# Romania open data catalogue (data.gov.ro)

> The full catalogue of data.gov.ro, Romania's national open data portal: every published dataset, every downloadable file with its direct URL, and every publishing institution, refreshed daily. Metadata is standardised on the way in — licenses mapped to canonical identifiers, file formats normalised and bucketed into families, Romanian titles given accent-folded search columns — so the whole portal is searchable and joinable with plain SQL, and any tabular file it lists is one read_csv_auto() away.

- Dataset id: `data-gov-ro`
- Update cadence: daily
- Upstream source: https://data.gov.ro/
- Tags: romania, open-data, government, catalogue, ckan, metadata
- Current version: ingest `01M0YSBF6PDT7M5K34WHS2V01Z`, snapshot 2026-08-26, generated 2026-08-26T10:23:51.732Z

## Licence and attribution

- Licence: Creative Commons Attribution 4.0 International (`CC-BY-4.0`)
- Licence text: https://data.gov.ro/pages/licence
- Required attribution: Catalogue metadata from data.gov.ro, the Romanian Government open data portal.
- Notes: This dataset republishes catalogue METADATA only (titles, descriptions, URLs, formats, license ids) — none of the files the catalogue points at. Each underlying dataset carries its own license, recorded per row in license_id (verbatim) and license_std (standardised): mostly CC-BY-4.0 and OGL-ROU-1.0 (both attribution licenses requiring credit to the publishing institution named in the organizations table), but not-specified occurs and at least one dataset is CC-BY-NC-4.0 (non-commercial). Anyone reusing the underlying files must check license_std for that dataset first — this catalogue's own license does not extend to them.

Reproduce the attribution line verbatim anywhere you republish this data.

## Manifest

- Start here for how this directory works: https://data.bunnytech.app/llms.txt
- This dataset: https://data-directory.fsn1.your-objectstorage.com/public/data-gov-ro/manifest.json
- All datasets: https://data-directory.fsn1.your-objectstorage.com/public/index.json
- Human page: https://data.bunnytech.app/datasets/data-gov-ro

The manifest is the contract: file URLs, row counts, column types, min/max statistics and
example queries all live there. Re-read it to discover a new version.

## Fastest path — Parquet exports

Immutable, physically sorted snapshot exports. No catalog fetch, best row-group pruning,
safe to cache forever. Register one view per table:

```sql
INSTALL httpfs; LOAD httpfs;

CREATE OR REPLACE VIEW datasets AS
  SELECT * FROM read_parquet('https://data-directory.fsn1.your-objectstorage.com/public/data-gov-ro/v/01M0YSBF6PDT7M5K34WHS2V01Z/datasets.parquet');

CREATE OR REPLACE VIEW resources AS
  SELECT * FROM read_parquet('https://data-directory.fsn1.your-objectstorage.com/public/data-gov-ro/v/01M0YSBF6PDT7M5K34WHS2V01Z/resources.parquet');

CREATE OR REPLACE VIEW organizations AS
  SELECT * FROM read_parquet('https://data-directory.fsn1.your-objectstorage.com/public/data-gov-ro/v/01M0YSBF6PDT7M5K34WHS2V01Z/organizations.parquet');
```

Example queries:

-- Search the catalogue for a topic, diacritics-insensitively: Query title_lc, never lower(title): the folded column matches "sanatate" against "sănătate" and is the sort column, so a prefix predicate prunes.
```sql
ATTACH IF NOT EXISTS 'ducklake:https://data-directory.fsn1.your-objectstorage.com/catalog/catalog.ducklake' AS "datadirectory" (READ_ONLY);

SELECT d.title, o.title AS institution, d.license_std, d.resource_count, d.dataset_url
FROM datadirectory.data_gov_ro.datasets d
LEFT JOIN datadirectory.data_gov_ro.organizations o ON o.organization_id = d.organization_id
WHERE d.title_lc LIKE '%sanatate%'
ORDER BY d.metadata_modified DESC NULLS LAST
LIMIT 25
```

-- Files DuckDB can load directly, for one dataset: format_group = 'tabular' is the standardised filter; download_url feeds read_csv_auto / st_read straight from the portal.
```sql
ATTACH IF NOT EXISTS 'ducklake:https://data-directory.fsn1.your-objectstorage.com/catalog/catalog.ducklake' AS "datadirectory" (READ_ONLY);

SELECT r.name, r.format, r.size_bytes, r.download_url
FROM datadirectory.data_gov_ro.resources r
JOIN datadirectory.data_gov_ro.datasets d ON d.dataset_id = r.dataset_id
WHERE d.name = 'siruta' AND r.format_group = 'tabular'
ORDER BY r.resource_id
```

## Live lake — DuckLake catalog with time travel

Attach the published catalog read-only over plain https. This works in native DuckDB AND in
duckdb-wasm in the browser — the reader catalog resolves its data files through public https
URLs, so no credentials and no backend are involved.

```sql
INSTALL ducklake; LOAD ducklake;
ATTACH 'ducklake:https://data-directory.fsn1.your-objectstorage.com/catalog/data-gov-ro/catalog.ducklake' AS data_gov_ro (READ_ONLY); SELECT * FROM data_gov_ro.main.datasets LIMIT 5;
USE data_gov_ro;
```

- Latest catalog: https://data-directory.fsn1.your-objectstorage.com/catalog/data-gov-ro/catalog.ducklake
- This exact version: https://data-directory.fsn1.your-objectstorage.com/catalog/data-gov-ro/v/01M0YSBF6PDT7M5K34WHS2V01Z/catalog.ducklake (sha256 929935accedb206644751c8e25d5073ef8dc208a2ef1648af480f9dd66804c88)
- Written by DuckDB v1.5.5
- Time travel: this ingest is snapshot 4 — e.g. `SELECT * FROM data_gov_ro.datasets AT (VERSION => 4)`

## Wholesale — S3 access

For bulk downloads, read the bucket prefix directly over S3 with httpfs:

```sql
INSTALL httpfs; LOAD httpfs;
CREATE SECRET hetzner (
  TYPE s3,
  ENDPOINT 'fsn1.your-objectstorage.com',
  URL_STYLE 'path',
  KEY_ID '<public-read-not-required>', SECRET ''
);
SELECT * FROM read_parquet('s3://data-directory/public/data-gov-ro/v/01M0YSBF6PDT7M5K34WHS2V01Z/**/*.parquet');
```

## Schema

### datasets
One row per dataset (CKAN "package") published on data.gov.ro, carrying the most recent metadata the portal reports for it. A dataset is a titled collection of one or more downloadable files (see the resources table) owned by one public institution (see organizations). Rows are updated in place as institutions edit their metadata; first_seen_date / last_seen_date give the window over which this catalogue observed the dataset, so newly published datasets are the rows with a recent first_seen_date.
- 5,226 rows · 482 KB · write mode `upsert`
- Physically sorted by: `title_lc`, `dataset_id`
- Primary key: `dataset_id`
- Text search columns: `title`
- Files: 1
  - https://data-directory.fsn1.your-objectstorage.com/public/data-gov-ro/v/01M0YSBF6PDT7M5K34WHS2V01Z/datasets.parquet (5,226 rows)

Columns:
- `dataset_id` VARCHAR NOT NULL — CKAN package id — a UUID for datasets created since ~2015, a legacy slug for the oldest ones. Stable across metadata edits; the primary key, and what resources.dataset_id joins to. (range: "00138538-d1f2-4209-a14c-e4c12b6792f4" … "ffe7edc5-8f06-461e-8e74-c06fc9350b43") (~5845 distinct)
- `name` VARCHAR NOT NULL — URL slug of the dataset, unique on the portal. dataset_url is derived from it.
- `title` VARCHAR NOT NULL — Display title as written by the publishing institution, usually in Romanian. Falls back to the slug when the portal reports no title.
- `title_lc` VARCHAR NOT NULL — title, lowercased and accent-folded (ș/ş/s all match), for case- and accent-insensitive search. Query this, not lower(title) — a function on a column cannot use parquet statistics, and Romanian titles mix cedilla and comma diacritics inconsistently. (range: "- lista organisme private autorizate romane autorizate de anpdca" … "– pni anghel saligny – stadiu implementare – 30 iunie 2023")
- `description` TEXT — Free-text description ("notes") as written by the publisher, usually in Romanian. Quality varies from careful documentation to empty.
- `organization_id` VARCHAR — Id of the publishing institution; joins to organizations.organization_id. Null only for the rare orphaned dataset the portal reports without an owner.
- `organization_name` VARCHAR — URL slug of the publishing institution, denormalised for filtering without a join.
- `license_id` VARCHAR — The portal's raw license id, verbatim. Messy by inheritance: 'uk-ogl' is stock CKAN's id that data.gov.ro reuses for the Romanian Open Government License, and null / 'notspecified' both occur. Prefer license_std for filtering.
- `license_std` VARCHAR NOT NULL — Standardised license identifier, derived: CC-BY-4.0, OGL-ROU-1.0, ODC-PDDL-1.0, CC0-1.0, CC-BY-NC-4.0, ODbL-1.0 …; not-specified when the publisher chose none, other:<raw-id> for anything unmapped. 'uk-ogl' maps to OGL-ROU-1.0 because the portal itself titles it exactly that (its license page: datasets published before the 2020 CC-BY-4.0 recommendation are under Licența pentru o Guvernare Deschisă OGL-ROU 1.0 and remain so). Check this column before reusing the underlying files.
- `tags` JSON — Tag names as a JSON string array, e.g. ["CAEN", "NACE"]. Null when the publisher set none — most datasets. Free vocabulary, not a taxonomy.
- `groups` JSON — Portal category (CKAN group) names as a JSON string array. Null when the dataset is in no category — the common case on this portal.
- `resource_count` INTEGER NOT NULL [resources] — Number of usable resource rows this dataset contributes to the resources table.
- `metadata_created` TIMESTAMP — When the dataset was created on the portal, as reported — a naive local-time wall clock with no timezone, stored verbatim. Null on the oldest datasets, migrated in bulk in 2013 before the portal kept creation times.
- `metadata_modified` TIMESTAMP — When the dataset metadata was last edited, as reported by the portal (naive local-time wall clock, stored verbatim). The freshness signal: recently maintained datasets have a recent value.
- `dataset_url` VARCHAR NOT NULL — Human-readable dataset page on the portal.
- `first_seen_date` DATE NOT NULL — First run day (UTC) on which this catalogue saw the dataset. Maintained by the upsert loader — never overwritten, only widened. A recent value means a newly published dataset.
- `last_seen_date` DATE NOT NULL — Most recent run day (UTC) on which the portal still listed the dataset. A stale value means the dataset was removed or hidden upstream; rows are never deleted here.

### resources
One row per downloadable file or endpoint ("resource") attached to a dataset on data.gov.ro. This is the actionable table: download_url is a direct link DuckDB can often read in place (e.g. read_csv_auto(download_url) for format = 'CSV'). Filter format_group = 'tabular' for files that load as tables. Rows update in place as institutions replace files; join datasets via dataset_id for the title, license and publisher.
- 39.5K rows · 3.2 MB · write mode `upsert`
- Physically sorted by: `dataset_id`, `resource_id`
- Primary key: `resource_id`
- Text search columns: `name`
- Files: 1
  - https://data-directory.fsn1.your-objectstorage.com/public/data-gov-ro/v/01M0YSBF6PDT7M5K34WHS2V01Z/resources.parquet (39.5K rows)

Columns:
- `resource_id` VARCHAR NOT NULL — CKAN resource id (UUID). Stable across file replacements; the primary key. (range: "0000b018-ad09-4632-b069-3b54822e23f4" … "ffff0368-2a82-49a4-82f2-d3a5654656f0") (~31360 distinct)
- `dataset_id` VARCHAR NOT NULL — Id of the dataset this file belongs to; joins to datasets.dataset_id. (range: "00138538-d1f2-4209-a14c-e4c12b6792f4" … "ffe7edc5-8f06-461e-8e74-c06fc9350b43")
- `name` VARCHAR — File name or resource title as the publisher wrote it.
- `name_lc` VARCHAR — name, lowercased and accent-folded, for case- and accent-insensitive search. Query this, not lower(name).
- `description` TEXT — Free-text description of the file, usually Romanian; often empty.
- `format` VARCHAR — File format, standardised from the portal's free text: trimmed, leading dot stripped, uppercased — 'csv', '.csv' and 'CSV' all become 'CSV'. Publisher-declared, so it can still lie about the bytes behind the URL.
- `format_group` VARCHAR — Format bucketed into a standard family, derived from format: tabular (CSV/XLSX/ODS/PARQUET…), structured (JSON/XML/RDF…), geospatial (GeoJSON/SHP/KML…), document (PDF/DOC/HTML…), archive (ZIP/RAR…), image, api, or other. The one-predicate answer to "which files can I actually load as data".
- `mimetype` VARCHAR — MIME type as recorded by the portal.
- `size_bytes` BIGINT [bytes] — File size in bytes as recorded by the portal. Null for external links and datastore resources; BIGINT because bulk uploads exceed 2 GiB.
- `download_url` VARCHAR — Direct download URL. For uploaded files it is served by the portal itself; for linked resources it points at the institution's own server and can rot. This is the column that turns the catalogue into data: pass it to read_csv_auto / read_parquet / read_json_auto in DuckDB.
- `created` TIMESTAMP — When the resource was added, as reported by the portal (naive local-time wall clock, stored verbatim).
- `last_modified` TIMESTAMP — When the file behind the resource last changed, as reported (naive local-time wall clock, stored verbatim). Null when the portal never recorded a modification.
- `datastore_active` BOOLEAN — True when the file is also loaded into the portal's own DataStore (queryable via its datastore_search API), not just stored as a blob.
- `first_seen_date` DATE NOT NULL — First run day (UTC) on which this catalogue saw the resource. Maintained by the upsert loader — never overwritten, only widened.
- `last_seen_date` DATE NOT NULL — Most recent run day (UTC) on which the portal still listed the resource. A stale value means the file was removed or replaced upstream.

### organizations
One row per publishing institution ("organization") on data.gov.ro — ministries, agencies, county councils, city halls. dataset_count makes it the leaderboard of who actually publishes; join from datasets.organization_id for everything else. Rows update in place; institutions are effectively never deleted.
- 130 rows · 16 KB · write mode `upsert`
- Physically sorted by: `title_lc`, `organization_id`
- Primary key: `organization_id`
- Text search columns: `title`
- Files: 1
  - https://data-directory.fsn1.your-objectstorage.com/public/data-gov-ro/v/01M0YSBF6PDT7M5K34WHS2V01Z/organizations.parquet (130 rows)

Columns:
- `organization_id` VARCHAR NOT NULL — CKAN organization id (UUID). Stable; the primary key. (range: "001b671a-b2fe-4a10-9036-ed1183699bcd" … "fe2a426f-a95a-4a5b-9115-6754f9abfb2c") (~137 distinct)
- `name` VARCHAR — URL slug of the institution on the portal. organization_url is derived from it.
- `title` VARCHAR NOT NULL — Institution name as displayed on the portal, in Romanian. Falls back to the slug, then the id, when the portal reports no title.
- `title_lc` VARCHAR NOT NULL — title, lowercased and accent-folded, for case- and accent-insensitive search. Query this, not lower(title). (range: "administratia fluviala a dunarii de jos galati" … "unitatea executiva pentru finantarea invatamantului superior, cercetarii, dezvoltarii si inovarii")
- `description` TEXT — Self-description of the institution, usually Romanian; often empty.
- `image_url` VARCHAR — Logo URL as configured on the portal.
- `created` TIMESTAMP — When the institution joined the portal, as reported (naive local-time wall clock, stored verbatim).
- `dataset_count` INTEGER NOT NULL [datasets] — Number of active datasets this institution had in the most recent catalogue snapshot. Recomputed every run, so it tracks the portal, not our history.
- `organization_url` VARCHAR — Institution page on the portal, listing all its datasets.
- `first_seen_date` DATE NOT NULL — First run day (UTC) on which this catalogue saw the institution. Maintained by the upsert loader.
- `last_seen_date` DATE NOT NULL — Most recent run day (UTC) on which the institution still had at least one dataset listed.


## Performance notes

- There are no indexes in a lakehouse. The physical sort order plus Parquet row-group min/max
  statistics ARE the index. Predicates on the leading sort columns prune whole row groups;
  everything else is a full column scan.
- `datasets` prunes on `title_lc` → `dataset_id`.
- `resources` prunes on `dataset_id` → `resource_id`.
- `organizations` prunes on `title_lc` → `organization_id`.
- `ILIKE '%needle%'` cannot prune: it decompresses that whole column. Prefer prefix predicates
  on the lowercased search companion columns where they exist.
- Every file URL under `/v/<ingestId>/` is immutable. Cache it as aggressively as you like.
- Files are never rewritten in place — a new ingest writes a new version path. Re-read the
  manifest to pick up updates rather than polling the file URLs.
- Pull only the columns you need. Parquet is columnar and the reader range-GETs what it reads.

## Do not

- Do not scrape https://data.bunnytech.app HTML. Every number on those pages comes from the manifest above.
- Do not hit the upstream source (https://data.gov.ro/) in bulk. This mirror exists so you do not have to.
- Do not read DuckLake internal data files with a plain Parquet reader. They come with
  positional delete files, and ignoring those silently returns deleted rows. Use the ducklake
  extension, or use the published exports above.
- Do not republish without the attribution line, and do not imply endorsement.