chore: clean remaining ruff/mypy findings and enforce lint in CI (#10)

* Add files via upload

* Add files via upload

Author: AmirhosseinHonardoust

.github/workflows/ci.yml

@@ -6,6 +6,34 @@ on:
   workflow_dispatch:
 
 jobs:
+  lint:
+    name: Lint and type-check
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Check out repository
+        uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+          cache: "pip"
+
+      - name: Install tooling
+        run: |
+          python -m pip install --upgrade pip
+          pip install ruff black mypy
+
+      - name: Ruff
+        run: ruff check src app.py tests
+
+      - name: Black
+        run: black --check src app.py tests
+
+      - name: Mypy
+        run: mypy src app.py
+
   test-and-smoke:
     name: Test and smoke workflow on Python ${{ matrix.python-version }}
     runs-on: ubuntu-latest

app.py

@@ -16,17 +16,17 @@
 if str(PROJECT_ROOT) not in sys.path:
     sys.path.append(str(PROJECT_ROOT))
 
-from src.config import METRICS_DIR, PROCESSED_DATA_DIR, TARGET_COL  # type: ignore
-from src.dashboard_utils import (  # type: ignore
+from src.config import METRICS_DIR, PROCESSED_DATA_DIR, TARGET_COL  # noqa: E402
+from src.dashboard_utils import (  # noqa: E402
     add_risk_band,
     build_model_metadata,
     format_float,
     format_percent,
     summarize_scored_transactions,
 )
-from src.reason_codes import shap_reason_codes, split_reason_codes  # type: ignore
-from src.score_new_transactions import score_dataframe  # type: ignore
-from src.validation import (  # type: ignore
+from src.reason_codes import shap_reason_codes, split_reason_codes  # noqa: E402
+from src.score_new_transactions import score_dataframe  # noqa: E402
+from src.validation import (  # noqa: E402
     REQUIRED_FEATURE_COLUMNS,
     DataValidationError,
     validate_scoring_dataframe,
@@ -148,19 +148,13 @@ def explain_single_transaction(model, df_scored: pd.DataFrame, row_idx: int):
     try:
         import scipy.sparse as sp
 
-        if sp.issparse(x_transformed):
-            x_for_shap = x_transformed.toarray()
-        else:
-            x_for_shap = x_transformed
+        x_for_shap = x_transformed.toarray() if sp.issparse(x_transformed) else x_transformed
     except ImportError:
         x_for_shap = x_transformed
 
     shap_vals = explainer.shap_values(x_for_shap)
 
-    if isinstance(shap_vals, list):
-        shap_for_fraud_class = shap_vals[1][0]
-    else:
-        shap_for_fraud_class = shap_vals[0]
+    shap_for_fraud_class = shap_vals[1][0] if isinstance(shap_vals, list) else shap_vals[0]
 
     fig = plot_single_shap_bar(shap_for_fraud_class, feature_names)
     reasons = shap_reason_codes(shap_for_fraud_class, feature_names, max_reasons=5)

src/evaluate.py

@@ -113,7 +113,9 @@ def compute_baseline_metrics(
         else:
             proba = clf.predict(X_test)
 
-        metrics = compute_probability_metrics(y_test.to_numpy(), proba)
+        metrics: dict[str, float | str] = dict(
+            compute_probability_metrics(y_test.to_numpy(), proba)
+        )
         metrics["strategy"] = strategy
         baselines[name] = metrics
 

src/explain.py

@@ -34,10 +34,7 @@ def compute_and_plot_global_shap(
     For tree models (RandomForest), TreeExplainer is appropriate.
     """
     # Subsample for speed if needed
-    if len(X) > max_samples:
-        X_sample = X.sample(max_samples, random_state=0)
-    else:
-        X_sample = X
+    X_sample = X.sample(max_samples, random_state=0) if len(X) > max_samples else X
 
     # Extract underlying estimator after preprocessing
     preprocessor = model.named_steps["preprocess"]
@@ -46,10 +43,7 @@ def compute_and_plot_global_shap(
     X_transformed = preprocessor.transform(X_sample)
 
     # Convert sparse matrices to dense for SHAP summary plot
-    if sp.issparse(X_transformed):
-        X_for_shap = X_transformed.toarray()
-    else:
-        X_for_shap = X_transformed
+    X_for_shap = X_transformed.toarray() if sp.issparse(X_transformed) else X_transformed
 
     explainer = shap.TreeExplainer(clf)
     shap_values = explainer.shap_values(X_for_shap)

src/generate_synthetic_data.py

@@ -182,8 +182,8 @@ def generate_synthetic_fraud_dataset(
     device_risk_score[low_risk_fraud] = rng.beta(1.5, 4.5, size=int(low_risk_fraud.sum()))
     ip_risk_score[low_risk_fraud] = rng.beta(1.5, 4.2, size=int(low_risk_fraud.sum()))
 
-    device_risk_score = np.round(np.clip(device_risk_score, 0, 1), 4)
-    ip_risk_score = np.round(np.clip(ip_risk_score, 0, 1), 4)
+    device_risk_score = np.round(np.clip(device_risk_score, 0, 1), 4).astype(float)
+    ip_risk_score = np.round(np.clip(ip_risk_score, 0, 1), 4).astype(float)
 
     transaction_type = _sample_by_class(
         rng,

src/threshold_policy.py

@@ -33,7 +33,7 @@ def _policy_row(name: str, row: dict[str, Any], rationale: str) -> dict[str, Any
         "tn",
         "fn",
     ]
-    result = {"policy": name, "rationale": rationale}
+    result: dict[str, Any] = {"policy": name, "rationale": rationale}
     for key in keys:
         result[key] = row.get(key)
     return result
@@ -91,7 +91,8 @@ def add(name: str, row: dict[str, Any] | None, rationale: str) -> None:
     add(
         "cost_optimized",
         cost_optimized,
-        "Minimizes expected business cost under the configured false-positive and false-negative costs.",
+        "Minimizes expected business cost under the configured "
+        "false-positive and false-negative costs.",
     )
 
     balanced_f1 = _max_by(
@@ -135,7 +136,8 @@ def add(name: str, row: dict[str, Any] | None, rationale: str) -> None:
     add(
         "high_precision",
         high_precision,
-        f"Maintains precision of at least {target_precision:.0%} while preserving as much recall as possible.",
+        f"Maintains precision of at least {target_precision:.0%} "
+        "while preserving as much recall as possible.",
     )
 
     review_capacity_pool = [
@@ -182,8 +184,10 @@ def build_threshold_policy_summary(
         "policy_candidates": candidates,
         "notes": [
             "These policies are decision-support artifacts, not automatic approval rules.",
-            "Thresholds should be reviewed with business, compliance, and operations stakeholders before deployment.",
-            "The demo dataset is synthetic; threshold values should not be reused for real banking data without validation.",
+            "Thresholds should be reviewed with business, compliance, and operations "
+            "stakeholders before deployment.",
+            "The demo dataset is synthetic; threshold values should not be reused "
+            "for real banking data without validation.",
         ],
     }
 
@@ -260,7 +264,10 @@ def _render_markdown_policy(summary: dict[str, Any]) -> str:
 
     for row in candidates:
         lines.append(
-            "| {policy} | {threshold} | {precision} | {recall} | {fpr} | {flagged} | {cost} | {rationale} |".format(
+            (
+                "| {policy} | {threshold} | {precision} | {recall} | "
+                "{fpr} | {flagged} | {cost} | {rationale} |"
+            ).format(
                 policy=row.get("policy", "n/a"),
                 threshold=_format_rate(row.get("threshold")),
                 precision=_format_rate(row.get("precision")),