Files
mindfulness/server/tests/test_rerank_freqcap.py
2026-02-02 16:47:37 +08:00

131 lines
4.1 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
from __future__ import annotations
import pytest
from app.features.personalized_reco.content_repository.types import ContentProfileDTO
from app.features.personalized_reco.rerank_freqcap.rerank import rerank_and_freqcap
from app.features.personalized_reco.rerank_freqcap.types import ScoredCandidate
def _c(
*,
content_id: int,
score: float,
author_id: str | None = None,
template_id: str | None = None,
stage: str = "general",
need_key: str = "emotional_support",
context_key: str = "family",
) -> ScoredCandidate:
cp = ContentProfileDTO(
content_id=content_id,
text="t",
stage=stage, # type: ignore[arg-type]
emotion_score=None,
context_suitability={context_key: 1.0},
need_suitability={need_key: 1.0},
personalization_power=0.0,
risk_flags=[],
author_id=author_id,
template_id=template_id,
)
return ScoredCandidate(
content_id=content_id,
final_score=score,
author_id=author_id,
template_id=template_id,
content_profile=cp,
)
def test_dedup_normalizes_str_int_ids() -> None:
cands = [_c(content_id=1, score=0.9), _c(content_id=2, score=0.8), _c(content_id=3, score=0.7)]
r = rerank_and_freqcap(
scene="push",
scored_candidates=cands,
already_recommended_ids=["1", "bad"],
touched_or_viewed_ids=[2],
k=10,
)
got_ids = [x.content_id for x in r.ranked_items]
assert got_ids == [3]
assert r.meta.candidate_pool_size_after_dedup == 1
assert r.meta.freqcap_filtered_counts["sentence"] == 2
def test_freqcap_missing_recent_author_template_is_recorded() -> None:
cands = [
_c(content_id=1, score=0.9, author_id="a1", template_id="t1"),
_c(content_id=2, score=0.8, author_id="a2", template_id="t2"),
]
r = rerank_and_freqcap(
scene="push",
scored_candidates=cands,
already_recommended_ids=[],
touched_or_viewed_ids=[],
k=10,
recent_author_ids=None,
recent_template_ids=None,
)
assert r.meta.missing_history_fields == ["author", "template"]
assert "author" not in r.meta.freqcap_filtered_counts # 未提供则不执行该维度
assert "template" not in r.meta.freqcap_filtered_counts
def test_freqcap_filters_by_recent_author_when_provided() -> None:
cands = [
_c(content_id=1, score=0.9, author_id="a1", template_id="t1"),
_c(content_id=2, score=0.8, author_id="a2", template_id="t2"),
]
r = rerank_and_freqcap(
scene="widget",
scored_candidates=cands,
already_recommended_ids=[],
touched_or_viewed_ids=[],
k=10,
recent_author_ids=["a1"],
recent_template_ids=None,
)
got_ids = [x.content_id for x in r.ranked_items]
assert got_ids == [2]
assert r.meta.missing_history_fields == ["template"]
assert r.meta.freqcap_filtered_counts["author"] == 1
def test_feed_mmr_picks_diverse_second_item() -> None:
# 构造Top1 是 a1c2 分数略高但同作者c3 分数略低但不同作者/阶段
c1 = _c(content_id=1, score=1.0, author_id="a1", template_id="t1", stage="general")
c2 = _c(content_id=2, score=0.99, author_id="a1", template_id="t2", stage="general")
c3 = _c(content_id=3, score=0.95, author_id="a2", template_id="t3", stage="expecting")
r = rerank_and_freqcap(
scene="feed",
scored_candidates=[c1, c2, c3],
already_recommended_ids=[],
touched_or_viewed_ids=[],
k=2,
)
got_ids = [x.content_id for x in r.ranked_items]
assert got_ids[0] == 1
assert got_ids[1] == 3
def test_push_topk_sorted_by_score_after_filters() -> None:
cands = [
_c(content_id=1, score=0.1),
_c(content_id=2, score=0.9),
_c(content_id=3, score=0.8),
]
r = rerank_and_freqcap(
scene="push",
scored_candidates=cands,
already_recommended_ids=[],
touched_or_viewed_ids=[],
k=2,
recent_author_ids=[],
recent_template_ids=[],
)
got_ids = [x.content_id for x in r.ranked_items]
assert got_ids == [2, 3]