# Hugging Face Publishing

## Canonical target

- Account: `PureOne`
- Repository: `PureOne/dirichlet-tree-polya-equality-rigidity`
- Repository type: `dataset`
- Visibility: **public**
- URL after publication: https://huggingface.co/datasets/PureOne/dirichlet-tree-polya-equality-rigidity

A dataset repository is used as a neutral research-artifact container because the release is a mathematical paper + code + machine-readable metadata, not a trained model.

## One-click Windows publication

Run:

```text
PUBLISH_HUGGINGFACE.bat
```

The script installs `huggingface_hub` if required, validates the release, asks for the write token using Python's hidden `getpass` prompt, creates the public dataset repository if necessary, uploads the complete folder, and verifies that the resulting repository is public.

The token is **not written to disk** by the publishing script.

PowerShell users may instead run:

```powershell
./PUBLISH_HUGGINGFACE.ps1
```

Cross-platform direct invocation:

```bash
python -m pip install -r requirements.txt
python publish_huggingface.py
```

## Dry-run validation

```bash
python publish_huggingface.py --dry-run
```

## Search/index optimization already included

The root `README.md` is a Hugging Face repository card with focused mathematics/search tags. The release additionally contains:

- full paper in PDF, LaTeX, and Markdown;
- `AI_AGENT_GUIDE.md`;
- `AI_CONTEXT.md`;
- `llms.txt`;
- structured `metadata/research_manifest.json`;
- structured `metadata/claims.json`;
- theorem and source/dependency registries;
- reproducibility code and data;
- citation metadata;
- explicit expert-review and claim-boundary files.

These files are intended to make the project easy to discover, parse, cite, audit, and ingest by researchers and AI agents.