Parquet archive
Every Hyperliquid fill and every 5-minute position snapshot as one Parquet file per UTC day, plus a latest positions file replaced every 5 minutes.
The Hyperliquid file archive is bulk history as Parquet: one file per UTC day for each of two datasets, and one
latest positions file. There is no folder to browse. /archive/datasets
lists every file that exists, and /archive/url answers with a temporary link
to one of them. The one open folder is the free samples; /archive/datasets returns its address as samples_base.
| File | What it holds | First day |
|---|---|---|
fills, one per UTC day | Every fill of every wallet that day. | 2026-05-01 |
positions, one per UTC day | Every 5-minute snapshot of every open position that day. | 2026-05-08, a partial day |
latest/positions.parquet | The current open positions of every asset, replaced every 5 minutes. |
Where to start
- Download the Hyperliquid file archive: ask for a link, download one whole file, read it with DuckDB, pyarrow or pandas, and the traps that cost a wrong number.
- Archive file index: every day with its rows and bytes, and the column schema of both datasets.
- Archive download link: one temporary link to one file. It needs no key and works for 300 seconds.