Skip to main content
Parameters, response columns and types are generated from the SQL catalog and render under Endpoints → Screens in the sidebar. This page carries what a schema cannot: what each screen measures, what it was measured against, and what else looks the same.
These are signals, not verdicts — the same contract as the forensic screens. A row is a fund whose public filing record crossed a stated threshold. It is not a score, a rating or a finding that a rule was broken. Every row carries screen and params; confirm against the fund’s own filings before repeating anything.

​
Overview

The dashboard’s /filings page calls these three functions, so the API function is still the only definition. Their defaults are published in screens in catalog(), with the allowed ranges and a meaning that says what else produces the row.

​
A fund is its CNPJ, never its name

FNET rows carry no CNPJ. A document belongs to a fund only because FNET returned it when Silo queried cnpjFundo = that CNPJ, in a sweep that reaches every FII and FIDC once a fortnight. screen_restatements and screen_late_filers read only those links. fund_name is FNET’s label, served so you can read it and never compared with anything. A document the sweep has not linked yet is not in either screen.

​
Restatements

Defaults: the last 12 months of delivery days, at least 3 re-filings, at least 20% of the fund’s documents. restatements_re (Reapresentação Espontânea, voluntary) and restatements_rc (Reapresentação por Exigência, required by CVM) split the count as FNET publishes modalidade.
The same row also comes from routine typo corrections, an administrator or custodian migration that re-submits a whole book, the resolution-175 adaptation, a FNET template change that forced a re-submission, one error carried through consecutive informes, or a CVM supervision sweep across one administrator’s funds. An RC says CVM asked; it does not say what was wrong. Pair a row with fund_restatements to see each re-filing and the version it replaced.

​
Late filers: measured against a cited rule

“Late” here is a delivery timestamp compared with a deadline Silo can cite: The citation rides on every row as deadline_rule. Months before a family’s adaptation deadline are not measured, because the predecessor instructions’ deadlines are not cited here; a window that ends before 2024-12 raises 22023. What a row measures:
  • the first delivery of an original (versao 1) Informe Mensal Estruturado for each reference month — a restatement never makes a month late;
  • a month is late when its delivery day minus (the month’s last day + 15) is at least p_min_days_late; the lag is delivery day minus the month’s last day;
  • informes is the months measured, informes_late the late ones, max_days_late the worst, median_lag_days the fund’s typical lag.
What it cannot see:
  • Calendar days, no holiday calendar. The text says dias, not dias úteis. A deadline on a weekend or holiday may legitimately roll forward, which is what the default p_min_days_late of 5 absorbs.
  • Extensions. CVM can grant them; the register does not say so.
  • Missing months. A month with no informe in the register is not counted late. FNET history is still being backfilled, and absence is the next screen’s question.
  • A fund with several classes is measured on its earliest filing for the month.

​
Silent filers: read from CVM, not FNET

screen_silent_filers answers absence from CVM’s own tables, which have deep history: the last period each fund filed — the informe diário for FI, the monthly informe for FIDC, FII and FIAGRO — against that family’s newest complete period (coverage().complete_through), never today, so a month CVM has not published yet is not silence. Only funds whose registry row is active are listed; registry_status is served as filed. Context columns, not criteria: registry_nav / registry_nav_date (the NAV the registry itself carries, with its date) and fnet_last_delivered_at, the newest FNET document linked to the CNPJ (FII and FIDC only) — a fund silent at CVM but still delivering to FNET. The same row also comes from a fund merged, incorporated or liquidated whose status CVM has not updated yet; reporting moved to a class CNPJ other than the fund’s by the resolution-175 adaptation; CVM’s dataset lagging the filing; or a Silo ingest gap — check coverage() landed_at before reading a family-wide silence. FIP files annually and is not screened.

​
Too many rows

Screens do not page. Above one 1,000-row page they raise 22023 (Conventions → Row caps). Narrow with the output filter (p_family, p_modalidade) or the thresholds; the silent screen over every family at once is the likeliest to need p_family.