Developers

Build AI products with SVGZO data.

Everything ships as files you already know how to read: parquet, tar, JSON-LD. The API follows once there is demonstrated demand.

Package layout

svgzo-<slug>-v<semver>/
  README.md            dataset card
  LICENSE.txt          tier-specific licence
  NOTICES/             OFL, UFL, Apache texts; font copyright lines
  croissant.json       schema.org Dataset + Croissant 1.0 + RAI
  provenance/          per-asset manifest; illustration sources; EU summary paragraph
  metadata/            parquet shards; index.csv
  svg/                 tar shards, 2 to 5 GB each
  svg_normalized/
  previews/            720px WebP
  taxonomy/
  quality/             validation, dedup and complexity reports
  checksums/           SHA256SUMS per shard
  CHANGELOG.md

Loading

from datasets import load_dataset
ds = load_dataset("parquet", data_files="metadata/*.parquet", split="train")

Syncing a licensed dataset

# scoped credentials are issued per licence grant
rclone sync svgzo:svgzo-infographics-v1.0.0 ./infographics --checksum --progress
sha256sum -c infographics/checksums/SHA256SUMS

Machine-readable metadata

Every dataset page embeds schema.org/Dataset JSON-LD that conforms to Croissant 1.0, and serves a full manifest with the record set at /datasets/{slug}/croissant.json.

API roadmap

Planned API endpoints
Endpoint	Returns	Phase
`GET /api/v1/datasets`	Catalogue with versions and licence tiers	3
`GET /api/v1/datasets/{id}`	Dataset object	3
`GET /api/v1/datasets/{id}/metadata`	Signed URL to parquet shards	3
`GET /api/v1/datasets/{id}/samples`	Sample pack	3
`GET /api/v1/assets/{id}`	Asset record and signed file URL	3

Later: Python SDK, dataset streaming, usage tracking.

Package layout

Loading

Syncing a licensed dataset

Machine-readable metadata

API roadmap

Request API access