Mintmark 0.1.0
The first tagged release of the Mintmark engine.
What this is
A generation engine for fully synthetic, Turkish-first labeled datasets. Every
value is derived from a seed by a deterministic function, with no model anywhere
in the generation path. Spans are labeled against a closed eighteen-label
taxonomy pinned by digest, and each dataset carries a manifest that lets anyone
re-derive the identical bytes.
Installing
Mintmark is not on PyPI, and the name on that index is unclaimed. Install
from this repository:
uv tool install git+https://github.com/lokomotifai/mintmark@v0.1.0A command naming the package index instead would install something this project
did not build.
Attached
| Asset | What it is |
|---|---|
mintmark-0.1.0-py3-none-any.whl |
The wheel |
mintmark-0.1.0.tar.gz |
The source distribution |
sbom.json |
CycloneDX software bill of materials |
SHA256SUMS |
A checksum for each of the above |
Verified rather than assumed
Four facts this engine depends on live in public registries and specifications,
and each was checked against its source with the record kept in
docs/normative-verification.md: the VKN checksum algorithm against two
independent implementations over 200 000 inputs, the bank code reserved for
synthetic IBANs, the permanent UTC+3 offset for Turkey, and the negative result
that the Turkish numbering plan reserves no fictional mobile range.
Determinism, and what it covers
Identical engine version, pack digest, recipe, seed, identifier policy, and
output format produce byte-identical data files and label sidecars, on CPython
3.12 on Linux x86_64, Linux arm64, and macOS arm64. The manifest's provenance
block is excluded from that claim because it records when the mint ran.
Known limitations
Generated phone numbers can coincide with assigned ones, because the Turkish
numbering plan reserves no fictional range. Never contact a number from this
data.
The name Mintmark is provisional pending trademark screening.
Licensing
Code is Apache-2.0. Datasets minted by this engine carry the terms their pack
declares, written into every manifest along with a credit line. Every pack in
this family declares CC BY 4.0. See LICENSE-DATASETS.md.