Archive download link

GET https://data.vike.io/v1/hyperliquid/archive/url?dataset={dataset}&date={date}

Asks for a temporary link to one Hyperliquid archive file. No daily or latest file has a permanent URL and their folder cannot be listed, key or no key, so this is the only way to get one. Ask which days exist at /archive/datasets; ask here for the one you want. The free samples are the one part of the archive with a plain address, the samples_base of that index (Free samples).

The answer carries the link, when it stops being accepted, and the size of the file it names. dataset is fills or positions; date is a published UTC day, or latest for the 5-minute positions file.

# a link to one day of fills: the url is good for 300 seconds and needs no key
curl -s "https://data.vike.io/v1/hyperliquid/archive/url?dataset=fills&date=2026-10-19" \
  -H "X-API-KEY: $VIKE_API_KEY"

# one day of 5-minute position snapshots: 1.6 GB, so read bytes before you download
curl -s "https://data.vike.io/v1/hyperliquid/archive/url?dataset=positions&date=2026-10-19" \
  -H "X-API-KEY: $VIKE_API_KEY"

# the current open positions of every asset, a small file replaced every 5 minutes
curl -s "https://data.vike.io/v1/hyperliquid/archive/url?dataset=positions&date=latest" \
  -H "X-API-KEY: $VIKE_API_KEY"

# ask and download in one step: redirect=1 answers 302 with the link in Location, and -L follows it
curl -s -L "https://data.vike.io/v1/hyperliquid/archive/url?dataset=positions&date=latest&redirect=1" \
  -H "X-API-KEY: $VIKE_API_KEY" \
  -o positions-latest.parquet

# keep only the link, to hand to a downloader
curl -s "https://data.vike.io/v1/hyperliquid/archive/url?dataset=fills&date=2026-10-19" \
  -H "X-API-KEY: $VIKE_API_KEY" \
  | jq -r '.url'

These are the three links, captured from the running service. In each url the token between /dl/ and the next / was replaced with <exp>.<kid>.<sig>, because a real signed link is a credential for five minutes; every other value is as returned.

{
  "url": "https://data.vike.io/dl/<exp>.<kid>.<sig>/venue=hyperliquid/date=2026-10-19/fills.parquet",
  "expires_at": "2026-10-21T10:24:43Z",
  "expires_in": 300,
  "dataset": "fills",
  "date": "2026-10-19",
  "bytes": 358509915,
  "rows": 12022655
}

{
  "url": "https://data.vike.io/dl/<exp>.<kid>.<sig>/venue=hyperliquid/date=2026-10-19/positions.parquet",
  "expires_at": "2026-10-21T10:24:43Z",
  "expires_in": 300,
  "dataset": "positions",
  "date": "2026-10-19",
  "bytes": 1639446104,
  "rows": 102738790
}

{
  "url": "https://data.vike.io/dl/<exp>.<kid>.<sig>/venue=hyperliquid/latest/positions.parquet",
  "expires_at": "2026-10-21T10:24:43Z",
  "expires_in": 300,
  "dataset": "positions",
  "date": "latest",
  "bytes": 29578173,
  "rows": 427884
}

bytes and rows are the file's size and row count, so you can budget the download first. For latest they are the values of the version that existed when you asked; the file is replaced about every 5 minutes, so the bytes you receive can differ.

The link

The url is an ordinary HTTPS address:

• It needs no key. The link is the credential. Anyone holding it can download that one file until it expires, so keep it out of logs and shared tickets. Downloads are recorded against the key that asked for the link.
• It names exactly one file. The signature covers the path: changing the date or the token returns 403, and a path that matches no archive file returns 404.
• It is accepted for 300 seconds (expires_in) from the start of a request. A download that is running when the clock passes expires_at is not cut. A reader that opens new requests, such as DuckDB or pyarrow reading over HTTP, needs a link that is still valid for each one, so download the whole file, or ask again per file.
• A file is served whole or in ranges. Responses carry Accept-Ranges: bytes, Content-Length and Cache-Control: private, no-store, so curl -C - can resume an interrupted download (with a fresh link).

What it costs

The call is a normal /v1 request: 1 credit, one request of the hourly budget, and counted by the per-minute /v1 cap. The download itself is recorded as an archive_download and spends neither the hourly budget nor the per-minute cap; its cost is bandwidth, capped per connection like the rest of the archive.

There is an allowance of 20 GB of links per user per rolling 24 hours, all of a user's keys together. What is counted is the size (bytes) of each link issued, not the bytes you read, so a link you never use still counts. Past it the call answers 429 with {"detail": "..."} saying how much was issued. Unlike the rate-limit 429s it carries no Retry-After: the window rolls over 24 hours, so a retry minutes later is refused again. Ask again after earlier links have aged out of the window, or write to us if you need more.

Related

• /archive/datasets — which days exist, their rows and bytes, and the columns.
• Download the Hyperliquid file archive — download a file with curl or Python and read it.
• Rate limits — what a link request and a download each spend.