Skip to content

Getting started

Install

pip install "bdp-model-gate[structured]"

Pulls scikit-learn, fairlearn and shap, which the fairness checks and most metrics need.

pip install bdp-model-gate

Compliance, PII and prompt-injection checks work with just this. Checks needing the extra libraries report NOT_APPLICABLE rather than failing, and regression metrics still work — they are implemented in numpy.

Python 3.9 – 3.13.

Gate a model

from bdp_model_gate import ModelGate, StructuredGateContext

context = StructuredGateContext(
    model=my_model,
    X=X_val,
    y_true=y_val,
    y_pred=y_pred,  # positive-class probability, for binary
    protected_df=protected_val,  # optional — enables fairness
    model_card=my_model_card,  # optional — enables compliance
    task="binary",
)

report = ModelGate().run(context)
print(report.summary())
report.to_json("gate_report.json")

Gate status: NEEDS_REVIEW (2741ms, binary)
  roc_auc: 0.8637
  performance: 0 flag(s)
  compliance: 0 flag(s)
  security: 0 flag(s)
  fairness: 4 flag(s)

Only model (or a predict_fn), X, y_true and y_pred are required. Every other field is optional, and omitting one makes the checks that need it report NOT_APPLICABLE instead of failing. See Concepts.

Read the verdict

if report.gate_status == "BLOCKED":
    raise SystemExit("Model failed governance — see gate_report.json")

for flag in report.flags:
    print(flag.check_name, flag.detail)
Status Meaning Your pipeline should
PASS nothing flagged deploy automatically
NEEDS_REVIEW only non-blocking flags pause for human sign-off
BLOCKED a blocking check failed hard-fail the build

From the command line

bdp-model-gate \
  --model model.joblib \
  --data validation.csv \
  --target-col label \
  --protected protected.csv \
  --model-card model_card.json \
  --task binary \
  --metric roc_auc --min-score 0.80 \
  --output gate_report.json

Exit codes are 0, 2 and 1 for the three statuses, so a pipeline can tell deploy from ask a human. Full options in the CLI reference.

Next