# Python binding

The `n4m` package is a ctypes wrapper over **libn4m (ABI 2.0)**. It exposes ABI
introspection, the context / config lifecycles, and the full method surface
through role subpackages that mirror the `n4m.<role>` namespace. See
`bindings/python/README.md` for installation and loader rules, and the
[ABI 2.0 migration guide](../MIGRATION_ABI2.md) for the old→new mapping.

## Hello-version

```python
import n4m

print(n4m.abi_version())          # (2, 0, 0)

with n4m.Context() as ctx:
    ctx.seed = 42
    assert ctx.seed == 42
```

## Role packages (ABI 2.0)

`n4m.__init__` exposes only metadata/helpers and the role subpackages. Public
classes use plain role names (not `Native*`):

```python
from n4m.estimators.regression.regularized import Ridge, RidgePLS
from n4m.estimators.regression.latent import PLS, PCR
from n4m.transform.scatter import SNV, MSC
from n4m.transform.smoothing import SavitzkyGolay
from n4m.feature_selection.wrapper import CARS
from n4m.model_selection.splitters import KennardStone
from n4m.domain_adaptation.orthogonalization import EPO
from n4m.augmentation.noise import GaussianAdditiveNoise
from n4m.ensemble import AOMRidgeBlender
from n4m.compose.aom_superblock import AOMRidgePLSSuperblock
from n4m.decomposition import FlexiblePCA
```

The estimators/transformers are sklearn-compatible with zero-copy NumPy
`n4m_matrix_view_t` round-trips. The slim `pls4all` package keeps its name (the
subset contract) but calls the same ABI-2 symbols under the hood.