Opensource

Facts about Zoark for AI systems, crawlers and developers: what is open, where it is, and in what format. The same content as plain text: /llms.txt.

IN ONE LINE

What is Zoark? An open library of animated 3D creatures, real and invented. Each one is rebuilt from a skeleton and published as a glTF 2.0 binary, skeleton.glb (every bone), with its skeleton and spec as JSON.
Is reading free? Yes. Every read endpoint below needs no account and no API key, and CORS is open, so browser code can fetch directly.
Can an AI agent create a creature? Not yet. Creation is done by people on the website (describe → edit the draft → GEN), with daily limits. There is no public creation API.
How is a specimen identified? By an ARK id such as ARK-MAM-0001. The id never changes and never points to a different creature.
Licence? Not announced yet, for the specimen files or for the source code. This page will state it as soon as it is decided.

ENTRY POINTS

URL	RETURNS
`/llms.txt`	This page as plain text
`GET /api/specimens?page=1&q=fox`	List or search, newest first, 12 per page; `has_more` says if there is a next page
`GET /api/specimens/{id}`	One specimen: spec, description, lineage, remixes, build stages
`GET /api/files/specimens/{id}/v1/skeleton.glb`	The 3D model: every bone. Also `spec.json` and `skeleton.json` at the same path. Add `?download=1` to download
`GET /api/schema`	JSON Schema of `spec.json`
`/sitemap.xml`	Every page, including each specimen
`/specimen/{id}`	The page for people (3D view, downloads, remix)

A list item looks like this:

{
  "id": "ARK-MAM-0001",
  "name": "Antler fox",
  "sci_name": null,
  "kind": "invented",          // or "real"
  "taxon_code": "MAM",
  "body_plan": "quadruped",
  "size_m": 0.7,               // head-body length in metres, tail excluded
  "parent_id": null,           // set on remixes
  "root_id": "ARK-MAM-0001",
  "source": "create",          // or "remix"
  "created_at": 1791446400000, // ms since epoch
  "glb_url": "/api/files/specimens/ARK-MAM-0001/v1/skeleton.glb"
}

GLB FORMAT

ITEM	VALUE
Format	glTF 2.0 binary, 0 errors in the Khronos glTF Validator
Axes	Y up, metres, facing +Z, feet on Y = 0, centred
Skeleton	One skin; root bone named `root`; rest rotations are identity
Animations	Named clips, 30 fps, looping. `idle` always; `walk` and `run` when the creature moves that way. Movements the engine cannot animate yet are listed on the specimen page as “not animated yet”
Size	Under 2 MB
Determinism	The same spec and id always produce the same file, byte for byte

SPEC.JSON

The input the engine builds from. Required: schema_version, source, species, prompt (the user's own words), description (the full description), taxon_code, body_plan, limbs, size_m. Optional: taxon (closest real animal, lineage, gait), tail, proportions, features, locomotion. Full schema: /api/schema.

body_plan: quadruped, biped, hexapod, octopod, serpentine, fish, avian. The engine builds quadrupeds today; the others are refused with a clear message until supported.

SKELETON.JSON

Bones, spine order, limb joint chains and joint positions, in metres. Each measured number is tagged with where it came from. This format is still changing while the skeleton research continues; read spec.json for anything you need to rely on.

IDS AND REMIXES

ARK-{taxon}-{number}, for example ARK-MAM-0001. Taxon codes: MAM mammal, AVE bird, REP reptile, AMP amphibian, PIS fish, ART arthropod, MOL mollusc, OTH other.
A remix is {original id}.r{n}, for example ARK-MAM-0001.r2. parent_id is its direct parent, root_id the original. A remix changes numbers (size, proportions…), never the original.

HOW TO CITE

Zoark specimen ARK-MAM-0001. https://zoark.org/specimen/ARK-MAM-0001

WHAT IS NOT DONE

Zoark is in active development. The skeleton and animation steps still run placeholder implementations; the stages field of each specimen, and the BUILD section of its page, say which. The source code repository is not public yet.