33 lines
1.0 KiB
Python
33 lines
1.0 KiB
Python
import os
|
|
|
|
os.environ.setdefault("OPENAI_API_KEY", "test-key")
|
|
|
|
from yuxi.knowledge.eval.metrics import EvaluationMetricsCalculator, RetrievalMetrics
|
|
|
|
|
|
def test_retrieval_metrics_use_metadata_chunk_id():
|
|
retrieved_chunks = [
|
|
{"metadata": {"chunk_id": "chunk_a"}},
|
|
{"metadata": {"chunk_id": "chunk_b"}},
|
|
]
|
|
|
|
metrics = EvaluationMetricsCalculator.calculate_retrieval_metrics(
|
|
retrieved_chunks, ["chunk_b", "chunk_c"], k_values=[1, 3]
|
|
)
|
|
|
|
assert metrics["recall@1"] == 0.0
|
|
assert metrics["recall@3"] == 0.5
|
|
assert metrics["f1@3"] == RetrievalMetrics.f1_score_at_k(["chunk_a", "chunk_b"], ["chunk_b", "chunk_c"], 3)
|
|
|
|
|
|
def test_overall_score_uses_retrieval_metrics_only():
|
|
score = EvaluationMetricsCalculator.calculate_overall_score([{"recall@1": 1.0, "f1@1": 0.5}], [{"score": 0.25}])
|
|
|
|
assert score == 0.75
|
|
|
|
|
|
def test_overall_score_returns_none_without_retrieval_metrics():
|
|
score = EvaluationMetricsCalculator.calculate_overall_score([], [{"score": 1.0}])
|
|
|
|
assert score is None
|