Metadata-Version: 2.4
Name: kmds-modeling
Version: 0.1.0
Summary: KMDS modeling pipeline package for KMDS lifecycle and model selection.
Requires-Python: >=3.13
Description-Content-Type: text/markdown
Requires-Dist: kmds-featurization>=0.1.5
Requires-Dist: click>=8.0
Requires-Dist: pandas>=2.0
Requires-Dist: numpy>=1.26
Requires-Dist: scikit-learn>=1.4
Requires-Dist: PyYAML>=6.0
Requires-Dist: joblib>=1.3

# KMDS Modeling

`kmds-modeling` is a lightweight modeling package designed to work inside the KMDS ecosystem. It provides generic modeling infrastructure and pipeline utilities for KMDS-style workflows, while leaving domain-specific examples and workspace-specific implementations separate.

## What this package provides
- `src/kmds_modeling/core` — generic modeling package infrastructure
- `src/kmds_modeling/core/path_coordinator.py` — workspace-rooted path resolution for KMDS modeling
- `src/kmds_modeling/core/notebook_utils.py` — notebook-friendly workspace resolver
- `src/kmds_modeling/cli.py` — installable CLI glue for evaluation and export
- `models/sba_example` — an example SBA-specific modeling workflow kept outside the installed package

## Intended usage
This package is meant to be installed into a KMDS workspace and used against modeling artifacts generated by KMDS tools such as `kmds-featurization`. The package does not embed any domain-specific SBA implementation in the installable distribution.

## Installation
```bash
pip install kmds-modeling
```

## CLI commands
After installing, the package exposes the `kmds-modeling` CLI:

```bash
kmds-modeling evaluate --config /path/to/modeling_config.yaml
kmds-modeling export --config /path/to/modeling_config.yaml
```

## Working directory and configuration
KMDS modeling expects a `working_dir` and a `modeling_config.yaml` that defines the workspace layout. The package resolves paths using the `PathCoordinator` and writes modeling outputs into the workspace `models/` directory by default.

## Example workflow
1. Use KMDS featurization to generate `model_ready_numeric_data.csv` under `data/featurization/`.
2. Create `modeling_config.yaml` with a `working_dir` pointing to your KMDS workspace.
3. Run the package CLI to evaluate and export model artifacts.

## Packaging note
The published PyPI package should only contain the generic package code under `src/kmds_modeling/`. Workspace-specific examples such as `models/sba_example/` are intentionally kept outside the installable package source tree.

## Contributing
If you want to add another KMDS modeling example, put it under `models/<example_name>/` and leave the core package unchanged.
