imibare

Programmatic access to official government statistical data.

imibare gives you clean, typed, versioned access to official statistics straight from the institutions that publish them, as pandas or polars DataFrames. No scraping, no PDF wrangling.

Coverage today is Rwanda and Kenya: NISR, BNR, MINECOFIN, RSSB, RRA, RURA, RDB and NAEB in Rwanda, KNBS and CBK in Kenya. Every dataset ID is country-prefixed (rw.nisr.cpi.monthly, ke.knbs.cpi.monthly), so further countries slot in without breaking changes.

Install

pip install imibare

Optional extras:

pip install imibare[polars]    # return polars DataFrames
pip install imibare[iceberg]   # snapshot time travel via load(version=...)

Quick start

import imibare as imi

# Browse the catalog (works offline)
for d in imi.catalog(country="RW")[:5]:
    print(d.id, "-", d.name)

# Load a dataset as a pandas DataFrame
df = imi.load("rw.nisr.cpi.monthly")

# Filter by date range
recent = imi.load("rw.nisr.cpi.monthly", start="2023-01-01")

# Kenya works the same way
ke_cpi = imi.load("ke.knbs.cpi.monthly")

# Return a polars DataFrame (requires the [polars] extra)
fx = imi.load("rw.bnr.fx.daily", engine="polars")

Catalog

catalog() returns dataset metadata you can filter by country, topic, or frequency:

imi.catalog()                             # every dataset
imi.catalog(country="KE")                 # Kenya only
imi.catalog(country="RW", topic="prices")
imi.catalog(frequency="daily")

Each entry carries its id, producing institution, coverage window, columns, source URL, and license.

Dataset IDs

Every dataset has a stable, country-prefixed id of four segments:

{country}.{institution}.{topic}.{frequency}
rw.nisr.cpi.monthly

Ids are permanent; a breaking schema change ships under a new suffix (for example rw.nisr.cpi.monthly.v2).

Versioned data (time travel)

With the [iceberg] extra you can load a dataset as it stood on a given date:

imi.load("rw.nisr.cpi.monthly", version="2025-06-01")

Data access

Both catalog() and load() work with no setup. catalog() reads metadata bundled in the package; load() fetches the latest data from the public imibare API. The versioned load(version=...) path additionally requires the [iceberg] extra and R2 catalog credentials.

load() caches each download under ~/.imibare/cache/. A cached file is used for up to an hour, then checked against the API, and downloaded again only if the data has changed. Pass force_download=True to skip the cache.

Links

• Website: https://imibare.org
• Documentation: https://imibare.org/docs

License

Code is released under the MIT License. Data retrieved through this package remains under the license of each source institution; see each dataset's metadata for details.