Hugging Face API — Model Metadata Export

Hugging Face API for model metadata. Search public models and export clean records through the Apify API, schedules, webhooks, or one-click dataset downloads. This Actor uses the official Hub API and never mirrors model files, private repositories, or gated files.

Actor ID huggingface-governance-intelligence is a stable legacy identifier. The product scope is public model metadata search and export only, not governance scoring or compliance certification.

Why use this Actor

• Model ID, title, task, library, and Hugging Face URL
• License tags and public or gated status
• Downloads, likes, repository SHA, and update dates
• One normalized row per matching public model
• JSON, CSV, Excel, XML, RSS, and API access through Apify datasets

Use it for model catalog research, license reviews, public-model inventories, and scheduled metadata snapshots. It is a search-and-export API, not a model inference service or a persistent change-history database.

Quick start

1. Add one or more public model-search terms.
2. Choose a per-query result limit from 1 to 100.
3. Run the Actor and download the dataset or read it through the Apify API.

{
  "queries": ["text-generation"],
  "maxItems": 5
}

The query is sent to the official Hugging Face model search endpoint. Add up to 20 queries and request 1–100 records per query.

Example output

{
  "recordType": "result",
  "query": "text-generation",
  "sourceId": "daotthuanSE/cv_anytext_text_generation_editing",
  "title": "daotthuanSE/cv_anytext_text_generation_editing",
  "status": "public",
  "url": "https://huggingface.co/daotthuanSE/cv_anytext_text_generation_editing",
  "updatedAt": "2026-07-08T10:44:55.000Z",
  "metrics": {
    "downloads": 0,
    "likes": 0
  },
  "metadata": {
    "pipelineTag": "text-generation",
    "licenses": ["apache-2.0"],
    "gated": false
  },
  "sourceName": "Hugging Face Hub"
}

Fields can be absent when the source record does not publish them. The Actor preserves that uncertainty instead of inventing values.

Input

Field	Type	Required	Description
queries	string[]	Yes	Model names, tasks, authors, or other Hugging Face model-search terms. Up to 20.
maxItems	integer	No	Results stored per query, from 1 to 100. Default: 25.
canaryMode	boolean	No	Returns one no-network release-status row for deterministic platform QA.

Output data contract

Each row uses the stable fields recordType, query, sourceId, title, description, status, url, publishedAt, updatedAt, metrics, metadata, sourceName, sourceUrl, retrievedAt, and releaseState.

The OUTPUT record in the default key-value store reports the item count, empty-query count, network use, release state, and any mandatory notice.

Automate it

• Run scheduled model metadata snapshots with Apify Schedules
• Trigger downstream work with webhooks
• Read the dataset from the Apify API
• Export to JSON, CSV, Excel, or another supported dataset format

For a buyer-focused guide and integration examples, see the Hugging Face Model Metadata API page.

Limitations and responsible use

• Source: official Hugging Face Hub model API
• API documentation: Hugging Face Hub API
• Rate limits: Hugging Face Hub rate limits
• Terms: Hugging Face Terms of Service
• Independent product; not affiliated with, endorsed by, or sponsored by Hugging Face
• Public metadata only
• No model weights, binaries, model-card mirroring, or private repositories
• Gated model metadata can be returned, but the Actor cannot access or download gated files
• This is not a Hugging Face scraper for repository files or model artifacts
• HTTPS source hosts are allowlisted; redirects are refused
• Responses, retries, and timeouts are bounded
• Empty searches produce an explicit emptyResult row and are not charged

API comparison

Option	Best for	Important boundary
Official Hugging Face API	Direct source integration and custom clients	You own normalization, storage, retries, and downstream exports
This Actor	Bounded model searches with normalized Apify datasets and automation	Snapshot export only; it does not store a historical diff database

Validation evidence

On 2026-07-23, the bounded source canary returned 100 complete public model metadata records, including 95 declared license tags and 12 models marked gated by the source. On 2026-07-24, actor-specific live QA returned five non-empty records from the deployed build.

Pricing

The primary pay-per-event charge is result-item.

• List price: $0.002 per successfully stored source record
• Apify plan tiers: $0.0018, $0.0016, or $0.0014 per record
• Canary, empty-result, and failed rows are not charged
• Maximum valid input: 20 queries × 100 records = 2,000 source records, or $4.00 maximum list-price PPE charge per run
• Duplicate query strings are removed before requests; if the same source record matches two different queries, each stored row is charged

Apify platform compute charges and plan credits are shown separately by Apify.

Support

For a reproducible issue, include the public query, expected behavior, run ID, and the relevant source URL. Never send private model tokens or gated repository credentials.

FAQ

Does this run Hugging Face models?

No. It searches public model metadata. It does not perform inference.

Does it download model files?

No. It returns metadata and source URLs only.

Can it monitor changes?

You can schedule repeated snapshots and compare datasets downstream. The Actor itself does not maintain a historical diff database.