Raw parquet downloads
Raw parquet files — served from Cloudflare R2. Readable directly with Polars / DuckDB / pandas.
| Path | Description | Size | |
|---|---|---|---|
silver/weather_hourly.parquet | 5-year hourly weather (NASA POWER + Senai METAR), 43,848 rows, 20 cols. UTC + MYT timestamps, t_canonical, GHI/DNI/DHI, wind, RH. | 1.2 MB | Download |
gold/load_hourly_state.parquet | Synthesised Johor hourly load (4.5 yr), 39,424 rows. DOSM-anchored monthly + ETOU peak share, with cooling-CDH temperature term. | 772 KB | Download |
Read with Polars
import polars as pl
BASE = 'https://jb-vpp-research.pages.dev/api/download?path='
w = pl.read_parquet(BASE + 'silver/weather_hourly.parquet')
print(w.head())
l = pl.read_parquet(BASE + 'gold/load_hourly_state.parquet')
print(l.describe())
- Files are served from a Cloudflare R2 bucket (
jb-vpp-raw) via a Pages Function at/api/raw/<path>. - Only paths under
silver/andgold/are exposed; bronze remains internal until provenance metadata is published. - The site itself queries D1 (faster, paginated) — these downloads are for offline / notebook / pipeline integration.