README.md

# ContentVault

Content packs for [Lucky Robots](https://luckyrobots.com). Each pack contains a `metadata.yaml`, a `thumbnail.png`, and a `.zip` with the pack contents.

## Content Packs

| Pack | Description |
|------|-------------|
| ExamplePack | Example content pack |
| Go1VelocityTracking | Unitree Go1 velocity tracking |
| MujocoExample | MuJoCo simulation example |
| Oscillator | Oscillator environment |
| Panda | Franka Panda robot |
| Piper | Piper robot |
| Piper-Pick-Place | Piper pick and place task |
| PiperUnscrewCap | Piper cap unscrewing task |
| Skies-Vol-1 | Sky HDRIs volume 1 |
| TheBungalow | Bungalow environment |
| TheLoft | Loft environment |
| TheOffice | Office environment |
| UnitreeG1 | Unitree G1 humanoid |
| UnitreeGo1 | Unitree Go1 quadruped |
| Welcome | Welcome / intro pack |

## CDN

All content packs are served globally via Cloudflare CDN:

```
https://contentvault.luckyrobots.com/{PackName}/{filename}
```

For example:
```
https://contentvault.luckyrobots.com/ExamplePack/metadata.yaml
https://contentvault.luckyrobots.com/TheLoft/TheLoft.zip
```

## How sync works

A Cloudflare Worker (`worker/`) listens for webhooks from HuggingFace. When a push is made to the [HuggingFace dataset repo](https://huggingface.co/datasets/luckyrobots/ContentVault), the Worker:

1. Lists all files in the repo via the HuggingFace API
2. Streams each file directly to Cloudflare R2 (no buffering, handles large zip files)
3. R2 serves the files through Cloudflare's CDN with a 30-day cache, cached at 300+ edge locations worldwide

## Worker setup

```bash
cd worker
npm install
```

### Secrets (set via wrangler)

```bash
npx wrangler secret put HF_WEBHOOK_SECRET
npx wrangler secret put HF_TOKEN
```

### Deploy

```bash
npx wrangler deploy
```

### Commands

## Setup

```bash
# Install huggingface_hub with fast upload backend
pip install -U "huggingface_hub[hf_xet]"

# Login (requires write-access token from https://huggingface.co/settings/tokens)
hf auth login
```

---

## CLI Commands

### Upload entire folder
```bash
hf upload <repo-id> <local-folder> . --repo-type dataset --commit-message "Your message"
```

### Upload a single pack/subfolder
```bash
hf upload <repo-id> <local-folder>/<pack-name> <pack-name> --repo-type dataset --commit-message "Add <pack-name>"
```

### Upload a single file
```bash
hf upload <repo-id> <local-file> <path-in-repo> --repo-type dataset --commit-message "Add file"
```

### Upload large folder (resumable, multi-threaded, auto-retry)
```bash
hf upload-large-folder <repo-id> <local-folder> --repo-type=dataset
```

### Delete files remotely (without re-uploading everything)
```bash
hf upload <repo-id> . . --repo-type dataset --include="" --delete="<folder-name>/*" --commit-message "Remove <folder-name>"
```

### Create a new dataset repo
```bash
hf repo create <repo-name> --type dataset
```

### Create under an organization
```bash
hf repo create <repo-name> --type dataset --organization <org-name>
```

---

## Python One-Liners

### Upload large folder (resumable)
```bash
python -c "from huggingface_hub import HfApi; HfApi().upload_large_folder(repo_id='<repo-id>', repo_type='dataset', folder_path='<local-folder>')"
```

### Upload a single folder/pack
```bash
python -c "from huggingface_hub import HfApi; HfApi().upload_folder(folder_path='<local-folder>', path_in_repo='<path-in-repo>', repo_id='<repo-id>', repo_type='dataset', commit_message='Your message')"
```

### Upload a single file
```bash
python -c "from huggingface_hub import HfApi; HfApi().upload_file(path_or_fileobj='<local-file>', path_in_repo='<path-in-repo>', repo_id='<repo-id>', repo_type='dataset', commit_message='Your message')"
```

### Delete a folder
```bash
python -c "from huggingface_hub import HfApi; HfApi().delete_folder('<folder-name>', repo_id='<repo-id>', repo_type='dataset', commit_message='Remove <folder-name>')"
```

### Delete a single file
```bash
python -c "from huggingface_hub import HfApi; HfApi().delete_file('<file-path>', repo_id='<repo-id>', repo_type='dataset', commit_message='Remove <file-path>')"
```

### Create a repo
```bash
python -c "from huggingface_hub import HfApi; HfApi().create_repo(repo_id='<repo-id>', repo_type='dataset')"
```

---

## Python Script Examples

### Upload large folder (full script with options)
```python
from huggingface_hub import HfApi

api = HfApi()
api.upload_large_folder(
    repo_id="<repo-id>",
    repo_type="dataset",
    folder_path="<local-folder>",
)
```

### Upload a folder to a specific path in the repo
```python
from huggingface_hub import HfApi

api = HfApi()
api.upload_folder(
    folder_path="<local-folder>",
    path_in_repo="<path-in-repo>",
    repo_id="<repo-id>",
    repo_type="dataset",
    commit_message="Your message",
    ignore_patterns=["*.cache", ".git/*"],
)
```

### Batch delete + upload in one commit
```python
from huggingface_hub import HfApi, CommitOperationAdd, CommitOperationDelete

api = HfApi()
operations = [
    CommitOperationDelete(path_in_repo="<old-folder>/"),
    CommitOperationAdd(path_in_repo="<file-path>", path_or_fileobj="<local-file>"),
]
api.create_commit(
    repo_id="<repo-id>",
    repo_type="dataset",
    operations=operations,
    commit_message="Your message",
)
```

---

## Tips

- **Resumable uploads**: `upload_large_folder` and `hf upload-large-folder` are resumable. If interrupted, run the same command again.
- **Don't use raw `git push`** for large files -- use the CLI/API instead. They handle LFS/Xet automatically.
- **Write token required**: Tokens default to read-only. Enable write access at https://huggingface.co/settings/tokens.
- **50 GB max per file** on HuggingFace. Split larger files before uploading.
- **Ignore patterns**: Use `--include` and `--exclude` (CLI) or `ignore_patterns` (Python) to filter what gets uploaded.
- **Delete the `.cache/huggingface/` folder** inside your local folder to reset upload state if something goes wrong with `upload_large_folder`.