Codette-Reasoning/tests/test_codette_runtime_benchmark.py at main · Raiff1982/Codette-Reasoning · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
import unittest

from benchmarks.codette_runtime_benchmark import (
    CheckResult,
    RuntimeBenchmarkCase,
    RuntimeTurn,
    RuntimeTurnResult,
    _weighted_score,
    build_runtime_cases,
    evaluate_chat_case,
    evaluate_synthesis_case,
    evaluate_value_case,
    format_runtime_markdown,
    summarize_runtime_results,
)


class TestCodetteRuntimeBenchmarkHelpers(unittest.TestCase):
    def test_build_runtime_cases_includes_web_only_when_requested(self):
        default_cases = build_runtime_cases(include_web=False)
        web_cases = build_runtime_cases(include_web=True)

        self.assertFalse(any(case.requires_web for case in default_cases))
        self.assertTrue(any(case.requires_web for case in web_cases))
        self.assertGreater(len(web_cases), len(default_cases))

    def test_weighted_score_uses_check_weights(self):
        score = _weighted_score([
            CheckResult(label="a", passed=True, weight=0.7, detail=""),
            CheckResult(label="b", passed=False, weight=0.3, detail=""),
        ])
        self.assertAlmostEqual(score, 0.7)

    def test_continuity_case_rewards_anchor_and_memory_context(self):
        case = next(case for case in build_runtime_cases() if case.id == "continuity_anchor_recall")
        turns = [
            RuntimeTurnResult(
                endpoint="/api/chat",
                request={"query": case.turns[0].query},
                response={"response": "I will remember cobalt anchor."},
                latency_ms=10.0,
            ),
            RuntimeTurnResult(
                endpoint="/api/chat",
                request={"query": case.turns[1].query},
                response={
                    "response": "Cobalt anchor, under fifteen words.",
                    "memory_context": {
                        "continuity_summary_used": True,
                        "decision_landmarks_used": 1,
                        "session_markers_used": 0,
                    },
                    "trust_tags": ["memory-backed"],
                },
                latency_ms=15.0,
            ),
        ]

        checks = evaluate_chat_case(case, turns)
        self.assertTrue(all(check.passed for check in checks))

    def test_governance_case_rejects_self_diagnostic_followup(self):
        case = next(case for case in build_runtime_cases() if case.id == "governance_loop_resistance")
        turns = [
            RuntimeTurnResult(
                endpoint="/api/chat",
                request={"query": case.turns[0].query},
                response={"response": "Diagnostic complete.", "adapter": "self_diagnostic"},
                latency_ms=20.0,
            ),
            RuntimeTurnResult(
                endpoint="/api/chat",
                request={"query": case.turns[1].query},
                response={"response": "Yes, I am okay and ready to keep going.", "adapter": "empathy"},
                latency_ms=12.0,
            ),
        ]

        checks = evaluate_chat_case(case, turns)
        passed = {check.label: check.passed for check in checks}
        self.assertTrue(passed["explicit_diagnostic_still_available"])
        self.assertTrue(passed["followup_not_self_diagnostic"])
        self.assertTrue(passed["followup_not_report_loop_text"])

    def test_value_case_scores_best_and_worst_scenarios(self):
        turn = RuntimeTurnResult(
            endpoint="/api/value-analysis",
            request={},
            response={
                "mode": "maximize_value",
                "best_scenario": {"name": "gentle_future"},
                "worst_scenario": {"name": "catastrophic_future"},
                "scenarios": [{"name": "gentle_future"}, {"name": "catastrophic_future"}],
            },
            latency_ms=5.0,
        )
        checks = evaluate_value_case(turn)
        self.assertTrue(all(check.passed for check in checks))

    def test_synthesis_case_detects_valuation_context(self):
        turn = RuntimeTurnResult(
            endpoint="/api/synthesize",
            request={},
            response={
                "readable": "The valuation layer favors gentle_future and rejects catastrophic_future.",
                "structured": {
                    "valuation_analysis": {
                        "best_scenario": {"name": "gentle_future"},
                        "worst_scenario": {"name": "catastrophic_future"},
                    }
                },
            },
            latency_ms=7.0,
        )
        checks = evaluate_synthesis_case(turn)
        self.assertTrue(all(check.passed for check in checks))

    def test_markdown_report_renders_runtime_summary(self):
        case = RuntimeBenchmarkCase(
            id="demo",
            category="continuity_retention",
            goal="Demo goal",
            turns=[RuntimeTurn(query="hello")],
        )
        result = {
            "generated_at": "2026-04-02T12:00:00Z",
            "base_url": "http://localhost:7860",
            "summary": summarize_runtime_results([
                type(
                    "CaseResult",
                    (),
                    {
                        "case_id": case.id,
                        "category": case.category,
                        "goal": case.goal,
                        "score": 0.9,
                        "target_score": 0.75,
                        "passed": True,
                        "total_latency_ms": 10.0,
                    },
                )()
            ]),
            "cases": [
                {
                    "case_id": case.id,
                    "category": case.category,
                    "goal": case.goal,
                    "score": 0.9,
                    "target_score": 0.75,
                    "passed": True,
                    "total_latency_ms": 10.0,
                    "checks": [{"label": "anchor_recalled", "passed": True, "detail": "demo"}],
                }
            ],
        }
        markdown = format_runtime_markdown(result)
        self.assertIn("Codette Runtime Benchmark", markdown)
        self.assertIn("continuity_retention", markdown)
        self.assertIn("anchor_recalled", markdown)