Skip to main content
Parameters, response columns and types are generated from the SQL catalog and render under Endpoints in the sidebar. This page carries what a schema cannot: what a row means and how far to trust it. Check the live catalog() before using either function: a merge alone does not apply analytical SQL.

​
Overview

  • credit_market_history(p_code, p_from, p_to, p_as_of) — the market observations B3 published for one debenture code, as the warehouse held them at p_as_of (default: now).
  • credit_implied_rate(p_code, p_from, p_to) — the yield calculated from each traded average PU of one code. It is not an observation: is_calculated is always true.
TEST01 is an illustrative code, not a claim of stored observations.

​
credit_market_history

One row per (instrument_code, trade_date, settlement_date, trade_classification), with the nine stored metrics side by side: quantity, min_price, avg_price, max_price, last_price, reference_price, trade_count, volume_brl and oscillation_pct. units gives the unit of each. NULL means unpublished.
  • Knowledge time. The function picks the latest complete, successfully reconciled capture per trade date, before it filters the code. A newer complete capture that drops a code makes the code absent. This is what the warehouse knew at p_as_of, not proof of what was public at the time.
  • No mixing. Prices and economic groups are never combined. last_price and reference_price repeat instrument-wide values: do not sum them or weight them by volume. A reference price can be modeled, and never replaces a traded price. Volume is not outstanding principal.
  • Provenance on every row: ISIN, issuer name as reported, source identifier, capture ID, raw payload hash, row hash, observed_at and available_at.
  • Errors. An unknown code at the cutoff raises 22023 (HTTP 404). A known code with an empty window returns zero rows. More than 1,000 groups raises 22023 (HTTP 400); narrow the window. There is no cursor, because several groups can share a date.

​
credit_implied_rate

One row per (trade_date, settlement_date, trade_classification, method_version), oldest first. implied_rate_pct and rate_basis give the yield: pre_aa, ipca_real_aa and di_spread_aa are percent a year, pct_di is percent of DI. The row also carries the contracted contract_rate_pct, the PU, quantity, volume and trade count it used, flows_used, b3_capture_id and computed_at.
  • What it is. The yield of an average PU, read as a dirty price. It is not a traded or contracted rate. A trade group the method cannot price has no row. method_version says how: v1 uses 252 business days, the ANBIMA holiday table and a flat-forward pre curve from B3 TaxaSwap.
  • How far to trust a row. Three tags: confidence (checked, divergent or unchecked), outlier and thin_trade. checked means within 1 pp of ANBIMA’s indicative rate for the same code and day, divergent means beyond that. unchecked means no comparable rate existed, so no check was possible; it does not mean the row is good. outlier means divergent by more than 5 pp. thin_trade means volume under BRL 20 thousand.
  • ANBIMA’s own rate and the gap to it are not served.
  • Errors. An unknown code raises 22023. More than 1,000 rows raises 22023; narrow the window.