Our daily update pipeline is live, and the API now delivers data at daily granularity.
What changed
Data from 1 August 2026 onwards is served as one Parquet file per calendar day. Everything up to and including 31 July 2026 stays as it is — one file per calendar month.
The changeover falls on a month boundary, so any given month is served entirely one way or the other. A date range that spans it returns the earlier months as monthly files, then a daily file for each day from 1 August.
What you need to do
Nothing. Request parameters are unchanged — you still pass start_date and end_date, and the API returns every file covering that range, in chronological order. The Python SDK handles the mix transparently and still hands you a single DataFrame.
If you consume the REST API directly, note that entries in files now carry an optional day field:
{
"files": [
{ "year": 2026, "month": 7, "url": "..." },
{ "year": 2026, "month": 8, "day": 1, "url": "..." },
{ "year": 2026, "month": 8, "day": 2, "url": "..." }
]
}
day is present only on daily files. Monthly files omit it entirely, so responses for older ranges are unchanged. If you key or sort on year/month alone, include day so same-month files stay in order.
Why
Daily files mean new data lands sooner after the exchange day closes, and a re-run only has to reprocess a single day rather than a whole month.