solyanka 0.0.1

Transaction pattern utilities and dataset for statement generators

Solyanka

Shared helpers and datasets that power the statement generators and downstream synthetic-data pipelines (including LLM fine-tuning sets). The first module exported by this repository focuses on transaction patterns: the curated YAML files, the schema that keeps them in shape, and a tiny loader to consume them from any application (FastAPI, CLI tools, notebooks, etc).

Installation

pip install solyanka

For local development (tests + linters):

pip install -e ".[dev]"

Usage

from solyanka import PatternsService, EEA_COUNTRIES

service = PatternsService()
general = service.load_general_patterns()
eea = service.load_eea_patterns()
thailand_specific = service.load_country_patterns("Thailand")

# Preferred helper: auto-mix general + (eea) + country overrides
full_set = service.get_country_patterns("Germany")

# Advanced filtering (e.g. validation scripts)
custom = service.get_patterns(country="Germany", include="general,eea")

# API-friendly dicts
payload = service.get_pattern_dicts(country="Spain")

If you need to point the loader to different files (e.g. while editing YAML locally), either pass PatternsService(base_dir=Path("./transaction_patterns")) or set the TRANSACTION_PATTERNS_DIR environment variable.

Pattern sets

• General – always included via load_general_patterns().
• EEA – supplement auto-applied for EEA countries via load_eea_patterns().
• Country-specific – call load_country_patterns("Germany"); the helper normalizes slugs internally.

get_country_patterns(country) mixes general + (EEA when applicable) + country-level patterns and accepts empty/None values. get_patterns(country, include) exposes finer control when tooling needs only certain slices. include accepts a comma-separated subset of {"general", "eea", "country"}; invalid values raise ValueError. get_pattern_dicts(...) mirrors these parameters but returns normalized dictionaries that are ready for JSON responses.

Development

• pytest validates every YAML file against schema.json.
• GitHub Actions run tests on push/pull_request and build/publish artifacts when a semver tag is pushed.

Contributions should keep the dataset human-friendly: no generated UUIDs in the patterns, clear field names, and comments that help reviewers understand why a pattern exists.

Tests

poetry run pytest