Class OpsMetrics

src/evaluation/models.py:57–66 · view source on GitHub ↗

Per-task operational metrics derived from a trajectory.

Source from the content-addressed store, hash-verified

55
56
57	class OpsMetrics(BaseModel):
58	"""Per-task operational metrics derived from a trajectory."""
59
60	turn_count: int = 0
61	tool_call_count: int = 0
62	unique_tools: list[str] = Field(default_factory=list)
63	tokens_in: int = 0
64	tokens_out: int = 0
65	duration_ms: float \| None = None
66	est_cost_usd: float \| None = None
67
68
69	class ScorerResult(BaseModel):

_resultFunction · 0.90

test_handles_none_trajectoryMethod · 0.90

test_sums_and_percentilesMethod · 0.90

test_cost_only_when_some_presentMethod · 0.90

_resultFunction · 0.90

test_build_report_includes_score_summaryFunction · 0.90

metrics_from_trajectoryFunction · 0.85

_from_sdk_trajectoryFunction · 0.85

_from_plan_executeFunction · 0.85

no outgoing calls

_resultFunction · 0.72

test_handles_none_trajectoryMethod · 0.72

test_sums_and_percentilesMethod · 0.72

test_cost_only_when_some_presentMethod · 0.72

_resultFunction · 0.72

test_build_report_includes_score_summaryFunction · 0.72