Open Dataset Finder: Data.gov, Socrata & Hugging Face Search

Open Dataset Finder: Data.gov, Socrata & Hugging Face Search

Open Dataset Finder searches the big open-data catalogs at once (Data.gov, hundreds of Socrata city, state and agency portals, and Hugging Face), then dedupes and ranks the results so you get the few datasets that actually fit, with license, formats, last update, popularity, column names, API endpoint and download links on every row. An optional add-on shows 3 real sample records so you can check the data before you download anything.

Built for data scientists and ML engineers, analysts and BI teams, journalists, students and researchers, civic-tech builders, and AI agents that need data on demand.

• One query, three catalogs, one ranked list (relevance = title, keyword and description hits + freshness + popularity).
• Filters: formats (csv, json, geojson, parquet, api...), open licenses only, updated within N days, a single Socrata portal.
• Column names for Socrata datasets; ready-made API and CSV download URLs.
• Add-on (off by default): 3 sample rows per Socrata dataset (the portal's robots.txt is checked first).

No login, no API key, no proxies: each catalog's documented public search API, called with an identifying User-Agent. Data.gov's 10-second crawl delay is respected.

Use it as a dataset search engine API, a Data.gov scraper / CKAN search alternative, a Socrata open data portal search across every city and state, or a Hugging Face dataset finder for machine learning and LLM training data: find public datasets by topic, license and format (CSV, JSON, GeoJSON, Parquet, API) and export the list to CSV, Excel or JSON.

Why this one

	Open Dataset Finder	Single-catalog scraper
Catalogs	Data.gov + Socrata network + Hugging Face	One
Ranking + dedupe	Yes	Raw search order, duplicates
License filter	Recognised open licenses	No
Usable links	API endpoint + CSV download + columns	Landing page only
Preview	Sample rows add-on	No

Pricing (pay per event)

Event	Price	When
Dataset	$0.002	Each ranked dataset returned
Sample rows add-on	+$0.002 per dataset	Only if addSampleRows is on and samples were fetched
Actor start	$0.00005	Apify's standard start event

How that compares: catalog and Hugging Face scrapers on the Store run about $1–$5 per 1,000 raw results from one catalog. A ranked top-50 across three catalogs costs $0.10 here.

Billing examples: no incremental mode; you pay per dataset returned: 25 datasets = $0.05; 25 with sample rows = $0.10; no matches = start event only. Set Max charge per run to cap spend.

Input examples

{ "query": "wildfire risk", "openLicenseOnly": true, "maxItems": 25 }

{ "query": "traffic crashes", "sources": ["socrata"], "formats": ["csv"], "updatedWithinDays": 30, "addSampleRows": true }

{ "query": "medical question answering", "sources": ["huggingface"], "maxItems": 20 }

Output (one row)

relevance, source (data.gov, socrata, huggingface), id, title, description, publisher, portal, url, license, formats, keywords, lastUpdated, popularity (views or downloads), columns, apiEndpoint, downloadUrls, plus sampleRows[] with the add-on.

Column lists and types

Socrata results list every column name in columns (up to 100). They also include columnTypes (column → type such as Text, Number or Calendar date), the portal's category, pageViewsTotal and downloadCount. All of it comes straight from the Socrata Discovery API catalog record, so you can check whether a dataset has the fields you need before downloading it. Data.gov and Hugging Face catalogs don't publish column lists, so columns is empty for those rows.

Notes

• Data.gov runs slower on purpose (10 s between pages, per its robots.txt); limit maxResultsPerSource for faster runs.
• Gated and private Hugging Face datasets are skipped.
• Licenses are shown as each catalog reports them; always read the dataset's own terms before reuse. openLicenseOnly keeps only recognised open licenses, so unknown, custom or missing licenses are excluded rather than assumed open.
• lastUpdated is the catalog's metadata date; it doesn't prove the underlying data is fresh. Sample rows show format, not completeness or fitness for your use.
• Sample rows are published open data and are shown as-is; some city datasets include names or addresses that the city chose to publish.

FAQ

How do I find open datasets on a topic? Type a topic in query (e.g. "wildfire risk", "housing permits", "air quality"). The Actor searches Data.gov, the Socrata network of city, county and state portals, and Hugging Face in one run, removes duplicates and returns a ranked list with license, formats, last update and links.

Can I search only one portal, such as NYC Open Data or the Chicago Data Portal? Yes. Set sources to ["socrata"] and socrataDomain to the portal (e.g. data.cityofnewyork.us, data.cityofchicago.org).

Can I find datasets for machine learning or LLM training? Yes. Use sources: ["huggingface"], or keep all three catalogs and turn on openLicenseOnly to keep only recognised open licenses. Always check each dataset's own terms before training a model.

Can I filter by file format or freshness? Yes: formats (csv, json, geojson, parquet, api...) and updatedWithinDays.

Can I preview the data before downloading? Turn on addSampleRows to get 3 real records for each Socrata API dataset, plus columns and columnTypes.

How much does it cost? $0.002 per dataset returned ($2 per 1,000); a ranked top-25 costs $0.05. Sample rows add $0.002 per dataset.

Can AI agents use it? Yes. Call it through the Apify API or the Apify MCP server so an agent can look up datasets on demand, or schedule it to watch for new datasets on a topic.

Data sources and legal notes

• Data.gov catalog (GSA): public search, robots.txt allows it, 10 s crawl delay honored.
• Socrata Discovery API (api.us.socrata.com): documented public API.
• Hugging Face Hub API: documented public API; robots.txt allows.
• Zenodo was considered and excluded because its robots.txt disallows /api.
• Not affiliated with or endorsed by GSA, Tyler Technologies (Socrata) or Hugging Face.

Disclaimer

Results are catalog metadata provided as is. This Actor doesn't verify that data is accurate, current or licensed for your use, and it isn't legal advice. Check each dataset's license and terms before commercial use or model training.

Support

Want another catalog (CKAN portals, Dataverse, EU Open Data)? Open an issue on this Actor's Issues tab. More Clearfeed Labs data feeds: clearfeedlabs.com.

Related Clearfeed Labs Actors

• 311 Complaints by District: Pain-Point & Spike Tracker: 311 complaint spikes by council district
• USAspending Scraper: Federal Contract & Grant Award Alerts: federal contract and grant awards with recompete details
• Grants.gov Scraper & Grant Finder API: NSF, NIH Fellowships: federal grants and fellowships with deadlines
• All Clearfeed Labs Actors