# Packrift Packingsolver-Compatible Ecommerce Cartonization Fixture Pack

This publication kit is upload-ready for human-authenticated dataset venues. It does not publish, log in, create accounts, handle secrets, choose a license for Packrift, bypass CAPTCHA, or submit paid listings.

Canonical live page: https://packrift-benchmark-navigator.vercel.app/packingsolver-fixtures.html

## What is included

- 120 carton candidates from public Packrift box records.
- 350 generated scenario orders.
- 803 order item lines.
- 1050 generated packingsolver input files under `fixture-pack/instances/`.
- CSV, JSON, Data Package, README, quality report, checksums, Dataset JSON-LD, and OpenAPI descriptors.
- Citation File Format (`CITATION.cff`) and CodeMeta (`codemeta.json`) metadata for archives and citation tools.

## Boundaries

- Scenario groupings are generated from public Packrift product records and are not real customer orders.
- No optimal or quasi-optimal packing solution is claimed.
- Target carton fields are scenario anchors for testing parsers and solver workflows, not solver answers or fulfillment promises.
- No private customer data, supplier cost, margin, live freight promise, or warehouse operational secret is included.
- No separate open-data license is declared here. Reuse requires Packrift approval unless a later public venue record declares a license.

## Human-auth publication targets

1. Zenodo: strongest DOI/citation target once Farhan chooses license/access terms.
2. Hugging Face Datasets: useful for agent/AI discovery if the Packrift account can publish a dataset repo.
3. Figshare/OSF: useful academic/research repository mirrors after license approval.
4. Kaggle: useful if a public dataset listing is desired, but it should use conservative license wording.

## Count treatment after upload

Do not count this local kit. Count only a public no-login venue record that returns Packrift-specific evidence without noindex/auth/challenge markers. Count the public venue record as live, pending, blocked, verified-identity-only, or junk/low-value according to direct verification.

## Data Contract Addendum

- `data-contract.json` and `data-contract.csv` provide a field-level contract for the published tables.
- `csvw-metadata.json` adds W3C CSV on the Web metadata for the CSV files.
- `fixture-croissant.jsonld` adds Croissant-style JSON-LD metadata for agent and ML dataset discovery.
- `schemas/*.schema.json` provide JSON Schema Draft 2020-12 row contracts for cartons, orders, order summaries, and instance-index rows.

These files clarify input-data structure and validation. They do not add real customer orders, live prices, live inventory, freight promises, solver-optimality claims, upstream acceptance claims, or third-party endorsement claims.