Parquet & CSV Data Sources for Zetaris Quick-Starts

A catalog of the seven Parquet/CSV onboarding pairs, with licences, paths, verification status, and source documentation. See HOWTO.md for setup and checks.

Licence key: 🟢 open · 🟡 conditional or unresolved · 🔴 not included. These markers describe licensing, not runtime status.

SQL: runnable CREATE LIGHTNING FILESTORE TABLE examples follow the Zetaris Knowledge Base and Lightning SQL manuals. See HOWTO.md for public-bucket options and verification.

Script folders: active pairs live in sql/; recorded failures live in sql/known_to_fail/.

Quick-reference table

Verified means the CREATE and matching SELECT statements ran on a live Zetaris instance. Known to fail means a blocking issue was reproduced; see the source's issue. Do not treat an unverified script as live-tested.

The script IDs match their filenames; gaps are intentional. NOAA and AWS Public Blockchain are verified. Five sources have recorded failures: PUDL and Ookla fail at registration; Foursquare and GBIF register but queries fail; Overture supports a bounded sample but cache and analytical queries fail or hang. See each linked issue for details. An earlier installation record says PUDL succeeded; that contradiction remains unresolved.

Partition, snapshot, and release paths change. Before use, run the listing command in the SQL script header instead of relying on the date shown here.

2. 🟢 NOAA Global Historical Climatology Network – Daily (GHCN-D)

• What it is: daily station observations from 1763 onward, distributed as CSV.
• Licence: CC0-1.0. See the AWS registry.
• Docs: GHCN-Daily readme · column format spec
• Status: live-tested; registration and all example queries passed. CACHE TABLE is required for the demonstrated queries because uncached scans were too slow.

3. 🟢 Catalyst Cooperative PUDL (Public Utility Data Liberation Project) — known to fail

• What it is: U.S. electricity, gas, and financial data assembled from EIA, EPA, FERC, PHMSA, and SEC sources.
• Licence: CC-BY-4.0 — AWS Registry of Open Data entry.
• Format: Parquet and SQLite are published from the same nightly/stable build.
• Docs: PUDL Data Access guide (has the full current table list and version-pinning instructions) · Data Dictionary
• Status: registration fails because Zetaris's connector rejects the dotted bucket name pudl.catalyst.coop, even with path-style access. The files remain downloadable outside Zetaris. An earlier installation record reports success; the conflict is unresolved. See ISSUE-03.

4. 🟢 Foursquare Open Source Places — known to fail

• What it is: a global places-of-interest dataset hosted on Source Cooperative.
• Licence: Apache License 2.0 — Foursquare's OS Places docs.
• Docs: Access instructions · Source Cooperative repo browser
• Status: registration succeeds, but cache and query attempts, including a bounded sample, return HTTP 500. The cause is unconfirmed. Check ISSUE-04.

5. 🟡 Overture Maps Foundation — known to fail

• What it is: monthly GeoParquet for places, buildings, transport, addresses, and divisions.
• Licence: varies by theme. Places uses CDLA-Permissive-2.0; Base, Buildings, Divisions, and Transportation use ODbL; Addresses varies by country. See the attribution table.
• Docs: Quickstart · DuckDB guide (has the current release-folder name — it changes roughly monthly)
• Status: registration and bounded samples work. Cache and analytical queries fail with HTTP 500 or hang; a bounded sample is not evidence that analysis works. See ISSUE-05.

7. 🟡 Ookla Speedtest Global Fixed and Mobile Network Performance — known to fail

• What it is: quarterly global broadband and mobile speed-test results on roughly 610 m Web Mercator tiles.
• Licence: CC BY-NC-SA 4.0, non-commercial — AWS open-data-registry YAML. Fine for a hobbyist demo or tutorial repo; not for a commercial deliverable without a separate licence from Ookla.
• Docs: ookla-open-data GitHub · README (has the exact path pattern and schema reference)
• Status: both registrations return HTTP 500. The bucket name has no dots, so PUDL's cause does not apply; Ookla's cause is unconfirmed. See ISSUE-07.

8. 🟡 GBIF (Global Biodiversity Information Facility) species occurrences — known to fail

• What it is: more than 1.6 billion species records from museums, herbaria, and citizen-science networks, refreshed monthly.
• Licence: the AWS registry lists aggregate CC-BY-NC. Individual records may use CC0 or CC-BY; use CC-BY-NC as the conservative whole-snapshot terms.
• Docs: GBIF AWS public data guide (has the current snapshot-date convention) · Citation guidelines
• Status: registration and GUI caching succeed; every query, including the filtered verification query, returns HTTP 500. DESCRIBE succeeds; the cause is unknown. See ISSUE-08.

9. 🟡 AWS Public Blockchain Data

• What it is: daily, date-partitioned Parquet for Bitcoin, Ethereum, and other chains.
• Licence: unresolved. The registry links to an MIT code sample, not a data licence. Do not re-host the data until its terms are confirmed.
• Docs: AWS registry entry · Analytics guidance README (has the full per-chain path structure)
• Status: both tables registered and all example queries passed live.
• Caching: SQL CACHE TABLE worked on eth_transactions but failed on btc_transactions; the Data Explorer cache action had the opposite result. Both tables queried successfully. The timing difference was observed, but its cause is unknown; see the create script header for exact results.

See also: HOWTO.md for the Zetaris onboarding walkthrough, and docs/plans/FUTURES.md plus its recipes/*.md files for everything outside Parquet/CSV — Kafka, logs, JSON/REST, SQL RDBMS, PDFs, and the government open-data sections.