ruthless-efficiency
Copyright (c) 2026 Karsten S. Nielsen

This product is original work, released under the MIT License (see LICENSE).
It is not a fork. The name and tagline are a homage to Monty Python (see
"Assets and Cultural Attribution" below).


Third-Party Libraries (runtime)
-------------------------------

Installed with the default distribution. Each retains its own upstream license;
the SPDX expressions below were verified against the installed package metadata.

pydantic --- data validation / settings (MIT License).
  See: https://github.com/pydantic/pydantic

numpy --- numerical arrays and the seeded RNG used by RandomSearchStrategy
  (BSD-3-Clause License). Copyright (c) 2005-2024, NumPy Developers.
  Pinned `>=1.24,<3` so the random-search RNG stream stays stable for the
  determinism gate. See: https://github.com/numpy/numpy

PyYAML --- YAML config parsing, via `yaml.safe_load` only (MIT License).
  See: https://github.com/yaml/pyyaml


Development and Build Tooling
-----------------------------

Used to test, lint, type-check, and build the project; NOT redistributed in the
runtime wheel.

pytest (MIT) --- test runner.
hypothesis (MPL-2.0) --- property-based testing.
ruff (MIT) --- linter and formatter.
pyright (MIT) --- static type checker.
import-linter (BSD-2-Clause) --- enforces the hexagonal dependency contracts
  declared in `.importlinter`.
hatchling (MIT) --- PEP 517 build backend.


Optional Extras (not installed by default)
------------------------------------------

Declared in pyproject.toml under [project.optional-dependencies] and pulled in
only when the matching extra is requested. Each retains its upstream license;
none is bundled in the default distribution. These are integration points for
later phases and are not consumed by any Phase 1A public API.

optuna (MIT) --- `[optuna]`, Phase 2 (OptunaStrategy).
openevolve --- `[evolve]`, Plan 1B (the evolve strategy is our orchestration
  over OpenEvolve's loop).
paramiko, huggingface_hub, docker --- `[backends]`, Plan 1B (SSH / Hugging Face
  Jobs / Docker compute backends).


Mathematical / Methodological References
----------------------------------------

The built-in random-search strategy (ruthless/strategies/random_/) implements
the baseline described in: Bergstra, J., & Bengio, Y. (2012). "Random Search
for Hyper-Parameter Optimization." Journal of Machine Learning Research, 13,
281-305.

The log-uniform sampling option (the `log` flag on FloatRange in
ruthless/config.py) follows the convention popularised by Optuna's
`suggest_float(log=True)`: Akiba, T., Sano, S., Yanase, T., Ohta, T., &
Koyama, M. (2019). "Optuna: A Next-generation Hyperparameter Optimization
Framework." Proc. KDD '19. (Optuna itself is an optional Phase 2 extra, not a
Phase 1A dependency.)

The ports-and-adapters ("hexagonal") structure of the core follows: Cockburn,
A. (2005). "Hexagonal Architecture" (alistair.cockburn.us/hexagonal-architecture).
The one-way core/strategy/backend dependency direction is enforced in CI by
import-linter.


Documentation Tooling
---------------------

The C4 architecture diagram (docs/c4/architecture.html) is authored in
Structurizr DSL (docs/c4/architecture.dsl) and rendered locally via Structurizr
and PlantUML. These tools are used only to produce the documentation artifact
and are not part of the project's dependencies.


Assets and Cultural Attribution
-------------------------------

The hero image (assets/ruthless-efficiency.jpg) and the project name/tagline
("Our chief weapons are Ruthless Efficiency!") are an affectionate homage to
Monty Python's "The Spanish Inquisition" sketch (Monty Python's Flying Circus,
1970). No affiliation with or endorsement by the rights holders is implied.
