BeddrokClient

Python client for the Beddrok API: personnummer in, stable numbers out.

Installation

pip install BeddrokClient

Usage

from beddrokclient import BeddrokClient

with BeddrokClient(
    user="cint-clean",
    password="...",
    api_url="http://beddrok-api.rst-vdp-lake-beddrok-prod.svc:8000",
) as beddrok:
    pseudonyms = beddrok.pseudonymize(["199204062355", "920406-2355"])
    pseudonyms["199204062355"].number
    pseudonyms["199204062355"].removed

    resolved = beddrok.resolve([12345])
    resolved[12345].pin
    resolved[12345].status

What it does for you

• Logs in through POST /auth/token and renews the token 5 minutes before it expires, reading the expiry from the JWT. On a 401 it logs in again once and retries.
• Splits input into batches: 1000 pins per /pseudonymize call, 20 numbers per /resolve call, the API's default limits. Duplicates are sent once.
• Retries connection errors and 503 with exponential backoff, backoff_base ** attempt seconds, up to max_retries.

What it never does

• It never changes a pin. Pins are sent exactly as given. Beddrok gives 920406-2355 and 199204062355 two different numbers, so pass values exactly as the source has them.
• It never returns a partial answer. If any pin or number is missing from the response, it raises BeddrokError. It never falls back to the original value.
• It never puts a personnummer in an error. BeddrokError carries path, status and the HTTP reason only. Response bodies are never included, because a 422 echoes the request.

Configuration

BeddrokClient(
    user="...",
    password="...",
    api_url="...",
    timeout=30,
    max_retries=3,
    backoff_base=2.0,
    pseudonymize_batch_size=1000,
    resolve_batch_size=20,
    refresh_token_before_seconds=300,
)

Upgrading from 2.x

3.0 targets the Beddrok API as it is deployed. The 2.x methods called endpoints that API does not have (/encrypt, /decrypt, /validate) and have been removed: encrypt_pins, decrypt_pins, validate_pins, and the pandas and polars DataFrame helpers. Use pseudonymize and resolve instead. pandas, polars and roskarl are no longer dependencies.

Development

pip install -e ".[dev]"
ruff check . && ruff format --check .
pyright
pytest

The tests run against a local fake Beddrok server, no cluster needed.