Python API

godml tiene una API Python para usar desde notebooks y scripts, además del CLI.

CLI vs API

Para pipelines en producción usa el CLI (godml run -f godml.yml). La API Python es ideal para notebooks, exploración e integración en código existente.

---

Notebook API

from godml import notebook_api as nb

nb.quick_train()

Entrena un modelo sin YAML:

nb.quick_train(
    model_type="xgboost",
    hyperparameters={"max_depth": 6, "eta": 0.1},
    dataset_path="./data/churn.csv",
    target="churned",
)

nb.train_model()

Entrena con datos ya preparados:

import pandas as pd
from sklearn.model_selection import train_test_split
from godml import notebook_api as nb

df = pd.read_csv("./data/churn.csv")
X = df.drop(columns=["churned"])
y = df["churned"]
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)

result = nb.train_model(
    model_type="xgboost",
    X=X_train,
    y=y_train,
    hyperparams={"max_depth": 6, "n_estimators": 100},
)

nb.predict() y nb.evaluate()

predictions = nb.predict(result, X_test)

metrics = nb.evaluate(
    y_true=y_test,
    y_pred=predictions,
    metrics=["accuracy", "f1", "auc"],  # para binaria, la clave es 'auc' (no 'roc_auc')
)
# {'accuracy': 0.82, 'f1': 0.79, 'auc': 0.87}

nb.compare_models()

compare_models lee el resultado de result.metrics, que train_model() no rellena — evalúa cada modelo y asígnalo con setattr antes de comparar:

rf  = nb.train_model("random_forest", X_train, y_train)
xgb = nb.train_model("xgboost", X_train, y_train)

for res in (rf, xgb):
    preds = nb.predict(res, X_test)
    setattr(res, "metrics", nb.evaluate(y_test, preds, ["accuracy", "auc"]))

nb.compare_models([rf, xgb], by="auc")

nb.dataprep_run_inline()

recipe = {
    "inputs": [{"name": "raw", "connector": "csv", "uri": "./data/raw.csv"}],
    "steps": [
        {"op": "fillna",   "params": {"columns": ["age"], "value": 30.0}},
        {"op": "one_hot",  "params": {"columns": ["region"], "drop_first": True}},
    ],
    "outputs": [{"name": "clean", "connector": "csv", "uri": "./data/clean.csv"}],
}

df_clean = nb.dataprep_run_inline(recipe)

---

Pipeline desde código

from godml.config_service.loader import load_pipeline
from godml.providers.mlflow.executor import MLflowExecutor

pipeline = load_pipeline("godml.yml")
executor = MLflowExecutor(pipeline)
executor.validate()
executor.run()

---

Compliance desde código

from godml.compliance_service.pii_detector import PiiDetector
import pandas as pd

df = pd.read_csv("./data/train.csv")
detector = PiiDetector()
pii_cols = detector.detect_all(df)  # {columna: tipo_pii_detectado}

from godml.notebook_api import apply_compliance
df_clean = apply_compliance(df, standard="pci-dss")

---

DataPrep desde código

from godml.dataprep_service.dataprep import DataPrepService
import pandas as pd

df = pd.read_csv("./data/raw.csv")
svc = DataPrepService()
df_clean = svc.run(df, steps=[
    {"op": "fillna",          "params": {"columns": ["age"], "value": 28.0}},
    {"op": "one_hot",         "params": {"columns": ["region"]}},
    {"op": "drop_duplicates", "params": {}},
])

---

GodmlNotebook (OOP)

from godml.notebook_api import GodmlNotebook

g = GodmlNotebook()
g.create_pipeline(
    name="churn-nb",
    model_type="xgboost",
    hyperparameters={"max_depth": 6, "eta": 0.1},
    dataset_path="./data/churn.csv",
)
g.train()

---

Próximamente: referencia completa de clases con tipos y excepciones. hide_edit_link: true

Referencia completa de los endpoints de GODML Core. Todos los endpoints están disponibles en http://localhost:8000 (local) o tu dominio de producción.

La documentación interactiva Swagger está en /docs.

Auth

Método	Endpoint	Descripción
POST	/api/auth/register	Crear cuenta
POST	/api/auth/login	Login (JWT)
GET	/api/auth/me	Usuario actual

Registro

curl -X POST http://localhost:8000/api/auth/register \
  -H "Content-Type: application/json" \
  -d '{"email": "user@example.com", "password": "secret123", "full_name": "Mi Nombre"}'

Login

curl -X POST http://localhost:8000/api/auth/login \
  -H "Content-Type: application/json" \
  -d '{"email": "user@example.com", "password": "secret123"}'

Respuesta: {"access_token": "eyJ...", "token_type": "bearer"}

Projects

Método	Endpoint	Descripción
GET	/api/projects	Listar proyectos
POST	/api/projects	Crear proyecto
GET	/api/projects/{id}	Ver proyecto

Training

Método	Endpoint	Descripción
POST	/api/train/upload	Entrenar con CSV + YAML
POST	/api/train/config	Entrenar con config inline

Jobs

Método	Endpoint	Descripción
GET	/api/jobs	Listar jobs
GET	/api/jobs/{id}	Status de job
GET	/api/jobs/{id}/celery-status	Status en tiempo real

Health

Método	Endpoint	Descripción
GET	/health	Health check
GET	/ping	Simple ping