genoma-evolution 1.0.0

package/.claude/worktrees/phase-2-mcp/tests/collectors/test_hermes_collector.py

@@ -0,0 +1,167 @@
+"""Tests for HermesCollector — Phase 2 verification."""
+
+import pytest
+from backend.promethean.models import TraceRecord, CanonicalRun
+from backend.collectors.hermes_collector import HermesCollector
+
+
+@pytest.fixture
+def collector():
+    """Fixture for HermesCollector."""
+    return HermesCollector()
+
+
+@pytest.fixture
+def sample_trace():
+    """Sample TraceRecord for testing."""
+    return TraceRecord(
+        agent="hermes",
+        agent_version="2.1.143",
+        timestamp="2026-05-19T14:32:00Z",
+        task="Implement user authentication",
+        outcome="success",
+        error_signature=None,
+        context={"skill_name": "auth-middleware"},
+        trace_id="hermes-001"
+    )
+
+
+@pytest.fixture
+def failed_trace():
+    """TraceRecord with error."""
+    return TraceRecord(
+        agent="hermes",
+        agent_version="2.1.143",
+        timestamp="2026-05-19T14:32:00Z",
+        task="Test task",
+        outcome="failure",
+        error_signature="TypeError: Cannot read property 'foo' of undefined",
+        resolution="Rolled back changes",
+        trace_id="hermes-002"
+    )
+
+
+class TestHermesCollectorBasic:
+    """Basic HermesCollector functionality tests."""
+
+    def test_collector_instantiation(self, collector):
+        """Collector should instantiate without error."""
+        assert collector is not None
+        assert collector.VERSION == "0.1.0"
+        assert collector.AGENT_NAME == "hermes"
+
+    def test_collect_from_trace_success(self, collector, sample_trace):
+        """Convert successful TraceRecord to CanonicalRun."""
+        canonical = collector.collect_from_trace(sample_trace)
+
+        assert isinstance(canonical, CanonicalRun)
+        assert canonical.run_id == "hermes-001"
+        assert canonical.agent_name == "hermes"
+        assert canonical.collector == "hermes-trace-ingestor"
+        assert canonical.outcome == "success"
+        assert canonical.task_name == "Implement user authentication"
+        assert canonical.started_at == "2026-05-19T14:32:00Z"
+        assert canonical.provider == "hermes"
+
+    def test_collect_from_trace_with_error(self, collector, failed_trace):
+        """Convert failed TraceRecord with error_signature."""
+        canonical = collector.collect_from_trace(failed_trace)
+
+        assert canonical.outcome == "failure"
+        assert len(canonical.errors) == 1
+        assert canonical.errors[0]["signature"] == "TypeError: Cannot read property 'foo' of undefined"
+        assert canonical.resolution == "Rolled back changes"
+
+    def test_collect_from_trace_no_error(self, collector, sample_trace):
+        """TraceRecord without error_signature should have empty errors list."""
+        canonical = collector.collect_from_trace(sample_trace)
+
+        assert canonical.errors == []
+
+    def test_collect_batch(self, collector, sample_trace, failed_trace):
+        """Batch conversion should work."""
+        traces = [sample_trace, failed_trace]
+        canonicals = collector.collect_batch(traces)
+
+        assert len(canonicals) == 2
+        assert canonicals[0].run_id == "hermes-001"
+        assert canonicals[1].run_id == "hermes-002"
+        assert canonicals[0].outcome == "success"
+        assert canonicals[1].outcome == "failure"
+
+
+class TestCanonicalRunSerialization:
+    """CanonicalRun serialization tests."""
+
+    def test_to_dict(self, collector, sample_trace):
+        """CanonicalRun.to_dict() should produce valid dict."""
+        canonical = collector.collect_from_trace(sample_trace)
+        data = canonical.to_dict()
+
+        assert isinstance(data, dict)
+        assert data["run_id"] == "hermes-001"
+        assert data["agent_name"] == "hermes"
+        assert "collector" in data
+        assert "started_at" in data
+
+    def test_to_json(self, collector, sample_trace):
+        """CanonicalRun.to_json() should produce valid JSON."""
+        import json
+
+        canonical = collector.collect_from_trace(sample_trace)
+        json_str = canonical.to_json()
+
+        # Should be valid JSON
+        parsed = json.loads(json_str)
+        assert parsed["run_id"] == "hermes-001"
+        assert parsed["agent_name"] == "hermes"
+
+    def test_from_dict_roundtrip(self, collector, sample_trace):
+        """CanonicalRun should round-trip through dict."""
+        original = collector.collect_from_trace(sample_trace)
+        data = original.to_dict()
+        restored = CanonicalRun.from_dict(data)
+
+        assert restored.run_id == original.run_id
+        assert restored.agent_name == original.agent_name
+        assert restored.outcome == original.outcome
+        assert restored.task_name == original.task_name
+
+    def test_context_preservation(self, collector, sample_trace):
+        """TraceRecord context should be preserved in CanonicalRun."""
+        canonical = collector.collect_from_trace(sample_trace)
+
+        assert canonical.context == {"skill_name": "auth-middleware"}
+
+    def test_minimal_trace_conversion(self, collector):
+        """Minimal TraceRecord (no optional fields) should convert."""
+        minimal = TraceRecord(
+            agent="hermes",
+            agent_version="2.0.0",
+            timestamp="2026-05-19T10:00:00Z",
+            task="Minimal task",
+            outcome="unknown",
+            trace_id="minimal-001"
+        )
+
+        canonical = collector.collect_from_trace(minimal)
+
+        assert canonical.run_id == "minimal-001"
+        assert canonical.outcome == "unknown"
+        assert canonical.errors == []
+        assert canonical.context == {}
+        assert canonical.resolution is None
+
+
+class TestCanonicalRunRequiredFields:
+    """Test that required fields are always present."""
+
+    def test_required_fields_present(self, collector, sample_trace):
+        """All required fields must be present and non-None."""
+        canonical = collector.collect_from_trace(sample_trace)
+
+        required = {"run_id", "agent_name", "collector", "started_at", "task_name", "outcome"}
+        for field in required:
+            value = getattr(canonical, field)
+            assert value is not None, f"Required field '{field}' is None"
+            assert len(str(value)) > 0, f"Required field '{field}' is empty string"

package/.claude/worktrees/phase-2-mcp/tests/eval/test_engine.py

@@ -0,0 +1,234 @@
+"""Tests for evaluation engine — Phase 5 verification."""
+
+import tempfile
+from pathlib import Path
+
+import pytest
+
+from backend.eval.engine import EvaluationEngine
+from backend.promethean.models import CanonicalRun, RunMetrics
+from backend.storage import RunStore
+
+
+@pytest.fixture
+def temp_db():
+    """Temporary database for testing."""
+    with tempfile.TemporaryDirectory() as tmpdir:
+        yield Path(tmpdir) / "test.db"
+
+
+@pytest.fixture
+def store(temp_db):
+    """RunStore instance with temp database."""
+    return RunStore(db_path=temp_db)
+
+
+@pytest.fixture
+def engine(store):
+    """EvaluationEngine instance."""
+    return EvaluationEngine(store=store)
+
+
+@pytest.fixture
+def sample_run(store):
+    """Sample run stored in database."""
+    run = CanonicalRun(
+        run_id="run-001",
+        agent_name="hermes",
+        collector="hermes-trace-ingestor",
+        started_at="2026-05-19T10:00:00Z",
+        task_name="Task",
+        outcome="success",
+        metrics=RunMetrics(input_tokens=1000, output_tokens=500, tool_call_count=2),
+        errors=[],
+    )
+    store.upsert_run(run)
+    return run
+
+
+@pytest.fixture
+def failed_run(store):
+    """Failed run stored in database."""
+    run = CanonicalRun(
+        run_id="run-002",
+        agent_name="hermes",
+        collector="hermes-trace-ingestor",
+        started_at="2026-05-19T11:00:00Z",
+        task_name="Task",
+        outcome="failure",
+        metrics=RunMetrics(input_tokens=2000, output_tokens=800, tool_call_count=0),
+        errors=[{"signature": "Error", "message": "Failed"}],
+    )
+    store.upsert_run(run)
+    return run
+
+
+class TestEvaluationEngine:
+    """EvaluationEngine tests."""
+
+    def test_engine_instantiation(self, engine):
+        """Engine should instantiate with default scorers."""
+        assert engine is not None
+        assert len(engine.scorers) == 5  # 5 default scorers
+
+    def test_evaluate_single_run(self, engine, sample_run):
+        """evaluate() should return scores from applicable scorers."""
+        scores = engine.evaluate(sample_run)
+
+        assert len(scores) > 0  # At least some scorers apply
+        assert all(hasattr(s, "score") for s in scores)
+        assert all(hasattr(s, "passed") for s in scores)
+        assert all(0.0 <= s.score <= 1.0 for s in scores)
+
+    def test_evaluate_batch(self, engine, sample_run, failed_run):
+        """evaluate_batch() should process multiple runs."""
+        runs = [sample_run, failed_run]
+        result = engine.evaluate_batch(runs)
+
+        assert result["total"] == 2
+        assert result["evaluated"] == 2
+        assert result["errors"] == 0
+
+    def test_get_aggregate_score(self, engine, sample_run):
+        """get_aggregate_score() should return weighted average."""
+        agg = engine.get_aggregate_score(sample_run)
+
+        assert 0.0 <= agg <= 1.0
+        # Success run should score > 0.5
+        assert agg > 0.5
+
+    def test_aggregate_score_consistency(self, engine, sample_run):
+        """Aggregate score should be consistent."""
+        agg1 = engine.get_aggregate_score(sample_run)
+        agg2 = engine.get_aggregate_score(sample_run)
+
+        assert agg1 == agg2
+
+    def test_detect_regression_improvement(self, engine, store):
+        """detect_regression() should detect improvement."""
+        baseline = CanonicalRun(
+            run_id="baseline",
+            agent_name="hermes",
+            collector="hermes-trace-ingestor",
+            started_at="2026-05-19T10:00:00Z",
+            task_name="Task",
+            outcome="partial",
+        )
+        evolved = CanonicalRun(
+            run_id="evolved",
+            agent_name="hermes",
+            collector="hermes-trace-ingestor",
+            started_at="2026-05-19T11:00:00Z",
+            task_name="Task",
+            outcome="success",
+        )
+        store.upsert_run(baseline)
+        store.upsert_run(evolved)
+
+        result = engine.detect_regression("baseline", "evolved", threshold=0.05)
+
+        assert result["improvement"] is True
+        assert result["regression"] is False
+        assert result["delta"] > 0
+
+    def test_detect_regression_failure(self, engine, store):
+        """detect_regression() should detect regression."""
+        baseline = CanonicalRun(
+            run_id="baseline",
+            agent_name="hermes",
+            collector="hermes-trace-ingestor",
+            started_at="2026-05-19T10:00:00Z",
+            task_name="Task",
+            outcome="success",
+        )
+        evolved = CanonicalRun(
+            run_id="evolved",
+            agent_name="hermes",
+            collector="hermes-trace-ingestor",
+            started_at="2026-05-19T11:00:00Z",
+            task_name="Task",
+            outcome="failure",
+        )
+        store.upsert_run(baseline)
+        store.upsert_run(evolved)
+
+        result = engine.detect_regression("baseline", "evolved", threshold=0.05)
+
+        assert result["regression"] is True
+        assert result["improvement"] is False
+        assert result["delta"] < 0
+
+    def test_detect_regression_neutral(self, engine, store):
+        """detect_regression() should detect neutral change."""
+        baseline = CanonicalRun(
+            run_id="baseline",
+            agent_name="hermes",
+            collector="hermes-trace-ingestor",
+            started_at="2026-05-19T10:00:00Z",
+            task_name="Task",
+            outcome="success",
+        )
+        evolved = CanonicalRun(
+            run_id="evolved",
+            agent_name="hermes",
+            collector="hermes-trace-ingestor",
+            started_at="2026-05-19T11:00:00Z",
+            task_name="Task",
+            outcome="success",
+        )
+        store.upsert_run(baseline)
+        store.upsert_run(evolved)
+
+        result = engine.detect_regression("baseline", "evolved", threshold=0.05)
+
+        assert result["neutral"] is True
+        assert result["improvement"] is False
+        assert result["regression"] is False
+
+    def test_detect_regression_missing_run(self, engine):
+        """detect_regression() should handle missing runs."""
+        result = engine.detect_regression("nonexistent-1", "nonexistent-2")
+
+        assert "error" in result
+        assert result["baseline_found"] is False
+        assert result["evolved_found"] is False
+
+
+class TestEngineWithCustomScorers:
+    """Test engine with custom scorer configurations."""
+
+    def test_custom_scorers(self, store):
+        """Engine should work with custom scorer list."""
+        from backend.eval.scorers import OutcomeScorer
+
+        custom_scorers = [OutcomeScorer()]
+        engine = EvaluationEngine(store=store, scorers=custom_scorers)
+
+        run = CanonicalRun(
+            run_id="test",
+            agent_name="hermes",
+            collector="hermes-trace-ingestor",
+            started_at="2026-05-19T10:00:00Z",
+            task_name="Task",
+            outcome="success",
+        )
+
+        scores = engine.evaluate(run)
+        assert len(scores) == 1
+        assert scores[0].scorer == "outcome"
+
+    def test_empty_scorers(self, store):
+        """Engine with empty scorers should return empty list."""
+        engine = EvaluationEngine(store=store, scorers=[])
+
+        run = CanonicalRun(
+            run_id="test",
+            agent_name="hermes",
+            collector="hermes-trace-ingestor",
+            started_at="2026-05-19T10:00:00Z",
+            task_name="Task",
+            outcome="success",
+        )
+
+        scores = engine.evaluate(run)
+        assert len(scores) == 0

package/.claude/worktrees/phase-2-mcp/tests/eval/test_scorers.py

@@ -0,0 +1,249 @@
+"""Tests for evaluation scorers — Phase 5 verification."""
+
+import pytest
+
+from backend.eval.scorers import (
+    OutcomeScorer,
+    ToolEfficiencyScorer,
+    TokenCostScorer,
+    ErrorRecoveryScorer,
+)
+from backend.promethean.models import CanonicalRun, ToolCallRecord, RunMetrics
+
+
+@pytest.fixture
+def success_run():
+    """Successful run with no errors."""
+    return CanonicalRun(
+        run_id="run-success",
+        agent_name="hermes",
+        collector="hermes-trace-ingestor",
+        started_at="2026-05-19T10:00:00Z",
+        task_name="Task",
+        outcome="success",
+        errors=[],
+    )
+
+
+@pytest.fixture
+def failure_run():
+    """Failed run with error."""
+    return CanonicalRun(
+        run_id="run-failure",
+        agent_name="hermes",
+        collector="hermes-trace-ingestor",
+        started_at="2026-05-19T10:00:00Z",
+        task_name="Task",
+        outcome="failure",
+        errors=[{"signature": "TypeError", "message": "Type error"}],
+    )
+
+
+@pytest.fixture
+def run_with_tools():
+    """Run with tool calls."""
+    return CanonicalRun(
+        run_id="run-tools",
+        agent_name="claude-code",
+        collector="claude-code-session-collector",
+        started_at="2026-05-19T10:00:00Z",
+        task_name="Task",
+        outcome="success",
+        tool_calls=[
+            ToolCallRecord(id="1", name="Read"),
+            ToolCallRecord(id="2", name="Read"),
+            ToolCallRecord(id="3", name="Edit"),
+        ],
+    )
+
+
+@pytest.fixture
+def run_with_metrics():
+    """Run with token metrics."""
+    return CanonicalRun(
+        run_id="run-metrics",
+        agent_name="hermes",
+        collector="hermes-trace-ingestor",
+        started_at="2026-05-19T10:00:00Z",
+        task_name="Task",
+        outcome="success",
+        metrics=RunMetrics(
+            input_tokens=5000,
+            output_tokens=2000,
+            cache_tokens=500,
+            tool_call_count=0,
+        ),
+    )
+
+
+class TestOutcomeScorer:
+    """OutcomeScorer tests."""
+
+    def test_success_outcome(self, success_run):
+        """Success outcome should score 1.0."""
+        scorer = OutcomeScorer()
+        score = scorer.score(success_run)
+
+        assert score.score == 1.0
+        assert score.passed is True
+        assert score.scorer == "outcome"
+
+    def test_failure_outcome(self, failure_run):
+        """Failure outcome should score 0.0."""
+        scorer = OutcomeScorer()
+        score = scorer.score(failure_run)
+
+        assert score.score == 0.0
+        assert score.passed is False
+
+    def test_partial_outcome(self, success_run):
+        """Partial outcome should score 0.5."""
+        success_run.outcome = "partial"
+        scorer = OutcomeScorer()
+        score = scorer.score(success_run)
+
+        assert score.score == 0.5
+        assert score.passed is False
+
+    def test_unknown_outcome(self, success_run):
+        """Unknown outcome should score 0.3."""
+        success_run.outcome = "unknown"
+        scorer = OutcomeScorer()
+        score = scorer.score(success_run)
+
+        assert score.score == 0.3
+        assert score.passed is False
+
+    def test_applies_to_all(self, success_run):
+        """OutcomeScorer applies to all runs."""
+        scorer = OutcomeScorer()
+        assert scorer.applies_to(success_run) is True
+
+
+class TestToolEfficiencyScorer:
+    """ToolEfficiencyScorer tests."""
+
+    def test_no_tools(self, success_run):
+        """No tools = 1.0 score (efficient)."""
+        scorer = ToolEfficiencyScorer()
+        score = scorer.score(success_run)
+
+        assert score.score == 1.0
+        assert score.passed is True
+
+    def test_efficient_tools(self, run_with_tools):
+        """2 unique / 3 total = 0.67 ratio (pass)."""
+        scorer = ToolEfficiencyScorer()
+        score = scorer.score(run_with_tools)
+
+        assert score.score > 0.3  # Passes threshold
+        assert score.passed is True
+        assert score.details["unique_tools"] == 2
+        assert score.details["total_calls"] == 3
+
+    def test_applies_only_with_tools(self, success_run, run_with_tools):
+        """ToolEfficiencyScorer applies only with tool calls."""
+        scorer = ToolEfficiencyScorer()
+        assert scorer.applies_to(success_run) is False
+        assert scorer.applies_to(run_with_tools) is True
+
+
+class TestTokenCostScorer:
+    """TokenCostScorer tests."""
+
+    def test_low_tokens(self, run_with_metrics):
+        """Low token count scores high."""
+        scorer = TokenCostScorer()
+        score = scorer.score(run_with_metrics)
+
+        assert score.score > 0.5
+        assert score.passed is True
+
+    def test_high_tokens(self, run_with_metrics):
+        """High token count scores low."""
+        run_with_metrics.metrics.input_tokens = 40000
+        run_with_metrics.metrics.output_tokens = 15000
+        scorer = TokenCostScorer()
+        score = scorer.score(run_with_metrics)
+
+        assert score.score < 0.3
+        assert score.passed is False
+
+    def test_no_metrics(self, success_run):
+        """No metrics = 1.0 score."""
+        scorer = TokenCostScorer()
+        score = scorer.score(success_run)
+
+        assert score.score == 1.0
+        assert score.passed is True
+
+    def test_applies_only_with_metrics(self, success_run, run_with_metrics):
+        """TokenCostScorer applies only with metrics."""
+        scorer = TokenCostScorer()
+        assert scorer.applies_to(success_run) is False
+        assert scorer.applies_to(run_with_metrics) is True
+
+
+class TestErrorRecoveryScorer:
+    """ErrorRecoveryScorer tests."""
+
+    def test_success_no_errors(self, success_run):
+        """Success with no errors = 1.0."""
+        scorer = ErrorRecoveryScorer()
+        score = scorer.score(success_run)
+
+        assert score.score == 1.0
+        assert score.passed is True
+
+    def test_success_with_errors(self, success_run):
+        """Success with errors = 0.8."""
+        success_run.errors = [{"signature": "Warning"}]
+        scorer = ErrorRecoveryScorer()
+        score = scorer.score(success_run)
+
+        assert score.score == 0.8
+        assert score.passed is True
+
+    def test_failure(self, failure_run):
+        """Failure = 0.0."""
+        scorer = ErrorRecoveryScorer()
+        score = scorer.score(failure_run)
+
+        assert score.score == 0.0
+        assert score.passed is False
+
+    def test_applies_to_all(self, success_run):
+        """ErrorRecoveryScorer applies to all runs."""
+        scorer = ErrorRecoveryScorer()
+        assert scorer.applies_to(success_run) is True
+
+
+class TestScorerDeterminism:
+    """Verify scorers are deterministic."""
+
+    def test_outcome_scorer_determinism(self, success_run):
+        """Same run should score identically."""
+        scorer = OutcomeScorer()
+        score1 = scorer.score(success_run)
+        score2 = scorer.score(success_run)
+
+        assert score1.score == score2.score
+        assert score1.passed == score2.passed
+
+    def test_tool_scorer_determinism(self, run_with_tools):
+        """Same run should score identically."""
+        scorer = ToolEfficiencyScorer()
+        score1 = scorer.score(run_with_tools)
+        score2 = scorer.score(run_with_tools)
+
+        assert score1.score == score2.score
+        assert score1.passed == score2.passed
+
+    def test_token_scorer_determinism(self, run_with_metrics):
+        """Same run should score identically."""
+        scorer = TokenCostScorer()
+        score1 = scorer.score(run_with_metrics)
+        score2 = scorer.score(run_with_metrics)
+
+        assert score1.score == score2.score
+        assert score1.passed == score2.passed