step.md

# Panduan Penggunaan, CDN & Update Dataset tripleS Objekts

Repository Hugging Face: **[fadhilafif98/triples-objekts](https://huggingface.co/datasets/fadhilafif98/triples-objekts)**

---

## 📌 Status Terakhir Snapshot Data (Sync Checkpoint)

> **Catatan:** Data di bawah ini tersimpan otomatis di file [`sync_checkpoint.json`](./sync_checkpoint.json). Ketika ada rilis kartu baru dari Modhaus/COSMO, script update akan **otomatis membandingkan dengan checkpoint ini** sehingga **tidak akan men-download ulang** kartu atau gambar yang sudah ada.

| Parameter | Nilai Terakhir |
| :--- | :--- |
| **Tanggal Sync Terakhir** | `2026-08-29` |
| **Timestamp Data Terakhir (`latest_created_at`)** | `1787922020000` *(2026-08-28T13:00:20Z)* |
| **Total Objekts Terindeks** | `11,025` unik objekts |
| **Total Foto Photocard Ter-upload** | `21,340` file (`_front.png` & `_back.png`) |
| **Total Video Motion & Voice Ter-upload** | `125` file (`.mp4`) |
| **Daftar Season yang Sudah Masuk** | `Atom01`, `Atom02`, `Binary01`, `Binary02`, `Cream01`, `Cream02`, `Divine01`, `Ever01` |
| **Total Member Tercover** | 24 Member (S1–S24) + Semua Sub-unit |
| **File Tracker Status** | [`sync_checkpoint.json`](./sync_checkpoint.json) *(Track 11,025 slug unik)* |

---

## 1. Cara Menggunakan Dataset via Python `datasets`

### A. Load Standar

```bash
pip install datasets pillow
```

```python
from datasets import load_dataset

# 1. Download & load dataset
dataset = load_dataset("fadhilafif98/triples-objekts", split="train")
print(f"Total data: {len(dataset)}")

# 2. Akses item pertama
sample = dataset[0]
print("Member :", sample["member"])
print("Season :", sample["season"])
print("Class  :", sample["class"])
print("Side   :", sample["side"])

# 3. Tampilkan gambar photocard
img = sample["image"]
img.show()
```

---

## 2. Cara Menggunakan Hugging Face sebagai Direct CDN (Web / App / Bot)

Hugging Face menyediakan endpoint CDN global Cloudflare gratis berkecepatan tinggi.

### A. Format URL CDN Gambar (Photocard)

```text
https://huggingface.co/datasets/fadhilafif98/triples-objekts/resolve/main/images/{season}/{member}/{slug}_{side}.png
```

* **Front Image**: `https://huggingface.co/datasets/fadhilafif98/triples-objekts/resolve/main/images/Atom01/JiWoo/atom01-jiwoo-100z_front.png`
* **Back Image**: `https://huggingface.co/datasets/fadhilafif98/triples-objekts/resolve/main/images/Atom01/JiWoo/atom01-jiwoo-100z_back.png`

---

### B. Format URL CDN Video / Motion / Voice Objekts (MP4)

```text
https://huggingface.co/datasets/fadhilafif98/triples-objekts/resolve/main/media/{season}/{member}/{slug}.mp4
```

* **Motion Loop**: `https://huggingface.co/datasets/fadhilafif98/triples-objekts/resolve/main/media/Binary02/SeoYeon/binary02-seoyeon-503z.mp4`
* **Voice Clip**: `https://huggingface.co/datasets/fadhilafif98/triples-objekts/resolve/main/media/Binary02/SeoYeon/binary02-seoyeon-360z.mp4`

Contoh di HTML:
```html
<video autoplay loop muted playsinline width="300">
  <source 
    src="https://huggingface.co/datasets/fadhilafif98/triples-objekts/resolve/main/media/Binary02/SeoYeon/binary02-seoyeon-503z.mp4" 
    type="video/mp4" 
  />
</video>
```

---

### C. Helper Function di JavaScript / TypeScript

```javascript
const HF_CDN_BASE = "https://huggingface.co/datasets/fadhilafif98/triples-objekts/resolve/main";

// 1. Photocard Image CDN URL
export function getObjektImageUrl(season, member, slug, side = 'front', ext = 'png') {
  return `${HF_CDN_BASE}/images/${season}/${member}/${slug}_${side}.${ext}`;
}

// 2. Motion Video / Voice CDN URL
export function getObjektVideoUrl(season, member, slug) {
  return `${HF_CDN_BASE}/media/${season}/${member}/${slug}.mp4`;
}
```

---

### D. Akses Raw JSON / Metadata via CDN

* **Master JSON**: `https://huggingface.co/datasets/fadhilafif98/triples-objekts/resolve/main/all_objekts.json`
* **Summary Stats**: `https://huggingface.co/datasets/fadhilafif98/triples-objekts/resolve/main/summary.json`
* **Checkpoint Tracker**: `https://huggingface.co/datasets/fadhilafif98/triples-objekts/resolve/main/sync_checkpoint.json`
* **Per-Member (e.g. JiWoo)**: `https://huggingface.co/datasets/fadhilafif98/triples-objekts/resolve/main/by_member/JiWoo.json`
* **Per-Season (e.g. Atom01)**: `https://huggingface.co/datasets/fadhilafif98/triples-objekts/resolve/main/by_season/Atom01.json`

---

## 3. Cara Update Data Baru di Masa Depan (Tanpa Mulai dari Awal)

Sistem sudah dilengkapi **checkpoint anti-duplikasi**. Ketika Modhaus merilis season/objekts baru, jalankan 3 langkah ini:

### Langkah 1: Unduh Hanya Data & Gambar Baru (Delta Fetch)
Script otomatis membaca `sync_checkpoint.json` dan hanya mengunduh data/media yang **belum pernah tersimpan**:

```bash
# 1. Unduh metadata baru & foto baru
python3 fetch_objekts.py --download-images

# 2. Unduh video baru jika ada rilis Motion Class baru
python3 fetch_media.py
```

---

### Langkah 2: Sinkronkan Checkpoint & Index Hugging Face
Regenerasi index `metadata.jsonl` dan perbarui snapshot `sync_checkpoint.json`:

```bash
python3 prepare_hf_dataset.py
python3 generate_checkpoint.py
```

---

### Langkah 3: Upload Hanya File Baru ke Hugging Face
Script upload menggunakan chunking dan hanya mengunggah file-file baru (*delta changes*):

```bash
python3 upload_to_hf.py --repo-id fadhilafif98/triples-objekts --token YOUR_HF_TOKEN
```

*(Setelah upload selesai, Anda aman menghapus folder lokal `images/` atau `media/` lagi untuk menghemat kapasitas harddisk)*.

---

## 4. Struktur File Lokal & Fungsinya

| File / Direktori | Deskripsi |
| :--- | :--- |
| `sync_checkpoint.json` | **File state tracker** berisi riwayat 11,025 objekts yang sudah ter-upload agar update berikutnya tidak download ulang |
| `all_objekts.json` | Master JSON mentah seluruh objekts dari Apollo / COSMO API |
| `summary.json` | Ringkasan statistik jumlah objekts per member, season, dan class |
| `by_member/` | Direktori berisi file JSON yang dipisah per masing-masing member (S1-S24) |
| `by_season/` | Direktori berisi file JSON yang dipisah per masing-masing season |
| `images/` | Direktori lokal foto photocard HD (`_front.png` & `_back.png`) |
| `media/` | Direktori lokal video motion & voice (`.mp4`) |
| `metadata.jsonl` | File mapping index standar Hugging Face ImageFolder |
| `fetch_objekts.py` | Script fetcher dengan auto-checkpoint filter |
| `fetch_media.py` | Script downloader untuk video Motion MP4 |
| `prepare_hf_dataset.py` | Script mapping file ke `metadata.jsonl` |
| `upload_to_hf.py` | Script pengunggah multi-commit ke Hugging Face |
| `step.md` | Dokumentasi panduan penggunaan, CDN, dan update |
| `README.md` | Dataset Card resmi di Hugging Face |