Reproduce

Rebuild it yourself.

A reproducibility claim without an executable path is marketing. This is the path.

What is this?

The build surface: what the dataset is made of, the digest of every table, and the commands to rebuild it.

Why should I trust it?

The digests are read from the freeze manifest committed alongside the data, and a sampled rebuild has been checked byte-for-byte against the frozen output.

Where can I verify it?

Reproducibility metricsIncluding what is unverified
The reproduction recordEnvironment and rebuild result
The dataset cardCoverage and limitations
Fetch the payloadsDigest-addressed JSON

Canonical tables: 7
Total files: 1,985
Dataset size: 569.3 MiB
Rebuild: byte-identical

Identity

What you are rebuilding, and what a correct rebuild must match.

Dataset digest

43fd0e228b28ae6bc7e468c3acf68722768bd62b73798eb6631e9e6233b71ed9

Provenance digest

7282b7c34fbd6c2eae7dfca6757f46cdc497b88a9b6312089715479e67be8498

versionAdityaNet_v2_dataset_r1
commitbe0b7e5114db41e24f592f6483fa643d98586d02
specr6
archive2024-02-01 .. 2026-06-17 (UTC)
frozen2026-07-18T08:22:44.103343+00:00

What it is made of

Science

Frozen dataset, pre-registered protocol

SoLEXS
HEL1OS
Level-1 FITS
Parquet

Engineering

Fail-loud parsing, contract-checked

Python
Alembic
SHA-256
Pytest

Frontend

Static islands, evidence ships 0 KB JS

Astro
React
GSAP
Lenis

Infrastructure

Reproducible build, budget-gated CI

GitHub Actions
Static hosting
Strict CSP

Archive to evidence

01
ISSDC archive
Raw Level-1 products, as published
02
Parse & validate
Twenty fail-loud rules; no silent coercion
03
Freeze
Digested, versioned, immutable
04
Publish
Every figure traceable to an artifact

The reproduction recordPer-table digests, pinned environment, rebuild result How the pipeline worksStage by stage, with what each produced