What is hyperliquid-archive-notes?

This repository is a detailed set of notes about a specific, tricky data source: the free historical market data published by Hyperliquid, a crypto exchange for perpetual futures. Hyperliquid publishes minute-by-minute price and funding data going back to 2023, which is more generous than what most paid data vendors offer. The catch, according to the author, is that this data is very easy to read incorrectly, producing numbers that look perfectly normal but are actually wrong, with no error, no missing column, and nothing that would obviously warn you. The README explains that the author has hit many of these silent mistakes personally, and this project is an attempt to document every one of them so other people do not repeat them. Each problem listed comes with a small script in a probes folder that proves the claim using the real data, so readers do not have to take the author's word for it. The notes cover two different ways to pull data from Hyperliquid, a free info API for current market snapshots and settled funding rates, and a paid archive stored on Amazon S3 that holds the full per-minute history for around a dollar total. The traps described include things like a funding rate calculation that silently changed units partway through history, two different data sources that flip to a new format on two different timestamps, a clamp value in a formula that changed without being announced, and absent data being written as a plain zero instead of being left blank, which can make a dead market look like a real one with no activity. Other issues mentioned include delisted markets that keep reporting a funding rate forever, an API that quietly cuts results off at 500 rows without saying so, and index numbers in one data set that do not match their actual position in the list. Each trap is explained with the exact date or condition where it appears and why a normal check would miss it. This is not a general-purpose library. It is a reference document and verification toolkit for anyone building trading or research systems on top of Hyperliquid's historical data, aimed at catching mistakes before they quietly corrupt an analysis.