p2predict 0.9.0__py3-none-any.whl

p2predict/__init__.py

@@ -0,0 +1,88 @@
+"""P2Predict — parametric price benchmarking for procurement and engineering.
+
+This is the public Python API. The same functionality is also exposed via
+the ``p2predict`` and ``p2predict-train`` console scripts (installed as
+entry points), and via the upcoming MCP server. Pick the surface that fits
+your workflow:
+
+  • Python API (this module) — embed P2Predict in scripts, notebooks, or
+    agent code.
+  • CLI — interactive use, batch processing.
+  • MCP — let an AI agent call P2Predict on behalf of a procurement user.
+
+Typical Python usage::
+
+    import pandas as pd
+    from p2predict import auto_train, predict_interval, explain, load_model
+
+    # Train (returns a fitted pipeline ready to predict / explain).
+    data = pd.read_csv("purchases.csv")
+    features = ["Weight", "Region", "Supplier", "Size"]
+    model, info = auto_train(data, target="Price", features=features)
+
+    # Save and reload.
+    info.save("models/my_model.model")
+    loaded = load_model("models/my_model.model")
+
+    # Predict on a new part.
+    new_part = pd.DataFrame([{"Weight": 15, "Region": "EU",
+                              "Supplier": "A", "Size": "Standard"}])
+    pred = loaded.predict(new_part)
+
+    # Likely range + per-feature explanation, both axiomatically grounded.
+    intervals = loaded.predict_interval(new_part, coverage=0.90)
+    explanation = loaded.explain(new_part.iloc[[0]])
+"""
+
+from p2predict.explain import (
+    Explanation,
+    explain_batch,
+    explain_row as explain,
+    top_drivers,
+)
+from p2predict.intervals import IntervalResult, predict_interval
+from p2predict.outliers import (
+    POLICIES as OUTLIER_POLICIES,
+    apply_feature_outlier_policy,
+    apply_outlier_policy,
+)
+from p2predict.trained_model_io import (
+    LoadModel as load_model,
+    P2PREDICT_VERSION,
+    SaveModel as save_model,
+    Serialize_Trained_Model,
+)
+from p2predict.training import ALGORITHMS, auto_train, start_training
+from p2predict.whatif import WhatIfResult, compute_whatif as what_if
+
+__version__ = P2PREDICT_VERSION.lstrip("v")
+
+__all__ = [
+    # Training entry points.
+    "auto_train",
+    "start_training",
+    "ALGORITHMS",
+    # Prediction utilities (the fitted model itself has `.predict()`; we expose
+    # the extras here).
+    "predict_interval",
+    "explain",
+    "explain_batch",
+    "top_drivers",
+    "what_if",
+    # Persistence.
+    "save_model",
+    "load_model",
+    "Serialize_Trained_Model",
+    # Outlier handling (also wired into the train CLI; exposed here so
+    # programmatic users can run the same policy from Python).
+    "apply_outlier_policy",
+    "apply_feature_outlier_policy",
+    "OUTLIER_POLICIES",
+    # Result containers — useful for typed downstream code.
+    "Explanation",
+    "IntervalResult",
+    "WhatIfResult",
+    # Version (mirrors the persisted model metadata field).
+    "P2PREDICT_VERSION",
+    "__version__",
+]

p2predict/__main__.py

@@ -0,0 +1,11 @@
+"""Allow ``python -m p2predict`` to invoke the predict CLI without needing
+the console script entry point. Equivalent to ``p2predict`` after
+``pip install``.
+
+For training, use ``python -m p2predict.cli.train`` or the
+``p2predict-train`` console script.
+"""
+from p2predict.cli.predict import main
+
+if __name__ == "__main__":
+    main()

p2predict/cli/__init__.py

@@ -0,0 +1,9 @@
+"""Command-line entry points for P2Predict.
+
+These are the Click commands that ``pyproject.toml`` registers as
+``p2predict-train`` and ``p2predict`` console scripts. They are also
+importable for tests::
+
+    from p2predict.cli.train import train
+    from p2predict.cli.predict import main
+"""