UPLOAD_GUIDE.md

# Uploading this release to Hugging Face

Suggested repository name: **ORYNTHRA-H6**  
Suggested title: **ORYNTHRA-H6 — Branch-Covariant Compression and the Spectral Geometry of Dense Light**

Short description:

> A mathematical research release connecting future-compatible task compression with conserved-sector quantum memory. Includes an exact representation-deficit law under a stated coherent-control model, a six-port simplex register, instruction-activation examples, a 39-page manuscript, reference code, synthetic data, and verification records.

The release is prepared locally; these instructions do not mean it has already been uploaded. Select a **Dataset** repository. The dataset card declares five separately loadable synthetic tables; the manuscript, software, figures, and verification records accompany them. The published page will be `https://huggingface.co/datasets/YOUR_USERNAME/ORYNTHRA-H6`.

## Option 1: Windows launcher

1. Extract the downloaded ZIP. Open the `ORYNTHRA-H6` folder.
2. Install Python if needed. In a terminal opened in that folder, install the small uploader dependency:

   ```bat
   py -3 -m pip install -r requirements-upload.txt
   ```

3. Authenticate using a Hugging Face token with write access to your intended repository:

   ```bat
   hf auth login
   ```

4. Double-click `upload_to_hf.bat`. The launcher prints the action, publishes the release, and keeps its window open on completion or error. It derives your personal username from the logged-in account and uses `<username>/ORYNTHRA-H6`.

To choose an organization, another name, or a private new repository, run:

```bat
py -3 tools\hf_upload.py organization/ORYNTHRA-H6 --publish --private
```

The script preserves the audience of an existing repository. `--private` chooses the visibility of a newly created repository; it does not convert an existing repository.

## Option 2: Terminal on any platform

```bash
python -m pip install -r requirements-upload.txt
hf auth login
python tools/hf_upload.py
python tools/hf_upload.py YOUR_USERNAME/ORYNTHRA-H6 --publish
```

The command without `--publish` only checks local file hashes and prints a plan. It makes no network calls and requires no account. `upload_to_hf.sh` is a publishing launcher for Linux/macOS after setup.

## Option 3: Hugging Face CLI

After logging in, from the unmodified release folder:

```bash
hf repos create YOUR_USERNAME/ORYNTHRA-H6 --type dataset --exist-ok
hf upload YOUR_USERNAME/ORYNTHRA-H6 . . --type dataset --commit-message "ORYNTHRA-H6 research release v1.0.2"
```

The supplied Python uploader offers stronger local file selection: it checks hashes and uploads only the files declared in `SHA256SUMS.txt`, plus that manifest. Its default behavior does not delete any remote files. The CLI alternative relies on the release's root ignore rules; use the clean extracted folder.

## Option 4: Web interface

Open [Hugging Face's new repository page](https://huggingface.co/new), select **Dataset**, choose a name and visibility, then use **Files and versions → Add file → Upload files**. Upload the **contents** of the `ORYNTHRA-H6` folder so `README.md` sits at the repository root. Uploading the ZIP alone does not expose its README as the repository card. If your browser cannot preserve nested folders, use the supplied uploader.

## Authentication and errors

Tokens belong in Hugging Face's login storage or the `HF_TOKEN` environment variable, not in the package, README, command-line token arguments, or screenshots. No token is included in this release. The uploader reports error class/status without printing credentials or full network tracebacks.

The uploader retries a bounded number of transient server/rate-limit/transport failures. If it still fails, correct the connection or token permissions and rerun. A local receipt is written in `.upload_state/` after success and excluded from uploads. Successful publishing should show a repository URL. Windows-native execution and a real upload have not been tested in this preparation session.

If a checksum differs, use a fresh extraction or deliberately prepare a new version with a regenerated manifest. Reproduction commands overwrite generated files, so run them in another copy. Extra unlisted files are excluded by the Python uploader.

## Rights and claims

The original package did not select an additional reuse license. This edition preserves that status; [RIGHTS.md](RIGHTS.md) explains it. If you later choose a license, record it explicitly and update the card and release checksums.

The research card distinguishes theorem proofs, computational checks, novelty candidates, and hardware assumptions. Avoid replacing those with claims of established worldwide uniqueness, universal file compression, fault-tolerant optical memory, or a validated photonic device.

## Official references

- [Hub repository creation](https://huggingface.co/docs/hub/repositories-getting-started)
- [Repository card metadata](https://huggingface.co/docs/hub/datasets-cards)
- [File and folder upload API](https://huggingface.co/docs/huggingface_hub/en/guides/upload)
- [Current Hugging Face CLI](https://huggingface.co/docs/huggingface_hub/en/guides/cli)

Preparation checked the official documentation on 9 October 2026. Repository-hosting behavior and CLI details can change; the local mathematical release is independent of those changes.