tesouropy

📖 Documentation & guides: https://strategicprojects.github.io/tesouropy/

Access Brazilian National Treasury open-data APIs from Python.

tesouropy is a Python port of the tesouror R package. It gives you a single, consistent interface to six open-data sources of the Brazilian National Treasury (Tesouro Nacional) and related federal government APIs, returning tidy polars DataFrames.

Domain	What it covers	API style
SICONFI	Fiscal reports — RREO, RGF, DCA, MSC — and entity info	ORDS (hasMore/offset pagination)
CUSTOS	Federal government cost data	ORDS
SADIPEM	Public debt and credit operations (PVL)	ORDS
Transferências Constitucionais	Constitutional transfers to states/municipalities	Simple JSON
SIORG	Federal organizational structure	Simple JSON
SIOPE	Education spending data (FNDE/MEC)	OData

Features: automatic pagination, in-memory caching, retry with backoff, fault-tolerant state-wide loops (partial results instead of hard failures), and tidy snake_case output.

Installation

pip install tesouropy

Requires Python 3.9+. Runtime dependencies: polars and requests.

Bilingual interface (PT/EN)

Almost every function has two names: a Portuguese name with Portuguese parameters, and an English alias that maps English parameter names onto the Portuguese ones. Pick whichever you prefer — they return identical data.

import tesouropy as tn

# Portuguese
rreo = tn.get_rreo(
    an_exercicio=2022, nr_periodo=6,
    co_tipo_demonstrativo="RREO", no_anexo="RREO-Anexo 01",
    co_esfera="E", id_ente=17,
)

# English (same call)
rreo = tn.get_budget_report(
    fiscal_year=2022, period=6,
    report_type="RREO", appendix="RREO-Anexo 01",
    sphere="E", entity_id=17,
)

Quick start

import tesouropy as tn

# List all government entities (states, municipalities, DF)
entes = tn.get_entes()

# Fiscal Management Report (RGF) for a state
rgf = tn.get_rgf(
    an_exercicio=2022, in_periodicidade="Q", nr_periodo=3,
    co_tipo_demonstrativo="RGF", no_anexo="RGF-Anexo 01",
    co_esfera="E", co_poder="E", id_ente=17,
)

# Public debt verification requests for a state, then a detail query
pvl = tn.get_pvl(uf="PE")
schedule = tn.get_opc_cronograma_pagamentos(id_pleito=pvl["id_pleito"][0])

# Education spending (SIOPE), filtered server-side for speed
recife = tn.get_siope_dados_gerais(
    ano=2023, periodo=6, uf="PE", filter="NOM_MUNI eq 'Recife'",
)

# Federal cost data (CUSTOS) — always filter; SIORG codes auto-padded
custos = tn.get_custos_pessoal_ativo(ano=2023, organizacao_n1=244)

State-wide panels (fault tolerant)

The *_for_state helpers fetch data for every municipality of a state, looping with fault tolerance: if a municipality call fails after all retries, the failure is recorded and the loop continues.

rreo_es = tn.get_rreo_for_state(
    state_uf="ES", an_exercicio=2021, nr_periodo=6,
    co_tipo_demonstrativo="RREO", no_anexo="RREO-Anexo 01",
)

# Inspect partial failures / no-data municipalities
getattr(rreo_es, "failed", None)    # DataFrame of failed calls, if any
getattr(rreo_es, "no_data", None)   # municipalities that returned 0 rows

Similarly, ORDS pagination is fault tolerant: if a page after the first fails, you get a partial DataFrame with result.partial == True and result.last_page_error set, rather than losing everything fetched so far.

Reconciling RREO layout drift across years

SICONFI relabels RREO appendices and accounts over time. tidy_rreo() uses a bundled layout table to assemble a coherent series across years:

import polars as pl

frames = []
for yr in range(2019, 2024):
    rule = tn.rreo_layout().filter(
        (pl.col("topic") == "previdencia") & (pl.col("regime") == "rgps")
        & (pl.col("first_year") <= yr) & (pl.col("last_year") >= yr)
    )
    frames.append(tn.get_rreo(
        an_exercicio=yr, nr_periodo=6, co_tipo_demonstrativo="RREO",
        no_anexo=rule["no_anexo"][0], co_esfera="U", id_ente=1,
    ))
rreo = pl.concat(frames, how="diagonal_relaxed")
serie = tn.tidy_rreo(rreo, topic="previdencia", regime="rgps")

Caching, retries and logging

• Cache: every HTTP request is cached in memory by default (use_cache=True). Clear it with tn.tesouropy_clear_cache().

• Retries: 5 attempts with progressive backoff (3/6/9/12s) on HTTP 429/5xx and connection failures.

• Logging: the package logs progress through the "tesouropy" logger with a NullHandler, so it is silent by default. Enable it with:

  import logging
  logging.basicConfig(level=logging.INFO)

  Pass verbose=True (or call tn.set_verbose(True)) to log the full request URL of each call — handy for debugging.

Relationship to the R package

tesouropy mirrors the public API of tesouror (CRAN). The function names, parameters, endpoints, pagination and fault-tolerance behaviour are kept in sync. The main idiomatic differences: R tibbles become polars DataFrames, and R attributes (attr(x, "partial"), attr(x, "failed")) become instance attributes (x.partial, x.failed).

License

MIT © tesouropy authors. See LICENSE.