Preview, not deployed. Point in time, survivorship, price adjustment, reproducibility and a citation line, in the v2 gateway.
Preview, not deployed. The v2 gateway is built but not deployed yet. This page describes it as built.
Five things decide whether a result from SILO can be trusted in a paper or a backtest. Each one below names the parameter that controls it and its default. The calls use the base URL from the v2 quickstart:
With as_of, only documents CVM received before that date are read, each at the highest version known then. as_of is a hidden parameter: it works but is not in the main table. It does not apply to kind=versions, which lists every version on purpose.
The listed shares and units that traded on that date: the ticker and ISIN pairs whose first and last observed sessions span it, dead names included. Build a sample from this list, not from today’s tickers. With no as_of, the list holds every pair the B3 tape has seen since 2019-01-02.
prices defaults to adjust=split: the close adjusted for splits, groupings and bonus shares. It is not adjusted for dividends. No dividend-adjusted series exists in v2.
curl "$SILO_V2/prices?id=PETR4&adjust=none"
adjust=none is the raw close as traded. Every row carries adjusted (true or false), so the two never mix silently. Only kind=history is adjusted; latest, session, index and forward are never adjusted.
Every response echoes, in meta.request, every parameter the resource takes with the value the gateway applied, defaults included. Paste it into a methods section and the call can be replayed exactly. meta.endpoint names the api.* endpoint that answered.meta.data_version is planned in the spec but not built yet. Until it is, record the data version from coverage:
curl "$SILO_V2/coverage?dataset=quotes"
Its landed_at and landed_git_sha say when the data you read was loaded, and by which version of the pipeline. data_through is the newest elapsed period, and complete_through is how much the source has fully published.