Управление конфигурацией: JSONSchema, Pydantic и YAML

engineering

config-skills is a Claude Code skill that documents patterns and best practices for the LlamaFarm configuration module. It covers the full pipeline: defining JSONSchema with `$ref` references in `schema.yaml`, dereferencing them via `jsonref` into `schema.deref.yaml`, generating Pydantic v2 models with `datamodel-codegen`, and writing custom cross-field validators in `validators.py` for constraints that JSONSchema draft-07 cannot express. Comment-preserving YAML read/write relies on `ruamel.yaml`. The module consists of seven key files — `datamodel.py`, `schema.yaml`, `compile_schema.py`, `generate_types.py`, `validators.py`, `helpers/loader.py`, and `helpers/generator.py` — and supports YAML, TOML, and JSON formats. Aimed at developers building a robust schema-validated, type-safe configuration pipeline in Python 3.11+.