bnetza_bk6_scraper

bnetza_bk6_scraper mirrors the documents published by the German Bundesnetzagentur (BNetzA) Beschlusskammer 6 (BK6) into a structured, git-diffable directory tree. BK6 regulates electricity network access and is a constant source of consultations, rulings (Festlegungen) and their attachments. Because the agency publishes these as loose PDFs on HTML pages with no changelog, tracking what changed and when is painful. This tool discovers every BK6 proceeding, downloads its PDFs and a normalized HTML snapshot of each phase page, and records structured metadata. Committing the output to git turns every regulatory update into a reviewable diff.

Installation

pip install bnetza_bk6_scraper

Usage

The package installs a single console command, bnetza-bk6-scraper, with a mirror subcommand:

bnetza-bk6-scraper mirror --target <dir> [--concurrency N] [--year YYYY] [-v]

Example — mirror only the 2023 proceedings into ./mirror:

bnetza-bk6-scraper mirror --target ./mirror --year 2023 -v

Each run logs a summary such as run summary: 7 proceedings, 16 documents written, 0 failures.

Scoped mirroring (Python API)

The mirror CLI mirrors all proceedings. To mirror only a curated subset (e.g. the electricity GPKE/WiM/MaBiS Prozessdokumente), use the Python API: give the scraper a list of seed pages to crawl and a list of predicates (OR semantics — a document is downloaded if any predicate returns True). Predicates receive a CandidateDocument.

import asyncio
from bnetza_bk6_scraper import BnetzaBk6Scraper, CandidateDocument

GPKE = "https://www.bundesnetzagentur.de/DE/Beschlusskammern/BK06/BK6_83_Zug_Mess/831_gpke/gpke_node.html"

def is_prozessdokument_lesefassung(c: CandidateDocument) -> bool:
    name = c.filename.lower()
    return any(fw in name for fw in ("_gpke_", "_wim_", "_mabis_")) and "lesefassung" in name

asyncio.run(
    BnetzaBk6Scraper().mirror_seeds(
        target_dir=".", seeds=[GPKE], keep=[is_prozessdokument_lesefassung]
    )
)

Documents that carry an Aktenzeichen are written under <year>/<aktenzeichen>/; documents without one (e.g. the PID-Liste in the Datenformate tree) go under _other/…. A root manifest.json lists the kept documents.

Link detection is not limited to PDFs: .pdf, .xlsx and .xls are all mirrored. BNetzA publishes the Liste der Prüfidentifikatoren (PID) as both a PDF and a machine-friendly Excel file — the Excel is captured too, so a predicate like the one above picks up PID_..._info_....xlsx right next to its PDF.

Output layout

Proceedings are written under /{year}/{aktenzeichen}/, with a top-level index.json listing every mirrored proceeding:

<target>/
├── index.json                          # summary of all proceedings
└── 2023/
    └── BK6-23-241/
        ├── metadata.json               # structured proceeding metadata
        ├── BK6-23-241_beschluss.html   # normalized HTML snapshot of a phase page
        ├── BK6-23-241_beschluss_vom_07.05.26.pdf
        ├── BK6-23-241_bilarem.pdf
        └── BK6-23-241_anlage_bilarem.pdf

• metadata.json captures the Aktenzeichen, year, title, status, Stand (last-modified date), any submission deadline (Frist), the phase pages, and one entry per document (title, type, source URL, filename).
• The normalized *.html files are trimmed, stable snapshots of the source phase pages so that content changes surface as small diffs.
• The PDFs are the proceeding's documents, downloaded verbatim.

Change detection is intentionally "dumb": the tool always writes the current state, and git diff in the mirror repository reveals what changed.

Mirror repository

The scraper is designed to feed a separate mirror repository, Hochfrequenz/bnetza_bk6_mirror. A scheduled GitHub Action there will periodically:

pip install bnetza_bk6_scraper
bnetza-bk6-scraper mirror --target .
git add -A && git commit -m "update BK6 mirror"

so that regulatory changes at BK6 become visible as reviewable git diffs and commit history. That Action is future work and does not live in this repository.

WAF / browser User-Agent

The BNetzA website sits behind a Web Application Firewall that rejects non-browser clients by serving a 200 OK "The requested URL was rejected" page instead of the real content. To get through, the scraper sends browser-like User-Agent and Accept headers and treats the rejection page as a retryable error. No credentials or API keys are required.

Contribute

This project uses tox for all quality gates. Create a one-shot development environment with everything installed:

tox -e dev

Individual gates: tox -e tests, tox -e linting, tox -e type_check, tox -e coverage, and tox -e spell_check. Run the full suite with tox.