Hugging Face Scraper — Models, Datasets & Spaces

Hugging Face Scraper is an Apify Actor for structured Hugging Face Hub lookups. Paste a Hugging Face URL or a text query, and the Actor auto-detects the target type and returns one record per resolved entity. For a single model, dataset, space, or paper URL, that means one deep record with fields such as downloads, likes, license, tags, language, lastModified, README text, lineage, evaluation results, or dataset config details. For user, org, and collection targets, it returns a bounded list of portfolio or collection records. The Actor is usable through Apify MCP.

Best fit and connected workflows

This Actor fits workflows where Hugging Face data needs to stay structured and machine-readable.

Use it when you want to:

• inspect a specific model, dataset, space, paper, user, organization, or collection from its canonical URL,
• search by text such as a model name, task, or library and get ranked results,
• review a Hugging Face user or organization together with portfolio counts and listed entities,
• open a collection and retrieve the collection metadata plus its items,
• extract paper metadata including authors, ArXiv ID, publication date, upvotes, and comment count.

Practical scenario

A machine learning engineer receives a link to https://huggingface.co/meta-llama/Llama-3.1-8B and wants to decide whether to add it to an internal registry. They run this Actor with that URL as the target. The returned record includes the model ID, author namespace, downloads, likes, pipeline tag, library name, license, base-model lineage, dataset citations, and model-index evaluation data. With those fields, they can verify the model identity, review its provenance, and continue with their registry import or review step.

Input

Field	Type	Purpose
targets	array of strings	Hugging Face URLs or plain-text queries. Each item is auto-detected as a model, dataset, space, paper, user, org, collection, or search target.
entityType	string	Entity type for text queries: models, datasets, spaces, papers, or all. Ignored for URLs.
resultsPerTarget	integer	Maximum records per target. Single entity URLs return one deep record. Search, user, org, and collection targets return up to this cap.
sortBy	string	Sort order for queries and portfolio listings: downloads, likes, modified, or trending.
filterTask	string	Filters model search queries by pipeline tag such as text-generation.
filterLibrary	string	Filters model search queries by library name such as transformers.
filterLanguage	string	Filters model and dataset queries by language code such as en or multilingual.
includeReadme	boolean	When true, includes full Markdown README text for models, datasets, and spaces.
responseFormat	string	concise returns a compact record shape. detailed returns the full parsed record.
proxyConfiguration	object	Apify proxy settings. The default Apify datacenter pool works for this Actor.

Focused JSON input example

{
  "targets": [
    "https://huggingface.co/meta-llama/Llama-3.1-8B",
    "qwen 3 instruct"
  ],
  "entityType": "models",
  "resultsPerTarget": 10,
  "sortBy": "downloads",
  "filterTask": "text-generation",
  "filterLibrary": "transformers",
  "filterLanguage": "en",
  "includeReadme": true,
  "responseFormat": "detailed",
  "proxyConfiguration": {
    "useApifyProxy": true
  }
}

Output

Each dataset item includes an itemType discriminator so consumers can branch by entity shape.

Field	Type	Meaning
itemType	string	Entity type such as model, dataset, space, paper, user, org, collection, collection_item, or search_result.
id	string or null	Canonical Hugging Face ID.
url	string	Canonical Hugging Face URL.
author	string or null	Namespace or author.
downloads	integer or null	Download count for models and datasets.
likes	integer or null	Community likes count.
trendingScore	number or null	Hugging Face trending score.
pipelineTag	string or null	Model task tag.
libraryName	string or null	ML library name.
license	string or null	License identifier.
tags	array of strings	Hugging Face tags.
language	array of strings	Language codes.
lastModified	string or null	ISO 8601 timestamp.
createdAt	string or null	ISO 8601 creation timestamp.
private	boolean or null	Visibility flag.
gated	string, boolean, or null	Gating status.
disabled	boolean or null	Disabled flag.
siblings	array of objects	Repo files for models and datasets.
modelIndex	array or null	Parsed evaluation results from model cards.
baseModels	array of strings	Parent or base model IDs.
adapterChildren	array of strings	Known adapter or finetune descendants.
quantizedChildren	array of strings	Known quantized descendants.
datasetsUsed	array of strings	Dataset IDs cited in model card metadata.
taskCategories	array of strings	Dataset task categories.
sizeCategories	array of strings	Dataset size buckets.
configs	array of objects	Dataset configs and splits with row counts.
sdk	string or null	Space SDK.
hardware	string or null	Space runtime hardware.
runtimeStage	string or null	Space runtime stage.
title	string or null	Paper title or display name for users, orgs, and collections.
summary	string or null	Paper abstract or collection description.
authors	array of objects	Paper authors with HF user links when available.
arxivId	string or null	ArXiv ID.
publishedAt	string or null	Paper publication date.
upvotes	integer or null	Paper upvotes.
commentsCount	integer or null	Paper discussion count.
fullname	string or null	User display fullname.
avatarUrl	string or null	Avatar or logo URL.
followers	integer or null	Follower count.
following	integer or null	Following count.
isPro	boolean or null	PRO subscriber flag.
orgs	array of strings	Organizations the user belongs to.
modelsCount	integer or null	Number of models in a portfolio.
datasetsCount	integer or null	Number of datasets in a portfolio.
spacesCount	integer or null	Number of spaces in a portfolio.
collectionSlug	string or null	Collection slug.
items	array of objects	Collection items.
readme	string or null	Full Markdown README when included.
cardData	object or null	Parsed YAML front matter.
_warnings	array of strings	Non-fatal extraction notes.

Illustrative JSON output record

{
  "itemType": "model",
  "id": "meta-llama/Llama-3.1-8B",
  "url": "https://huggingface.co/meta-llama/Llama-3.1-8B",
  "author": "meta-llama",
  "downloads": 12345678,
  "likes": 4321,
  "pipelineTag": "text-generation",
  "libraryName": "transformers",
  "license": "llama3.1",
  "tags": ["llama-3", "text-generation", "facebook"],
  "language": ["en"],
  "lastModified": "2026-04-22T18:51:00.000Z",
  "private": false,
  "gated": false,
  "disabled": false,
  "siblings": [{ "rfilename": "config.json", "size": 1234, "lfs": null }],
  "modelIndex": [{ "name": "...", "results": [] }],
  "baseModels": [],
  "adapterChildren": ["someone/llama-3.1-8b-lora-medical"],
  "quantizedChildren": ["bartowski/Meta-Llama-3.1-8B-GGUF"],
  "datasetsUsed": ["allenai/c4", "EleutherAI/pile"],
  "readme": "# Model Card for Llama 3.1 8B..."
}

How it works

The Actor uses Hugging Face public API endpoints and raw README fetches to build structured records. It auto-detects entity type from each target path when a URL is provided. For text queries, it searches the entity type you select and can return models, datasets, spaces, papers, or the combined model, dataset, and space union when entityType is set to all. For model records, it also parses lineage and evaluation metadata from the card. For datasets, it returns configs, splits, and row counts. For spaces, it captures SDK, hardware, and runtime stage. For users, orgs, and collections, it returns portfolio or item lists with the same record structure.

Pricing

This Actor uses Apify's Pay per event pricing model plus standard Apify platform usage. The primary billed event is each Result record pushed to the dataset. Actor start is also billed as a one-time event based on the Actor's memory allocation.

For example, if a run returns ten records, the execution includes ten Result events plus one Actor Start event.

For the current live pricing details, open the Pricing tab in the Apify Console.

Use with AI agents (MCP)

This Actor is available through Apify MCP as a structured Hugging Face data tool. The exact Actor identity is khadinakbar/huggingface-all-in-one-scraper.

Tool description: fetch canonical Hugging Face metadata, portfolio records, collection items, and deep card fields for URLs or search queries, with compact or detailed record shapes.

Fetch the Hugging Face model card for https://huggingface.co/meta-llama/Llama-3.1-8B, include the full README, and return a concise record with downloads, likes, license, pipeline tag, lineage, and evaluation fields.

Output interpretation:

• itemType tells the agent which schema branch to read.
• id and url identify the entity.
• downloads, likes, and lastModified support ranking and freshness checks.
• modelIndex, baseModels, adapterChildren, quantizedChildren, and datasetsUsed are the model provenance fields.
• configs and taskCategories are the dataset-specific fields.
• sdk, hardware, and runtimeStage are the Space-specific fields.
• readme contains the source card text when requested.
• _warnings can surface partial extraction notes.

Pagination and cost guidance:

• Single entity URLs return one deep record.
• Search, user, org, and collection targets can return multiple records up to resultsPerTarget.
• responseFormat: "concise" keeps records smaller for agent loops.
• includeReadme: false reduces per-record fetch work for bulk search tasks.

Use from code

JavaScript

const { ApifyClient } = require('apify-client');

const client = new ApifyClient({
  token: process.env.APIFY_TOKEN
});

const run = await client.actor('khadinakbar/huggingface-all-in-one-scraper').call({
  targets: [
    'https://huggingface.co/meta-llama/Llama-3.1-8B',
    'qwen 3 instruct'
  ],
  entityType: 'models',
  resultsPerTarget: 5,
  responseFormat: 'concise'
});

const items = await client.dataset(run.defaultDatasetId).listItems();
console.log(items.items);

Python

from apify_client import ApifyClient

client = ApifyClient(token=os.environ["APIFY_TOKEN"])

run = client.actor("khadinakbar/huggingface-all-in-one-scraper").call(
    run_input={
        "targets": [
            "https://huggingface.co/datasets/squad",
            "image classification"
        ],
        "entityType": "all",
        "resultsPerTarget": 50,
    }
)

dataset_items = list(client.dataset(run["defaultDatasetId"]).iterate_items())
print(dataset_items)

Best results and outcome guidance

Use URLs when you already know the exact Hugging Face entity and want one deep record. Use text queries when you want discovery across models, datasets, spaces, or papers. Set entityType to match the query intent, and use all when you want the combined model, dataset, and space search union. For model searches, filterTask, filterLibrary, and filterLanguage help narrow the result set. For bulk runs, responseFormat: "concise" keeps the items compact for downstream processing.

Focused standalone workflow

This Actor is designed as a focused standalone workflow.

Design note

I found that the dataset schema treats gated defensively as string | boolean | null, which matches the live contract and explains why the output parser normalizes that field carefully.

FAQ

When should I use a model URL versus a text query?
Use a model URL when you already know the exact repository and want one deep record. Use a text query when you want search results or want to compare candidates.

How does this Actor handle Hugging Face users and organizations?
It returns profile metadata plus portfolio listings. For those targets, sortBy affects the returned portfolio order.

What comes back for a collection target?
You get the collection metadata and the items inside it. Each item is represented in the same structured dataset format.

Can I request only models, datasets, or spaces from the dataset output?
Yes. The dataset output includes filtered views for models, datasets, and spaces.

Can I use this through Apify MCP?
Yes. It is an Apify Actor usable through Apify MCP, and the MCP-facing response shape is designed to stay compact when responseFormat is set to concise.

Responsible use

Use the Actor in line with Hugging Face's public access terms and the licenses attached to the entities you retrieve. The Actor works with public URLs and queries exposed by Hugging Face. For higher-volume usage, keep query scopes narrow and set resultsPerTarget thoughtfully so the returned dataset stays aligned with your downstream task.