> ## Documentation Index > Fetch the complete documentation index at: https://octo-98895abd.mintlify.site/llms.txt > Use this file to discover all available pages before exploring further. # fund_nav > Monthly NAV/flows series for one CNPJ, oldest first. Default window (p_to NULL) ends at the family's latest COMPLETE period per mv_period_completeness; an explicit p_to serves the window verbatim, partial months included. Row cap: more than 1000 rows RAISES 22023 (never trimmed) unless p_after pages: '' = first page, then the last row's period as 'YYYY-MM-DD'. PAGING REQUIRES p_entity_type — one CNPJ can file under two families in the same month (385 do), so a bare period is unique only within one family; whole-result mode serves both and labels each row. period keeps each family's own date convention: fi, fii and fiagro rows are dated the FIRST of the month (an fi row carries the last daily report filed in that month, so 2026-09-01 holds September's closing NAV), fidc rows the month-END (2026-08-31), fip rows 31 December of the year; the trailing period_month is the same month as api.panel keys it (first of month), equal to period for fi, fii and fiagro. Columns are per family (fact_fund_monthly arms): fi files quota, quotaholders, inflows, redemptions; fidc and fiagro file delinquency; fii files quotaholders, monthly_yield, assets; fip files nav only — a null outside that list is not applicable, not missing (catalog().applicability). fidc delinquency is filed from 2013-01, null for a fund with no tab VI row before 2020-11 and on every row from 2020-11 (regime break; catalog().regime_breaks). **Row cap.** A window producing more than 1000 rows raises SQLSTATE `22023` naming this function. Nothing is trimmed to fit — PostgREST would silently cut the response at 1000 and a cut-short series is indistinguishable from one that simply ends. This function pages: send `p_after = ""` for the first page, then the key copied from the last row; a page shorter than 1000 is the last. `p_after = null` is whole-result mode, which is what refuses. ## OpenAPI ````yaml /openapi.json post /rpc/fund_nav openapi: 3.1.0 info: title: SILO Data API version: '75' summary: >- Read-only Brazilian public financial data: CVM filings, BACEN series, B3 quotes and the B3 BDI group. description: > Schema `api` over PostgREST. Views are `GET /`; functions are `POST /rpc/` with a JSON body. **This document is generated** from the live Postgres catalog by `scripts/gen_openapi.py` — `pg_get_function_arguments` for the RPC signatures, `information_schema.columns` for the view columns, and the granted-to-`anon` privilege set for which objects exist at all. It is not hand-maintained, and `tests/test_openapi_spec.py` fails the build when it drifts from the SQL. ### Two different ceilings Reading one for the other is the most expensive mistake on this API. * **The row cap refuses.** Fifty-nine set-returning functions raise SQLSTATE `22023` when the window they were handed would produce more than 1000 rows. Nothing is trimmed, and the error says why and how to fix it (the message, plus PostgREST's `details` and `hint`). Five of them (`panel`, `quote_history`, `fund_nav`, `index_history`, `trade_consolidated_history`) take a `p_after` cursor so you can walk the series; the other fifty-four ask you to narrow the window. `fund_nav` paging additionally REQUIRES `p_entity_type` — its cursor is a bare period, which is unique only within one family, and CNPJs that file under both `fi` and `fidc` in the same month would otherwise be ambiguous. * **The tier ceiling clamps.** Three other functions (`search_funds`, `option_chain`, `option_exercises`) silently lower `p_limit` to a per-tier maximum. No error is raised, so the only way to know the result was cut is to know the ceiling. Signing in raises it. Separately from both, PostgREST's server-wide `db-max-rows = 1000` cuts every response, on every endpoint, for every tier. On the GET views `Content-Range` is the signal (`0-999/*`); `Prefer: count=exact` turns the `*` into a total. ### Retention Most of this warehouse is append-only history. The **B3 BDI group** — `short_interest`, `short_interest_by_sector`, `investor_flow`, `lending_trades`, `lending_participants` — is not: B3 keeps roughly 21 business days and publishes no archive. A missed session cannot be bought back. See Market Coverage. license: name: See repository url: https://github.com/PedroDnT/SILO-BZ servers: - url: https://zcjbtpxuhdekpwcxmepn.supabase.co/rest/v1 description: Supabase PostgREST, schema `api` security: - apikey: [] tags: - name: ANBIMA - name: Company events - name: Coverage & catalog - name: Derivatives - name: FIDC - name: FNET documents - name: Financials - name: Funds - name: Holdings - name: Inflation - name: Lookup - name: Macro - name: Macro expectations - name: Other - name: Panel - name: Portfolio - name: Quotes - name: Rates - name: Research - name: Screens - name: Securities lending - name: Short interest & flows paths: /rpc/fund_nav: post: tags: - Funds summary: fund_nav description: >- Monthly NAV/flows series for one CNPJ, oldest first. Default window (p_to NULL) ends at the family's latest COMPLETE period per mv_period_completeness; an explicit p_to serves the window verbatim, partial months included. Row cap: more than 1000 rows RAISES 22023 (never trimmed) unless p_after pages: '' = first page, then the last row's period as 'YYYY-MM-DD'. PAGING REQUIRES p_entity_type — one CNPJ can file under two families in the same month (385 do), so a bare period is unique only within one family; whole-result mode serves both and labels each row. period keeps each family's own date convention: fi, fii and fiagro rows are dated the FIRST of the month (an fi row carries the last daily report filed in that month, so 2026-09-01 holds September's closing NAV), fidc rows the month-END (2026-08-31), fip rows 31 December of the year; the trailing period_month is the same month as api.panel keys it (first of month), equal to period for fi, fii and fiagro. Columns are per family (fact_fund_monthly arms): fi files quota, quotaholders, inflows, redemptions; fidc and fiagro file delinquency; fii files quotaholders, monthly_yield, assets; fip files nav only — a null outside that list is not applicable, not missing (catalog().applicability). fidc delinquency is filed from 2013-01, null for a fund with no tab VI row before 2020-11 and on every row from 2020-11 (regime break; catalog().regime_breaks). **Row cap.** A window producing more than 1000 rows raises SQLSTATE `22023` naming this function. Nothing is trimmed to fit — PostgREST would silently cut the response at 1000 and a cut-short series is indistinguishable from one that simply ends. This function pages: send `p_after = ""` for the first page, then the key copied from the last row; a page shorter than 1000 is the last. `p_after = null` is whole-result mode, which is what refuses. operationId: rpc_fund_nav requestBody: required: true content: application/json: schema: type: object properties: p_cnpj: type: - string - 'null' p_from: type: - string - 'null' format: date description: Defaults to `'2019-01-01'::date`. default: '2019-01-01' p_to: type: - string - 'null' format: date description: Defaults to `NULL::date`. default: null p_entity_type: type: - string - 'null' description: Defaults to `NULL::text`. default: null p_after: type: - string - 'null' description: Defaults to `NULL::text`. default: null required: - p_cnpj responses: '200': description: >- The result set, oldest first. An unknown id is an empty array, not a 404 and never a guessed value. content: application/json: schema: type: array items: type: object properties: cnpj: type: - string - 'null' period: type: - string - 'null' format: date entity_type: type: - string - 'null' nav: type: - number - 'null' quota: type: - number - 'null' quotaholders: type: - integer - 'null' format: int32 delinquency: type: - number - 'null' monthly_yield: type: - number - 'null' inflows: type: - number - 'null' redemptions: type: - number - 'null' assets: type: - number - 'null' period_month: type: - string - 'null' format: date '400': description: >- Bad request. Contract violations raise SQLSTATE `22023` with a message naming the limit or the argument at fault. content: application/json: schema: $ref: '#/components/schemas/Error' '401': description: >- No usable `apikey`, or a request for a schema outside `api`. `Accept-Profile: public` and any `cvm_*` / `b3_cotahist` / `cia_*` path answer 401 for every caller, by grant — schema `api` is the entire public surface. content: application/json: schema: $ref: '#/components/schemas/Error' '404': description: No such endpoint in schema `api`. content: application/json: schema: $ref: '#/components/schemas/Error' components: schemas: Error: type: object description: >- PostgREST error body. `code` carries the SQLSTATE — `22023` is a contract refusal. properties: code: type: - string - 'null' examples: - '22023' message: type: - string - 'null' details: type: - string - 'null' hint: type: - string - 'null' securitySchemes: apikey: type: apiKey in: header name: apikey description: >- The publishable key (`sb_publishable_…`), on every request. It is NOT a JWT — sending it as `Authorization: Bearer` is rejected as an invalid JWT. It identifies the project, not you, and maps to the Postgres `anon` role. ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.