Datasets documentation

Open reference datasets, published as immutable Parquet and queryable in your browser with DuckDB: the catalogue, their open licences, and where to start.

Meridian Datasets is a small, curated commons of open reference data — the slow-changing tables you reach for again and again when you're resolving companies, industries, or securities. Each dataset is published once as an immutable Parquet file on Cloudflare R2 and served from a stable URL at openlake.meridian.online. Nothing to download, nothing to sign up for.

Every dataset has a live explorer at /datasets. Open one and a DuckDB engine boots in your browser tab — no server touches your query. Sort, filter, facet, run SQL, profile the columns, and export the result to CSV, all against the real file over HTTP range reads. The same file is addressable from anywhere: one ATTACH exposes the whole catalogue to DuckDB, Python, or R (see Use anywhere).

The point is not to be a data warehouse. It's to make a handful of authoritative reference tables instantly queryable, honestly sourced, and cheap to trust — so you can spend your time on the analysis, not on wrangling a download.

The catalogue

The published manifest always carries each dataset's current release, its row count and its published timestamp, so this page gives none. The NDC Directory's page says how it is refreshed; for the others we promise no cadence. See Data & provenance for what we promise.

Dataset	Licence	Source	Explore
GLEIF — Legal Entity Identifiers	CC0	GLEIF golden copy	/datasets/gleif/lei
SEC EDGAR — Company Tickers	Public domain	SEC EDGAR	/datasets/sec/company_tickers_exchange
NAICS — Industry Classification	Public domain	U.S. Census Bureau	/datasets/census/naics_2022
EDGAR ↔ GLEIF — Company-to-LEI Crosswalk	CC0	SEC EDGAR + GLEIF	/datasets/library/edgar_lei
HTS — Harmonized Tariff Schedule of the United States	Public domain	U.S. International Trade Commission	/datasets/usitc/hts
Schedule B — U.S. Export Classification Codes	Public domain	U.S. Census Bureau	/datasets/census/schedule_b
HCPCS Level II — Healthcare Common Procedure Coding System	Public domain	Centers for Medicare & Medicaid Services	/datasets/cms/hcpcs
NDC Directory — National Drug Code	CC0	U.S. Food and Drug Administration	/datasets/fda/ndc

GLEIF — the Global Legal Entity Identifier Foundation's public register: every legal entity with its 20-character LEI, legal name, country, jurisdiction, legal form, and registration status.
SEC EDGAR — every company with securities registered with the U.S. Securities and Exchange Commission, keyed by Central Index Key (CIK), with ticker and exchange.
NAICS — the 2022 North American Industry Classification System, from 20 broad sectors down to 1,012 national industries, each with its official title and description.
EDGAR ↔ GLEIF — the link between SEC EDGAR entities and their GLEIF LEI, which Meridian built from the two public files. Each link carries the tier that says how it was made.
HTS — the U.S. import tariff schedule from the U.S. International Trade Commission, from chapters down to ten-digit statistical lines, with duty rates as the Commission prints them.
Schedule B — the Census Bureau's classification of exported goods, one row per ten-digit commodity, with a row for each chapter, heading and subheading above it.
HCPCS Level II — the alpha-numeric codes the Centers for Medicare & Medicaid Services maintain for billing supplies, drugs, equipment and services.
NDC Directory — the FDA's National Drug Code directory: one row per labeler, product and package.

Open licences only

Everything here is open at the source. Today that means CC0 (GLEIF, the EDGAR ↔ GLEIF crosswalk and the NDC Directory) and U.S. public domain (SEC EDGAR, NAICS, HTS, Schedule B and HCPCS Level II) — free to use, including commercially, with no attribution required.

That is a deliberate gate, not an accident of what we happened to have. A dataset only ships if its licence is proven to permit redistribution; anything unproven is refused by default. Share-alike (ODbL) and proprietary identifier sets are out of scope for the open commons. See Data & provenance for the full policy and per-dataset provenance.