MonikaDvorackova commited on
Commit
0f16e07
·
unverified ·
1 Parent(s): 25dfeb9

Add governance risk inference and tests

Browse files
Files changed (3) hide show
  1. README.md +33 -0
  2. inference.py +87 -0
  3. tests/test_inference.py +37 -0
README.md CHANGED
@@ -155,3 +155,36 @@ Website: https://govbase.dev
155
  ## License
156
 
157
  Apache-2.0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
155
  ## License
156
 
157
  Apache-2.0
158
+
159
+ ## Inference
160
+
161
+ The repository includes `inference.py` for local predictions.
162
+
163
+ Example:
164
+
165
+ ```bash
166
+ python inference.py --json '{
167
+ "sector": "financial_services",
168
+ "impact": "high",
169
+ "decision_autonomy": "automated",
170
+ "human_oversight": "none",
171
+ "monitoring": "none",
172
+ "traceability": "none",
173
+ "technical_documentation": "partial"
174
+ }'
175
+ ```
176
+
177
+ Example output:
178
+
179
+ ```json
180
+ {
181
+ "risk_tier": "unacceptable",
182
+ "class_probabilities": {
183
+ "high": 0.24612,
184
+ "lower": 0.08766,
185
+ "unacceptable": 0.66622
186
+ }
187
+ }
188
+ ```
189
+
190
+ The returned class probabilities are model scores from the logistic-regression baseline. They should not be interpreted as calibrated probabilities of legal or real-world governance risk.
inference.py ADDED
@@ -0,0 +1,87 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import argparse
2
+ import json
3
+ from pathlib import Path
4
+
5
+ import joblib
6
+ import pandas as pd
7
+
8
+
9
+ MODEL_PATH = Path(__file__).parent / "model.joblib"
10
+
11
+ FEATURES = [
12
+ "sector",
13
+ "impact",
14
+ "decision_autonomy",
15
+ "human_oversight",
16
+ "monitoring",
17
+ "traceability",
18
+ "technical_documentation",
19
+ ]
20
+
21
+
22
+ def load_model():
23
+ if not MODEL_PATH.exists():
24
+ raise FileNotFoundError(
25
+ f"Model artifact not found: {MODEL_PATH}"
26
+ )
27
+
28
+ return joblib.load(MODEL_PATH)
29
+
30
+
31
+ def validate_input(data: dict) -> None:
32
+ missing = [feature for feature in FEATURES if feature not in data]
33
+
34
+ if missing:
35
+ raise ValueError(
36
+ "Missing required features: " + ", ".join(missing)
37
+ )
38
+
39
+
40
+ def predict_governance_risk(data: dict) -> dict:
41
+ validate_input(data)
42
+
43
+ model = load_model()
44
+
45
+ frame = pd.DataFrame(
46
+ [{feature: data[feature] for feature in FEATURES}]
47
+ )
48
+
49
+ prediction = model.predict(frame)[0]
50
+
51
+ result = {
52
+ "risk_tier": str(prediction),
53
+ }
54
+
55
+ if hasattr(model, "predict_proba"):
56
+ probabilities = model.predict_proba(frame)[0]
57
+ classes = model.classes_
58
+
59
+ result["class_probabilities"] = {
60
+ str(label): round(float(probability), 6)
61
+ for label, probability in zip(classes, probabilities)
62
+ }
63
+
64
+ return result
65
+
66
+
67
+ def main():
68
+ parser = argparse.ArgumentParser(
69
+ description="Run the AIGov governance risk classifier."
70
+ )
71
+
72
+ parser.add_argument(
73
+ "--json",
74
+ required=True,
75
+ help="Governance scenario encoded as JSON.",
76
+ )
77
+
78
+ args = parser.parse_args()
79
+
80
+ data = json.loads(args.json)
81
+ result = predict_governance_risk(data)
82
+
83
+ print(json.dumps(result, indent=2))
84
+
85
+
86
+ if __name__ == "__main__":
87
+ main()
tests/test_inference.py ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import sys
2
+ from pathlib import Path
3
+
4
+ ROOT = Path(__file__).resolve().parents[1]
5
+ sys.path.insert(0, str(ROOT))
6
+
7
+ from inference import predict_governance_risk
8
+
9
+
10
+ def test_high_risk_scenario():
11
+ result = predict_governance_risk({
12
+ "sector": "financial_services",
13
+ "impact": "high",
14
+ "decision_autonomy": "automated",
15
+ "human_oversight": "none",
16
+ "monitoring": "none",
17
+ "traceability": "none",
18
+ "technical_documentation": "partial",
19
+ })
20
+
21
+ assert result["risk_tier"] == "unacceptable"
22
+ assert "class_probabilities" in result
23
+
24
+
25
+ def test_lower_risk_scenario():
26
+ result = predict_governance_risk({
27
+ "sector": "enterprise_productivity",
28
+ "impact": "low",
29
+ "decision_autonomy": "assistive",
30
+ "human_oversight": "strong",
31
+ "monitoring": "strong",
32
+ "traceability": "strong",
33
+ "technical_documentation": "strong",
34
+ })
35
+
36
+ assert result["risk_tier"] == "lower"
37
+ assert "class_probabilities" in result