soloaudio" | "stable-baselines3" | "stanza" | "swarmformer" | "f5-tts" | "genmo" | "tensorflowtts" | "tabpfn" | "terratorch" | "tic-clip" | "timesfm" | "timm" | "transformers" | "transformers.js" | "trellis" | "ultralytics" | "unity-sentis" | "sana" | "vfi-mamba" | "voicecraft" | "whisperkit" | "yolov10" | "3dtopia-xl")[];
+export declare const ALL_DISPLAY_MODEL_LIBRARY_KEYS: ("adapter-transformers" | "allennlp" | "anemoi" | "asteroid" | "audiocraft" | "audioseal" | "ben2" | "bertopic" | "big_vision" | "birder" | "birefnet" | "bm25s" | "champ" | "chat_tts" | "colpali" | "cosmos" | "cxr-foundation" | "deepforest" | "depth-anything-v2" | "depth-pro" | "derm-foundation" | "diffree" | "diffusers" | "diffusionkit" | "doctr" | "cartesia_pytorch" | "cartesia_mlx" | "clipscope" | "cosyvoice" | "cotracker" | "edsnlp" | "elm" | "espnet" | "fairseq" | "fastai" | "fasttext" | "flair" | "gemma.cpp" | "gliner" | "glyph-byt5" | "grok" | "hallo" | "hezar" | "htrflow" | "hunyuan-dit" | "imstoucan" | "keras" | "tf-keras" | "keras-hub" | "k2" | "liveportrait" | "llama-cpp-python" | "mini-omni2" | "mindspore" | "mamba-ssm" | "mars5-tts" | "mesh-anything" | "mitie" | "ml-agents" | "mlx" | "mlx-image" | "mlc-llm" | "model2vec" | "moshi" | "nemo" | "open-oasis" | "open_clip" | "paddlenlp" | "peft" | "pxia" | "pyannote-audio" | "py-feat" | "pythae" | "recurrentgemma" | "relik" | "refiners" | "reverb" | "saelens" | "sam2" | "sample-factory" | "sapiens" | "sentence-transformers" | "setfit" | "sklearn" | "spacy" | "span-marker" | "speechbrain" | "ssr-speech" | "stable-audio-tools" | "diffusion-single-file" | "seed-story" | "soloaudio" | "stable-baselines3" | "stanza" | "swarmformer" | "f5-tts" | "genmo" | "tensorflowtts" | "tabpfn" | "terratorch" | "tic-clip" | "timesfm" | "timm" | "transformers" | "transformers.js" | "trellis" | "ultralytics" | "unity-sentis" | "sana" | "vfi-mamba" | "voicecraft" | "whisperkit" | "yolov10" | "3dtopia-xl")[];
 //# sourceMappingURL=model-libraries.d.ts.map

dist/esm/model-libraries.js

@@ -68,2 +68,9 @@ import * as snippets from "./model-libraries-snippets.js";
     },
+    ben2: {
+        prettyLabel: "BEN2",
+        repoName: "BEN2",
+        repoUrl: "https://github.com/PramaLLC/BEN2",
+        snippets: snippets.ben2,
+        filter: false,
+    },
     bertopic: {
@@ -70,0 +77,0 @@ prettyLabel: "BERTopic",

dist/esm/pipelines.d.ts

@@ -412,2 +412,8 @@ export declare const MODALITIES: readonly ["multimodal", "nlp", "cv", "audio", "tabular", "rl", "other"];
     };
+    "visual-document-retrieval": {
+        name: string;
+        modality: "multimodal";
+        color: "yellow";
+        hideInDatasets: true;
+    };
     "any-to-any": {
@@ -431,3 +437,3 @@ name: string;
 export declare const SUBTASK_TYPES: string[];
-export declare const PIPELINE_TYPES_SET: Set<"other" | "text-classification" | "token-classification" | "table-question-answering" | "question-answering" | "zero-shot-classification" | "translation" | "summarization" | "feature-extraction" | "text-generation" | "text2text-generation" | "fill-mask" | "sentence-similarity" | "text-to-speech" | "text-to-audio" | "automatic-speech-recognition" | "audio-to-audio" | "audio-classification" | "audio-text-to-text" | "voice-activity-detection" | "depth-estimation" | "image-classification" | "object-detection" | "image-segmentation" | "text-to-image" | "image-to-text" | "image-to-image" | "image-to-video" | "unconditional-image-generation" | "video-classification" | "reinforcement-learning" | "robotics" | "tabular-classification" | "tabular-regression" | "tabular-to-text" | "table-to-text" | "multiple-choice" | "text-retrieval" | "time-series-forecasting" | "text-to-video" | "image-text-to-text" | "visual-question-answering" | "document-question-answering" | "zero-shot-image-classification" | "graph-ml" | "mask-generation" | "zero-shot-object-detection" | "text-to-3d" | "image-to-3d" | "image-feature-extraction" | "video-text-to-text" | "keypoint-detection" | "any-to-any">;
+export declare const PIPELINE_TYPES_SET: Set<"other" | "text-classification" | "token-classification" | "table-question-answering" | "question-answering" | "zero-shot-classification" | "translation" | "summarization" | "feature-extraction" | "text-generation" | "text2text-generation" | "fill-mask" | "sentence-similarity" | "text-to-speech" | "text-to-audio" | "automatic-speech-recognition" | "audio-to-audio" | "audio-classification" | "audio-text-to-text" | "voice-activity-detection" | "depth-estimation" | "image-classification" | "object-detection" | "image-segmentation" | "text-to-image" | "image-to-text" | "image-to-image" | "image-to-video" | "unconditional-image-generation" | "video-classification" | "reinforcement-learning" | "robotics" | "tabular-classification" | "tabular-regression" | "tabular-to-text" | "table-to-text" | "multiple-choice" | "text-retrieval" | "time-series-forecasting" | "text-to-video" | "image-text-to-text" | "visual-question-answering" | "document-question-answering" | "zero-shot-image-classification" | "graph-ml" | "mask-generation" | "zero-shot-object-detection" | "text-to-3d" | "image-to-3d" | "image-feature-extraction" | "video-text-to-text" | "keypoint-detection" | "visual-document-retrieval" | "any-to-any">;
 //# sourceMappingURL=pipelines.d.ts.map

dist/esm/pipelines.js

@@ -630,2 +630,8 @@ export const MODALITIES = ["multimodal", "nlp", "cv", "audio", "tabular", "rl", "other"];
     },
+    "visual-document-retrieval": {
+        name: "Visual Document Retrieval",
+        modality: "multimodal",
+        color: "yellow",
+        hideInDatasets: true,
+    },
     "any-to-any": {
@@ -632,0 +638,0 @@ name: "Any-to-Any",

dist/esm/snippets/common.js

@@ -17,4 +17,4 @@ export function stringifyMessages(messages, opts) {
     return Object.entries(config)
-        .map(([key, val]) => `${quote}${key}${quote}${opts.attributeValueConnector}${val}`)
-        .join(`,${opts.indent}`);
+        .map(([key, val]) => `${quote}${key}${quote}${opts.attributeValueConnector}${val},`)
+        .join(`${opts.indent}`);
 }

dist/esm/snippets/curl.d.ts

@@ -1,7 +0,7 @@
-import { type InferenceProvider } from "../inference-providers.js";
+import { type SnippetInferenceProvider } from "../inference-providers.js";
 import type { PipelineType } from "../pipelines.js";
 import type { ChatCompletionInputMessage, GenerationParameters } from "../tasks/index.js";
 import type { InferenceSnippet, ModelDataMinimal } from "./types.js";
-export declare const snippetBasic: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
-export declare const snippetTextGeneration: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: {
+export declare const snippetBasic: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
+export declare const snippetTextGeneration: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: {
     streaming?: boolean;
@@ -13,6 +13,6 @@ messages?: ChatCompletionInputMessage[];
 }) => InferenceSnippet[];
-export declare const snippetZeroShotClassification: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
-export declare const snippetFile: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
-export declare const curlSnippets: Partial<Record<PipelineType, (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: Record<string, unknown>) => InferenceSnippet[]>>;
-export declare function getCurlInferenceSnippet(model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: Record<string, unknown>): InferenceSnippet[];
+export declare const snippetZeroShotClassification: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
+export declare const snippetFile: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
+export declare const curlSnippets: Partial<Record<PipelineType, (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: Record<string, unknown>) => InferenceSnippet[]>>;
+export declare function getCurlInferenceSnippet(model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: Record<string, unknown>): InferenceSnippet[];
 //# sourceMappingURL=curl.d.ts.map

dist/esm/snippets/curl.js

@@ -12,3 +12,3 @@ import { HF_HUB_INFERENCE_PROXY_TEMPLATE } from "../inference-providers.js";
             content: `\
-curl https://api-inference.huggingface.co/models/${model.id} \\
+curl https://router.huggingface.co/hf-inference/models/${model.id} \\
 	-X POST \\
@@ -24,3 +24,3 @@ -d '{"inputs": ${getModelInputSnippet(model, true)}}' \\
         const baseUrl = provider === "hf-inference"
-            ? `https://api-inference.huggingface.co/models/${model.id}/v1/chat/completions`
+            ? `https://router.huggingface.co/hf-inference/models/${model.id}/v1/chat/completions`
             : HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", provider) + "/v1/chat/completions";
@@ -53,3 +53,3 @@ // Conversational model detected, so we display a code snippet that features the Messages API
                     attributeValueConnector: ": ",
-                })},
+                })}
     "stream": ${!!streaming}
@@ -71,3 +71,3 @@ }'`,
             client: "curl",
-            content: `curl https://api-inference.huggingface.co/models/${model.id} \\
+            content: `curl https://router.huggingface.co/hf-inference/models/${model.id} \\
 	-X POST \\
@@ -87,3 +87,3 @@ -d '{"inputs": ${getModelInputSnippet(model, true)}, "parameters": {"candidate_labels": ["refund", "legal", "faq"]}}' \\
             client: "curl",
-            content: `curl https://api-inference.huggingface.co/models/${model.id} \\
+            content: `curl https://router.huggingface.co/hf-inference/models/${model.id} \\
 	-X POST \\
@@ -90,0 +90,0 @@ --data-binary '@${getModelInputSnippet(model, true, true)}' \\

dist/esm/snippets/js.d.ts

@@ -1,7 +0,7 @@
-import { type InferenceProvider } from "../inference-providers.js";
+import { type SnippetInferenceProvider } from "../inference-providers.js";
 import type { PipelineType } from "../pipelines.js";
 import type { ChatCompletionInputMessage, GenerationParameters } from "../tasks/index.js";
 import type { InferenceSnippet, ModelDataMinimal } from "./types.js";
-export declare const snippetBasic: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
-export declare const snippetTextGeneration: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: {
+export declare const snippetBasic: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
+export declare const snippetTextGeneration: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: {
     streaming?: boolean;
@@ -14,8 +14,8 @@ messages?: ChatCompletionInputMessage[];
 export declare const snippetZeroShotClassification: (model: ModelDataMinimal, accessToken: string) => InferenceSnippet[];
-export declare const snippetTextToImage: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
-export declare const snippetTextToAudio: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
-export declare const snippetAutomaticSpeechRecognition: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
-export declare const snippetFile: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
-export declare const jsSnippets: Partial<Record<PipelineType, (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: Record<string, unknown>) => InferenceSnippet[]>>;
-export declare function getJsInferenceSnippet(model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: Record<string, unknown>): InferenceSnippet[];
+export declare const snippetTextToImage: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
+export declare const snippetTextToAudio: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
+export declare const snippetAutomaticSpeechRecognition: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
+export declare const snippetFile: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
+export declare const jsSnippets: Partial<Record<PipelineType, (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: Record<string, unknown>) => InferenceSnippet[]>>;
+export declare function getJsInferenceSnippet(model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: Record<string, unknown>): InferenceSnippet[];
 //# sourceMappingURL=js.d.ts.map

dist/esm/snippets/js.js

@@ -34,3 +34,3 @@ import { openAIbaseUrl } from "../inference-providers.js";
 
-console.log(output)
+console.log(output);
 `,
@@ -45,3 +45,3 @@ },
 	const response = await fetch(
-		"https://api-inference.huggingface.co/models/${model.id}",
+		"https://router.huggingface.co/hf-inference/models/${model.id}",
 		{
@@ -121,3 +121,3 @@ headers: {
 	messages: ${messagesStr},
-	${configStr},
+	${configStr}
 	stream: true,
@@ -151,3 +151,4 @@ });
 
-console.log(chatCompletion.choices[0].message);`,
+console.log(chatCompletion.choices[0].message);
+`,
                 },
@@ -169,3 +170,4 @@ {
 
-console.log(chatCompletion.choices[0].message);`,
+console.log(chatCompletion.choices[0].message);
+`,
                 },
@@ -185,3 +187,3 @@ ];
 			const response = await fetch(
-				"https://api-inference.huggingface.co/models/${model.id}",
+				"https://router.huggingface.co/hf-inference/models/${model.id}",
 				{
@@ -230,3 +232,3 @@ headers: {
 	const response = await fetch(
-		"https://api-inference.huggingface.co/models/${model.id}",
+		"https://router.huggingface.co/hf-inference/models/${model.id}",
 		{
@@ -258,3 +260,3 @@ headers: {
 		const response = await fetch(
-			"https://api-inference.huggingface.co/models/${model.id}",
+			"https://router.huggingface.co/hf-inference/models/${model.id}",
 			{
@@ -334,3 +336,3 @@ headers: {
 	const response = await fetch(
-		"https://api-inference.huggingface.co/models/${model.id}",
+		"https://router.huggingface.co/hf-inference/models/${model.id}",
 		{
@@ -337,0 +339,0 @@ headers: {

dist/esm/snippets/python.d.ts

@@ -1,6 +0,6 @@
-import { type InferenceProvider } from "../inference-providers.js";
+import { type SnippetInferenceProvider } from "../inference-providers.js";
 import type { PipelineType } from "../pipelines.js";
 import type { ChatCompletionInputMessage, GenerationParameters } from "../tasks/index.js";
 import type { InferenceSnippet, ModelDataMinimal } from "./types.js";
-export declare const snippetConversational: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: {
+export declare const snippetConversational: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: {
     streaming?: boolean;
@@ -14,10 +14,10 @@ messages?: ChatCompletionInputMessage[];
 export declare const snippetZeroShotImageClassification: (model: ModelDataMinimal) => InferenceSnippet[];
-export declare const snippetBasic: (model: ModelDataMinimal) => InferenceSnippet[];
+export declare const snippetBasic: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
 export declare const snippetFile: (model: ModelDataMinimal) => InferenceSnippet[];
-export declare const snippetTextToImage: (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider) => InferenceSnippet[];
+export declare const snippetTextToImage: (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider) => InferenceSnippet[];
 export declare const snippetTabular: (model: ModelDataMinimal) => InferenceSnippet[];
 export declare const snippetTextToAudio: (model: ModelDataMinimal) => InferenceSnippet[];
 export declare const snippetDocumentQuestionAnswering: (model: ModelDataMinimal) => InferenceSnippet[];
-export declare const pythonSnippets: Partial<Record<PipelineType, (model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: Record<string, unknown>) => InferenceSnippet[]>>;
-export declare function getPythonInferenceSnippet(model: ModelDataMinimal, accessToken: string, provider: InferenceProvider, opts?: Record<string, unknown>): InferenceSnippet[];
+export declare const pythonSnippets: Partial<Record<PipelineType, (model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: Record<string, unknown>) => InferenceSnippet[]>>;
+export declare function getPythonInferenceSnippet(model: ModelDataMinimal, accessToken: string, provider: SnippetInferenceProvider, opts?: Record<string, unknown>): InferenceSnippet[];
 //# sourceMappingURL=python.d.ts.map

dist/esm/snippets/python.js

@@ -1,4 +0,33 @@
-import { HF_HUB_INFERENCE_PROXY_TEMPLATE, openAIbaseUrl } from "../inference-providers.js";
+import { openAIbaseUrl, } from "../inference-providers.js";
 import { stringifyGenerationConfig, stringifyMessages } from "./common.js";
 import { getModelInputSnippet } from "./inputs.js";
+const HFH_INFERENCE_CLIENT_METHODS = {
+    "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, provider) => `\
@@ -37,3 +66,3 @@ from huggingface_hub import InferenceClient
 	messages=messages, 
-	${configStr},
+	${configStr}
 	stream=True
@@ -60,3 +89,3 @@ )
 	messages=messages, 
-	${configStr},
+	${configStr}
 	stream=True
@@ -148,4 +177,22 @@ )
 };
-export const snippetBasic = (model) => {
+export const snippetBasic = (model, accessToken, provider) => {
     return [
+        ...(model.pipeline_tag && model.pipeline_tag in HFH_INFERENCE_CLIENT_METHODS
+            ? [
+                {
+                    client: "huggingface_hub",
+                    content: `\
+${snippetImportInferenceClient(accessToken, provider)}
+
+result = client.${HFH_INFERENCE_CLIENT_METHODS[model.pipeline_tag]}(
+	model="${model.id}",
+	inputs=${getModelInputSnippet(model)},
+	provider="${provider}",
+)
+
+print(result)
+`,
+                },
+            ]
+            : []),
         {
@@ -346,5 +393,2 @@ client: "requests",
             : [];
-        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) => {
@@ -357,3 +401,3 @@ return {
 
-API_URL = "${baseUrl}"
+API_URL = "${openAIbaseUrl(provider)}"
 headers = {"Authorization": ${accessToken ? `"Bearer ${accessToken}"` : `f"Bearer {API_TOKEN}"`}}
@@ -360,0 +404,0 @@

dist/esm/tasks/index.js

@@ -107,2 +107,3 @@ import { PIPELINE_DATA } from "../pipelines.js";
     "any-to-any": ["transformers"],
+    "visual-document-retrieval": ["transformers"],
 };
@@ -134,2 +135,3 @@ /**
     "document-question-answering": getData("document-question-answering", documentQuestionAnswering),
+    "visual-document-retrieval": getData("visual-document-retrieval", placeholder),
     "feature-extraction": getData("feature-extraction", featureExtraction),
@@ -136,0 +138,0 @@ "fill-mask": getData("fill-mask", fillMask),

dist/esm/tasks/text-generation/data.js

@@ -59,5 +59,3 @@ const taskData = {
     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" },
         {
@@ -64,0 +62,0 @@ description: "Smaller variant of one of the most powerful models.",

package.json

 {
   "name": "@huggingface/tasks",
   "packageManager": "pnpm@8.10.5",
-  "version": "0.15.8",
+  "version": "0.15.9",
   "description": "List of ML tasks for huggingface.co/tasks",
@@ -6,0 +6,0 @@ "repository": "https://github.com/huggingface/huggingface.js.git",

src/inference-providers.ts

@@ -1,6 +0,16 @@
-export const INFERENCE_PROVIDERS = ["hf-inference", "fal-ai", "replicate", "sambanova", "together"] as const;
+/// This list is for illustration purposes only.
+/// in the `tasks` sub-package, we do not need actual strong typing of the inference providers.
+const INFERENCE_PROVIDERS = [
+	"fal-ai",
+	"fireworks-ai",
+	"hf-inference",
+	"hyperbolic",
+	"replicate",
+	"sambanova",
+	"together",
+] as const;
 
-export type InferenceProvider = (typeof INFERENCE_PROVIDERS)[number];
+export type SnippetInferenceProvider = (typeof INFERENCE_PROVIDERS)[number] | string;
 
-export const HF_HUB_INFERENCE_PROXY_TEMPLATE = `https://huggingface.co/api/inference-proxy/{{PROVIDER}}`;
+export const HF_HUB_INFERENCE_PROXY_TEMPLATE = `https://router.huggingface.co/{{PROVIDER}}`;
 
@@ -12,6 +22,5 @@ /**
  */
-export function openAIbaseUrl(provider: InferenceProvider): string {
-	return provider === "hf-inference"
-		? "https://api-inference.huggingface.co/v1/"
-		: HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", provider);
+export function openAIbaseUrl(provider: SnippetInferenceProvider): string {
+	const url = HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", provider);
+	return provider === "hf-inference" ? `${url}/v1` : url;
 }

src/library-to-tasks.ts

@@ -6,3 +6,3 @@ import type { ModelLibraryKey } from "./model-libraries.js";
  * Mapping from library name to its supported tasks.
- * Inference API (serverless) should be disabled for all other (library, task) pairs beyond this mapping.
+ * HF-Inference API (serverless) should be disabled for all other (library, task) pairs beyond this mapping.
  * This mapping is partially generated automatically by "python-api-export-tasks" action in
@@ -9,0 +9,0 @@ * huggingface/api-inference-community repo upon merge. For transformers, the mapping is manually

src/model-libraries-snippets.ts

@@ -86,2 +86,16 @@ import type { ModelData } from "./model-data.js";
 
+export const ben2 = (model: ModelData): string[] => [
+	`import requests
+from PIL import Image
+from ben2 import AutoModel
+
+url = "https://huggingface.co/datasets/mishig/sample_images/resolve/main/teapot.jpg"
+image = Image.open(requests.get(url, stream=True).raw)
+
+model = AutoModel.from_pretrained("${model.id}")
+model.to("cuda").eval()
+foreground = model.inference(image)
+`,
+];
+
 export const bertopic = (model: ModelData): string[] => [
@@ -88,0 +102,0 @@ `from bertopic import BERTopic

src/model-libraries.ts

@@ -112,2 +112,9 @@ import * as snippets from "./model-libraries-snippets.js";
 	},
+	ben2: {
+		prettyLabel: "BEN2",
+		repoName: "BEN2",
+		repoUrl: "https://github.com/PramaLLC/BEN2",
+		snippets: snippets.ben2,
+		filter: false,
+	},
 	bertopic: {
@@ -114,0 +121,0 @@ prettyLabel: "BERTopic",

src/pipelines.ts

@@ -679,2 +679,8 @@ export const MODALITIES = ["multimodal", "nlp", "cv", "audio", "tabular", "rl", "other"] as const;
 	},
+	"visual-document-retrieval": {
+		name: "Visual Document Retrieval",
+		modality: "multimodal",
+		color: "yellow",
+		hideInDatasets: true,
+	},
 	"any-to-any": {
@@ -681,0 +687,0 @@ name: "Any-to-Any",

src/snippets/common.ts

@@ -37,4 +37,4 @@ import type { ChatCompletionInputMessage, GenerationParameters } from "../tasks/index.js";
 	return Object.entries(config)
-		.map(([key, val]) => `${quote}${key}${quote}${opts.attributeValueConnector}${val}`)
-		.join(`,${opts.indent}`);
+		.map(([key, val]) => `${quote}${key}${quote}${opts.attributeValueConnector}${val},`)
+		.join(`${opts.indent}`);
 }

src/snippets/curl.ts

@@ -1,2 +0,2 @@
-import { HF_HUB_INFERENCE_PROXY_TEMPLATE, type InferenceProvider } from "../inference-providers.js";
+import { HF_HUB_INFERENCE_PROXY_TEMPLATE, type SnippetInferenceProvider } from "../inference-providers.js";
 import type { PipelineType } from "../pipelines.js";
@@ -11,3 +11,3 @@ import type { ChatCompletionInputMessage, GenerationParameters } from "../tasks/index.js";
 	accessToken: string,
-	provider: InferenceProvider
+	provider: SnippetInferenceProvider
 ): InferenceSnippet[] => {
@@ -21,3 +21,3 @@ if (provider !== "hf-inference") {
 			content: `\
-curl https://api-inference.huggingface.co/models/${model.id} \\
+curl https://router.huggingface.co/hf-inference/models/${model.id} \\
 	-X POST \\
@@ -34,3 +34,3 @@ -d '{"inputs": ${getModelInputSnippet(model, true)}}' \\
 	accessToken: string,
-	provider: InferenceProvider,
+	provider: SnippetInferenceProvider,
 	opts?: {
@@ -47,3 +47,3 @@ streaming?: boolean;
 			provider === "hf-inference"
-				? `https://api-inference.huggingface.co/models/${model.id}/v1/chat/completions`
+				? `https://router.huggingface.co/hf-inference/models/${model.id}/v1/chat/completions`
 				: HF_HUB_INFERENCE_PROXY_TEMPLATE.replace("{{PROVIDER}}", provider) + "/v1/chat/completions";
@@ -70,11 +70,11 @@
     "messages": ${stringifyMessages(messages, {
-			indent: "\t",
-			attributeKeyQuotes: true,
-			customContentEscaper: (str) => str.replace(/'/g, "'\\''"),
-		})},
+					indent: "\t",
+					attributeKeyQuotes: true,
+					customContentEscaper: (str) => str.replace(/'/g, "'\\''"),
+				})},
     ${stringifyGenerationConfig(config, {
-			indent: "\n    ",
-			attributeKeyQuotes: true,
-			attributeValueConnector: ": ",
-		})},
+					indent: "\n    ",
+					attributeKeyQuotes: true,
+					attributeValueConnector: ": ",
+				})}
     "stream": ${!!streaming}
@@ -92,3 +92,3 @@ }'`,
 	accessToken: string,
-	provider: InferenceProvider
+	provider: SnippetInferenceProvider
 ): InferenceSnippet[] => {
@@ -101,3 +101,3 @@ if (provider !== "hf-inference") {
 			client: "curl",
-			content: `curl https://api-inference.huggingface.co/models/${model.id} \\
+			content: `curl https://router.huggingface.co/hf-inference/models/${model.id} \\
 	-X POST \\
@@ -114,3 +114,3 @@ -d '{"inputs": ${getModelInputSnippet(model, true)}, "parameters": {"candidate_labels": ["refund", "legal", "faq"]}}' \\
 	accessToken: string,
-	provider: InferenceProvider
+	provider: SnippetInferenceProvider
 ): InferenceSnippet[] => {
@@ -123,3 +123,3 @@ if (provider !== "hf-inference") {
 			client: "curl",
-			content: `curl https://api-inference.huggingface.co/models/${model.id} \\
+			content: `curl https://router.huggingface.co/hf-inference/models/${model.id} \\
 	-X POST \\
@@ -138,3 +138,3 @@ --data-binary '@${getModelInputSnippet(model, true, true)}' \\
 			accessToken: string,
-			provider: InferenceProvider,
+			provider: SnippetInferenceProvider,
 			opts?: Record<string, unknown>
@@ -173,3 +173,3 @@ ) => InferenceSnippet[]
 	accessToken: string,
-	provider: InferenceProvider,
+	provider: SnippetInferenceProvider,
 	opts?: Record<string, unknown>
@@ -176,0 +176,0 @@ ): InferenceSnippet[] {

src/snippets/js.ts

@@ -1,3 +0,3 @@
-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";
@@ -8,3 +8,3 @@ import { stringifyGenerationConfig, stringifyMessages } from "./common.js";
 
-const HFJS_METHODS: Record<string, string> = {
+const HFJS_METHODS: Partial<Record<WidgetType, string>> = {
 	"text-classification": "textClassification",
@@ -26,3 +26,3 @@ "token-classification": "tokenClassification",
 	accessToken: string,
-	provider: InferenceProvider
+	provider: SnippetInferenceProvider
 ): InferenceSnippet[] => {
@@ -45,3 +45,3 @@ return [
 
-console.log(output)
+console.log(output);
 `,
@@ -56,3 +56,3 @@ },
 	const response = await fetch(
-		"https://api-inference.huggingface.co/models/${model.id}",
+		"https://router.huggingface.co/hf-inference/models/${model.id}",
 		{
@@ -81,3 +81,3 @@ headers: {
 	accessToken: string,
-	provider: InferenceProvider,
+	provider: SnippetInferenceProvider,
 	opts?: {
@@ -147,3 +147,3 @@ streaming?: boolean;
 	messages: ${messagesStr},
-	${configStr},
+	${configStr}
 	stream: true,
@@ -176,3 +176,4 @@ });
 
-console.log(chatCompletion.choices[0].message);`,
+console.log(chatCompletion.choices[0].message);
+`,
 				},
@@ -194,3 +195,4 @@ {
 
-console.log(chatCompletion.choices[0].message);`,
+console.log(chatCompletion.choices[0].message);
+`,
 				},
@@ -210,3 +212,3 @@ ];
 			const response = await fetch(
-				"https://api-inference.huggingface.co/models/${model.id}",
+				"https://router.huggingface.co/hf-inference/models/${model.id}",
 				{
@@ -237,3 +239,3 @@ headers: {
 	accessToken: string,
-	provider: InferenceProvider
+	provider: SnippetInferenceProvider
 ): InferenceSnippet[] => {
@@ -263,3 +265,3 @@ return [
 	const response = await fetch(
-		"https://api-inference.huggingface.co/models/${model.id}",
+		"https://router.huggingface.co/hf-inference/models/${model.id}",
 		{
@@ -289,3 +291,3 @@ headers: {
 	accessToken: string,
-	provider: InferenceProvider
+	provider: SnippetInferenceProvider
 ): InferenceSnippet[] => {
@@ -297,3 +299,3 @@ if (provider !== "hf-inference") {
 		const response = await fetch(
-			"https://api-inference.huggingface.co/models/${model.id}",
+			"https://router.huggingface.co/hf-inference/models/${model.id}",
 			{
@@ -345,3 +347,3 @@ headers: {
 	accessToken: string,
-	provider: InferenceProvider
+	provider: SnippetInferenceProvider
 ): InferenceSnippet[] => {
@@ -374,3 +376,3 @@ return [
 	accessToken: string,
-	provider: InferenceProvider
+	provider: SnippetInferenceProvider
 ): InferenceSnippet[] => {
@@ -386,3 +388,3 @@ if (provider !== "hf-inference") {
 	const response = await fetch(
-		"https://api-inference.huggingface.co/models/${model.id}",
+		"https://router.huggingface.co/hf-inference/models/${model.id}",
 		{
@@ -414,3 +416,3 @@ headers: {
 			accessToken: string,
-			provider: InferenceProvider,
+			provider: SnippetInferenceProvider,
 			opts?: Record<string, unknown>
@@ -449,3 +451,3 @@ ) => InferenceSnippet[]
 	accessToken: string,
-	provider: InferenceProvider,
+	provider: SnippetInferenceProvider,
 	opts?: Record<string, unknown>
@@ -452,0 +454,0 @@ ): InferenceSnippet[] {

src/snippets/python.ts

@@ -1,3 +0,7 @@
-import { HF_HUB_INFERENCE_PROXY_TEMPLATE, openAIbaseUrl, type InferenceProvider } from "../inference-providers.js";
-import type { PipelineType } from "../pipelines.js";
+import {
+	HF_HUB_INFERENCE_PROXY_TEMPLATE,
+	openAIbaseUrl,
+	type SnippetInferenceProvider,
+} from "../inference-providers.js";
+import type { PipelineType, WidgetType } from "../pipelines.js";
 import type { ChatCompletionInputMessage, GenerationParameters } from "../tasks/index.js";
@@ -8,3 +12,33 @@ import { stringifyGenerationConfig, stringifyMessages } from "./common.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 =>
 	`\
@@ -21,3 +55,3 @@ from huggingface_hub import InferenceClient
 	accessToken: string,
-	provider: InferenceProvider,
+	provider: SnippetInferenceProvider,
 	opts?: {
@@ -58,3 +92,3 @@ streaming?: boolean;
 	messages=messages, 
-	${configStr},
+	${configStr}
 	stream=True
@@ -81,3 +115,3 @@ )
 	messages=messages, 
-	${configStr},
+	${configStr}
 	stream=True
@@ -171,4 +205,26 @@ )
 
-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)
+`,
+					},
+			  ]
+			: []),
 		{
@@ -207,3 +263,3 @@ client: "requests",
 	accessToken: string,
-	provider: InferenceProvider
+	provider: SnippetInferenceProvider
 ): InferenceSnippet[] => {
@@ -346,3 +402,3 @@ return [
 			accessToken: string,
-			provider: InferenceProvider,
+			provider: SnippetInferenceProvider,
 			opts?: Record<string, unknown>
@@ -385,3 +441,3 @@ ) => InferenceSnippet[]
 	accessToken: string,
-	provider: InferenceProvider,
+	provider: SnippetInferenceProvider,
 	opts?: Record<string, unknown>
@@ -398,7 +454,2 @@ ): InferenceSnippet[] {
 
-		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) => {
@@ -412,3 +463,3 @@ return {
 
-API_URL = "${baseUrl}"
+API_URL = "${openAIbaseUrl(provider)}"
 headers = {"Authorization": ${accessToken ? `"Bearer ${accessToken}"` : `f"Bearer {API_TOKEN}"`}}
@@ -415,0 +466,0 @@

src/tasks/audio-classification/about.md

@@ -36,3 +36,3 @@ ## Use Cases
 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"
 
@@ -39,0 +39,0 @@ def query(filename):

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

@@ -22,3 +22,3 @@ ## Use Cases
 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"
 
@@ -25,0 +25,0 @@ def query(filename):

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

@@ -28,3 +28,3 @@ ## Use Cases
 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"
 
@@ -31,0 +31,0 @@ def query(filename):

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

@@ -11,7 +11,14 @@ {
 			"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"
+					}
+				}
+			]
 		},
@@ -18,0 +25,0 @@ "normalize": {

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

@@ -79,3 +79,3 @@ ## Different Types of Vision Language Models
 ```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 \
@@ -82,0 +82,0 @@ -d '{"messages": [{"role": "user","content": [{"type": "image"}, {"type": "text", "text": "Can you describe the image?"}]}]}' \

src/tasks/index.ts

@@ -177,2 +177,3 @@ import type { PipelineType } from "../pipelines.js";
 	"any-to-any": ["transformers"],
+	"visual-document-retrieval": ["transformers"],
 };
@@ -206,2 +207,3 @@
 	"document-question-answering": getData("document-question-answering", documentQuestionAnswering),
+	"visual-document-retrieval": getData("visual-document-retrieval", placeholder),
 	"feature-extraction": getData("feature-extraction", featureExtraction),
@@ -208,0 +210,0 @@ "fill-mask": getData("fill-mask", fillMask),

src/tasks/sentence-similarity/about.md

@@ -25,3 +25,3 @@ ## Use Cases 🔍
 
-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}"}
@@ -55,3 +55,3 @@
 
-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}"}
@@ -58,0 +58,0 @@

src/tasks/text-generation/data.ts

@@ -64,5 +64,3 @@ import type { TaskDataCustom } from "../index.js";
 	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" },
 		{
@@ -69,0 +67,0 @@ description: "Smaller variant of one of the most powerful models.",

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

@@ -22,3 +22,3 @@ ## Use Cases
 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"
 
@@ -25,0 +25,0 @@ def query(payload):

No alert changes

Improved metrics

Total package byte prevSize

1777289

increased by0.56%

Lines of code

44181

increased by0.63%

No dependency changes