diff --git a/routes/dashboard_routes.py b/routes/dashboard_routes.py index d624426..74d0717 100644 --- a/routes/dashboard_routes.py +++ b/routes/dashboard_routes.py @@ -51,6 +51,10 @@ REVIEW_STATUS_OPTIONS = [ 'refresh_unit_comparable', 'identity_veto', 'low_score', + 'refresh_low_score', + 'recoverable_low_score', + 'true_low_confidence', + 'protected_existing_match', 'expired_match', 'refresh_no_result', 'no_result', @@ -62,7 +66,8 @@ REVIEW_STATUS_OPTIONS = [ 'statuses': ('unit_comparable', 'refresh_unit_comparable'), }, {'key': 'identity_veto', 'label': '已排除', 'statuses': ('identity_veto',)}, - {'key': 'low_score', 'label': '低信心', 'statuses': ('low_score',)}, + {'key': 'low_score', 'label': '低信心', 'statuses': ('low_score', 'refresh_low_score', 'recoverable_low_score', 'true_low_confidence')}, + {'key': 'protected_existing_match', 'label': '既有保護', 'statuses': ('protected_existing_match',)}, {'key': 'expired_match', 'label': '價格過期', 'statuses': ('expired_match',)}, {'key': 'no_result', 'label': '找不到同款', 'statuses': ('no_result', 'refresh_no_result')}, ] @@ -249,24 +254,30 @@ def _build_pchome_match_status(attempt=None, ineligible=None): candidate_count = int(attempt.get('candidate_count') or 0) score_text = f"最佳候選 {round(score * 100)}%" if score is not None else "尚無候選分數" - if status in {'low_score', 'refresh_low_score'}: + if status in {'low_score', 'refresh_low_score', 'recoverable_low_score', 'true_low_confidence'}: diagnostic_text = attempt.get('error_message') or '' label, summary = _diagnostic_match_rejection_label( diagnostic_text, score_text, blocked=False, ) + if status == 'recoverable_low_score': + label = '近門檻可回收' + summary = '同品線證據已足夠，但分數仍略低於正式採用門檻' + elif status == 'true_low_confidence': + label = '證據不足' + summary = '目前候選仍缺乏足夠身份證據，先保守不採用' return { 'label': label, 'tone': 'neutral', 'summary': summary, 'detail': f'{candidate_count} 筆候選', } - if status in {'needs_review', 'refresh_needs_review'}: + if status in {'needs_review', 'refresh_needs_review', 'protected_existing_match'}: return { - 'label': '配對衝突待審', + 'label': '既有配對保護', 'tone': 'neutral', - 'summary': '新候選與既有配對不同，需人工確認後再覆蓋', + 'summary': '新候選合理，但正式環境已存在更強既有配對，需人工確認後才覆蓋', 'detail': f'{score_text} / {candidate_count} 筆候選', } if status in {'no_result', 'no_match', 'refresh_no_match'}: diff --git a/services/competitor_intel_repository.py b/services/competitor_intel_repository.py index 434b628..869b691 100644 --- a/services/competitor_intel_repository.py +++ b/services/competitor_intel_repository.py @@ -28,6 +28,10 @@ ACTIONABLE_ATTEMPT_STATUSES = { "refresh_unit_comparable", "identity_veto", "low_score", + "refresh_low_score", + "recoverable_low_score", + "true_low_confidence", + "protected_existing_match", "expired_match", "refresh_no_result", "no_result", @@ -35,7 +39,8 @@ ACTIONABLE_ATTEMPT_STATUSES = { REVIEW_STATUS_FILTER_GROUPS = { "unit_comparable": ("unit_comparable", "refresh_unit_comparable"), "identity_veto": ("identity_veto",), - "low_score": ("low_score",), + "low_score": ("low_score", "refresh_low_score", "recoverable_low_score", "true_low_confidence"), + "protected_existing_match": ("protected_existing_match",), "expired_match": ("expired_match",), "no_result": ("no_result", "refresh_no_result"), } @@ -44,6 +49,10 @@ ATTEMPT_STATUS_LABELS = { "refresh_unit_comparable": "需單位價比較", "identity_veto": "身份否決", "low_score": "低信心待審", + "refresh_low_score": "刷新後仍低信心", + "recoverable_low_score": "近門檻可救回", + "true_low_confidence": "證據不足待觀察", + "protected_existing_match": "既有較強配對保護中", "expired_match": "價格過期待刷新", "refresh_no_result": "刷新找不到商品", "no_result": "找不到同款", @@ -58,6 +67,10 @@ ATTEMPT_ACTION_LABELS = { "refresh_unit_comparable": "人工確認檔期、贈品與單位價", "identity_veto": "確認是否為不同商品線或規格", "low_score": "人工審核候選商品身份", + "refresh_low_score": "檢查 refresh 後是否還有更好的同款候選", + "recoverable_low_score": "優先回放這批近門檻同品線候選", + "true_low_confidence": "保守保留，等待更明確的身份證據", + "protected_existing_match": "比較新舊候選證據，避免覆蓋較強正式配對", "expired_match": "重新刷新 PChome 價格", "refresh_no_result": "調整搜尋詞後重抓", "no_result": "補充搜尋詞或品牌關鍵字", @@ -755,9 +768,11 @@ def _review_queue_cte_and_filter( CASE WHEN la.attempt_status IN ('unit_comparable', 'refresh_unit_comparable') THEN 0 WHEN la.attempt_status = 'identity_veto' THEN 1 - WHEN la.attempt_status = 'low_score' THEN 2 - WHEN la.attempt_status = 'expired_match' THEN 3 - ELSE 4 + WHEN la.attempt_status IN ('recoverable_low_score', 'low_score', 'refresh_low_score') THEN 2 + WHEN la.attempt_status = 'protected_existing_match' THEN 3 + WHEN la.attempt_status = 'true_low_confidence' THEN 4 + WHEN la.attempt_status = 'expired_match' THEN 5 + ELSE 6 END AS priority_rank FROM latest_momo lm JOIN latest_attempt la ON la.sku = lm.sku @@ -902,6 +917,10 @@ def _fetch_competitor_review_queue_uncached(engine, limit: int = 12) -> list[dic 'refresh_unit_comparable', 'identity_veto', 'low_score', + 'refresh_low_score', + 'recoverable_low_score', + 'true_low_confidence', + 'protected_existing_match', 'expired_match', 'refresh_no_result', 'no_result' @@ -910,9 +929,11 @@ def _fetch_competitor_review_queue_uncached(engine, limit: int = 12) -> list[dic CASE WHEN la.attempt_status IN ('unit_comparable', 'refresh_unit_comparable') THEN 0 WHEN la.attempt_status = 'identity_veto' THEN 1 - WHEN la.attempt_status = 'low_score' THEN 2 - WHEN la.attempt_status = 'expired_match' THEN 3 - ELSE 4 + WHEN la.attempt_status IN ('recoverable_low_score', 'low_score', 'refresh_low_score') THEN 2 + WHEN la.attempt_status = 'protected_existing_match' THEN 3 + WHEN la.attempt_status = 'true_low_confidence' THEN 4 + WHEN la.attempt_status = 'expired_match' THEN 5 + ELSE 6 END, lm.momo_price DESC NULLS LAST, la.best_match_score DESC NULLS LAST, diff --git a/services/competitor_price_feeder.py b/services/competitor_price_feeder.py index fff8307..87d3e5b 100644 --- a/services/competitor_price_feeder.py +++ b/services/competitor_price_feeder.py @@ -45,6 +45,19 @@ BATCH_SIZE = 30 # 每批 DB 寫入筆數 RATE_DELAY = float(os.getenv("PCHOME_FEEDER_RATE_DELAY", "1.0")) # 每次 PChome 請求間隔（秒） TTL_HOURS = 6 # competitor_prices 快取有效期 REQUEST_TIMEOUT = float(os.getenv("PCHOME_FEEDER_TIMEOUT", "12")) # 避免外部搜尋 API 長時間卡住排程 +RECOVERABLE_LOW_SCORE_FLOOR = max(MIN_MATCH_SCORE - 0.03, 0.72) +RECOVERABLE_DIAGNOSTIC_REASONS = { + "strong_product_line_match", + "strong_exact_spec_match", + "shared_identity_anchor", + "shared_identity_anchor_no_spec", + "shared_identity_anchor_packaging_variant", + "shared_identity_anchor_marketing_variant", + "shared_identity_anchor_core_line", + "shared_identity_anchor_variant_safe", + "shared_model_token", + "spec_name_alignment", +} # ── Feeder 結果 ─────────────────────────────────────── @dataclass @@ -59,6 +72,28 @@ class FeederResult: attempts_written: int = 0 +def _has_recoverable_identity_signal(diagnostics) -> bool: + if not diagnostics: + return False + reasons = set(getattr(diagnostics, "reasons", ()) or ()) + if reasons & RECOVERABLE_DIAGNOSTIC_REASONS: + return True + return ( + getattr(diagnostics, "brand_score", 0) >= 0.95 + and getattr(diagnostics, "token_score", 0) >= 0.56 + and getattr(diagnostics, "sequence_score", 0) >= 0.50 + and getattr(diagnostics, "comparison_mode", "exact_identity") == "exact_identity" + ) + + +def _classify_low_score_attempt(score: float, diagnostics) -> str: + if getattr(diagnostics, "hard_veto", False): + return "identity_veto" + if score >= RECOVERABLE_LOW_SCORE_FLOOR and _has_recoverable_identity_signal(diagnostics): + return "recoverable_low_score" + return "true_low_confidence" + + def _extract_tags(pchome_product) -> list: """ 從 PChomeProduct 物件提取語意標籤 @@ -716,7 +751,8 @@ class CompetitorPriceFeeder: 這條路徑不重新搜尋，只用前次留下的 PChome product_id 批次查詢最新商品資料， 適合把舊 scorer 卡在 0.70~0.759 的真同款重新推進正式比價。 - 已重評後仍不足門檻的 refresh_low_score 不再重複進隊列，避免排程空轉。 + 僅重跑明顯仍在 exact identity 軌道內、具回收價值的候選； + 真正低信心與 hard veto 不再反覆空轉。 """ if self.engine is None: raise RuntimeError("需要注入 SQLAlchemy engine") @@ -769,7 +805,7 @@ class CompetitorPriceFeeder: AND COALESCE(cp.tags, '[]'::jsonb) ? 'identity_v2' WHERE lm.rn = 1 AND cp.sku IS NULL - AND la.attempt_status = 'low_score' + AND la.attempt_status IN ('low_score', 'refresh_low_score', 'recoverable_low_score') AND la.best_competitor_product_id IS NOT NULL AND la.best_competitor_product_id <> '' AND COALESCE(la.best_match_score, 0) >= :min_score @@ -1212,7 +1248,7 @@ class CompetitorPriceFeeder: continue if score < MIN_MATCH_SCORE and not manual_accept_override: - attempt_status = "identity_veto" if getattr(diagnostics, "hard_veto", False) else "low_score" + attempt_status = _classify_low_score_attempt(score, diagnostics) logger.debug( f"[Feeder] {sku} 比對分數過低 ({score:.3f} < {MIN_MATCH_SCORE})，" f"{_format_match_diagnostics(diagnostics)}" @@ -1263,7 +1299,7 @@ class CompetitorPriceFeeder: momo_price=momo_price, search_terms=search_terms, candidate_count=len(products), - attempt_status="needs_review", + attempt_status="protected_existing_match", best_product=best_product, best_score=score, diagnostics=diagnostics, @@ -1439,7 +1475,7 @@ class CompetitorPriceFeeder: momo_price=momo_price, search_terms=attempt_terms, candidate_count=max(1, recovery_candidate_count), - attempt_status="refresh_needs_review", + attempt_status="protected_existing_match", best_product=best_product, best_score=score, diagnostics=diagnostics, @@ -1571,7 +1607,7 @@ class CompetitorPriceFeeder: momo_price=momo_price, search_terms=attempt_terms, candidate_count=candidate_count, - attempt_status="refresh_needs_review", + attempt_status="protected_existing_match", best_product=best_product, best_score=score, diagnostics=diagnostics, @@ -1611,7 +1647,7 @@ class CompetitorPriceFeeder: attempts_written += 1 continue - attempt_status = "identity_veto" if getattr(diagnostics, "hard_veto", False) else "refresh_low_score" + attempt_status = _classify_low_score_attempt(score, diagnostics) self._record_match_attempt( sku, momo_name, @@ -1651,7 +1687,7 @@ class CompetitorPriceFeeder: momo_price=momo_price, search_terms=search_terms, candidate_count=1, - attempt_status="refresh_needs_review", + attempt_status="protected_existing_match", best_product=best_product, best_score=score, diagnostics=diagnostics, diff --git a/services/marketplace_product_matcher.py b/services/marketplace_product_matcher.py index 49b33b0..3de1430 100644 --- a/services/marketplace_product_matcher.py +++ b/services/marketplace_product_matcher.py @@ -244,6 +244,7 @@ SEARCH_NOISE_TOKENS = { } SEARCH_IDENTITY_ANCHORS = ( + "時尚潮流美甲片", "止汗爽身噴霧", "止汗爽身乳膏pro", "零粉感超持久粉底棒", @@ -258,6 +259,7 @@ SEARCH_IDENTITY_ANCHORS = ( "裸光幻閃亮采餅", "絕對持久定妝噴霧", "兒童防曬氣墊粉餅", + "勝過眼皮十色眼影盤", "提提亮膚打亮液", "甜甜嫩頰腮紅液", "自動武士刀眉筆", @@ -330,6 +332,22 @@ SEARCH_BROAD_ANCHORS = { "香氛融蠟燈", } +VARIANT_SENSITIVE_KEYWORDS = { + "美甲片", + "眼影盤", + "唇釉", + "唇膏", + "唇凍", + "潤唇膏", + "眉筆", + "眼線筆", + "腮紅液", + "打亮液", + "蜜粉餅", + "粉底棒", + "遮瑕棒", +} + SEARCH_AMBIGUOUS_PRODUCT_TERMS = { "保護膜", "保護貼", @@ -368,6 +386,11 @@ PRODUCT_TYPES = { "潔膚露": ("潔膚露", "浴潔露", "護潔露", "沐浴露", "wash"), "私密噴霧": ("私密噴霧", "抑菌噴霧", "醒肌抑菌噴霧"), "私密凝露": ("凝露", "激淨凝露", "緊實凝露", "亮白凝露"), + "氣墊粉餅": ("氣墊粉餅", "cushion"), + "眼影盤": ("眼影盤",), + "打亮液": ("打亮液",), + "腮紅液": ("腮紅液",), + "護唇膏": ("護唇膏", "潤唇膏"), "唇釉": ("唇釉", "唇彩", "lip tint", "lip glaze"), "粉底棒": ("粉底棒", "foundation stick"), "精華": ("精華", "精華液", "essence", "serum", "安瓶"), @@ -1381,6 +1404,20 @@ def score_marketplace_match( ): score += 0.02 reasons.append("shared_identity_anchor_core_line") + if ( + shared_anchor + and len(shared_anchor.replace(" ", "")) >= 6 + and brand_score >= 0.95 + and not hard_veto + and price_penalty == 0 + and type_score >= 0.55 + and spec_score >= 0.45 + and token_score >= 0.58 + and sequence_score >= 0.50 + and not variant_descriptor_conflict + ): + score += 0.025 + reasons.append("shared_identity_anchor_variant_safe") if ( brand_score >= 0.95 and not hard_veto @@ -1403,7 +1440,7 @@ def score_marketplace_match( ): score += 0.04 reasons.append("shared_model_token") - if variant_descriptor_conflict and spec_score < 0.85 and not shared_anchor and not shared_models: + if variant_descriptor_conflict and spec_score < 0.85: score -= 0.05 reasons.append("variant_descriptor_conflict") if ( @@ -1549,8 +1586,27 @@ def _variant_descriptors(identity: ProductIdentity) -> set[str]: return {token for token in descriptors if token} +def _is_variant_sensitive_identity( + left: ProductIdentity, + right: ProductIdentity, + shared_anchor: str, +) -> bool: + corpus = ( + shared_anchor, + left.product_type or "", + right.product_type or "", + left.searchable_name, + right.searchable_name, + ) + return any(keyword in text for text in corpus for keyword in VARIANT_SENSITIVE_KEYWORDS if text) + + def _has_variant_descriptor_conflict(left: ProductIdentity, right: ProductIdentity, shared_anchor: str) -> bool: - if shared_anchor and shared_anchor not in SEARCH_BROAD_ANCHORS: + if ( + shared_anchor + and shared_anchor not in SEARCH_BROAD_ANCHORS + and not _is_variant_sensitive_identity(left, right, shared_anchor) + ): return False if _shared_model_tokens(left, right): return False @@ -1558,7 +1614,13 @@ def _has_variant_descriptor_conflict(left: ProductIdentity, right: ProductIdenti right_descriptors = _variant_descriptors(right) if not left_descriptors or not right_descriptors: return False - return not bool(left_descriptors & right_descriptors) + if left_descriptors & right_descriptors: + return False + for left_descriptor in left_descriptors: + for right_descriptor in right_descriptors: + if left_descriptor in right_descriptor or right_descriptor in left_descriptor: + return False + return True def _search_core_score(token: str, all_tokens: set[str]) -> tuple[int, int, str]: @@ -1636,6 +1698,12 @@ def build_search_terms(name: str, max_terms: int = 3) -> list[str]: terms: list[str] = [] def primary_brand_phrase() -> str: + if {"dashing", "diva"} <= identity.brand_tokens: + return "dashing diva" + if {"rom", "nd"} <= identity.brand_tokens: + return "romand" + if {"im", "meme"} <= identity.brand_tokens: + return "im meme" chinese = sorted( (token for token in identity.brand_tokens if re.search(r"[\u4e00-\u9fff]", token)), key=lambda token: (-len(token), token), @@ -1656,6 +1724,8 @@ def build_search_terms(name: str, max_terms: int = 3) -> list[str]: core_phrases = _ranked_search_core_phrases(identity, limit=4) core_short = " ".join(core_phrases[:2]) core_primary = core_phrases[0] if core_phrases else "" + variant_descriptors = sorted(_variant_descriptors(identity), key=lambda token: (len(token), token)) + variant_primary = variant_descriptors[0] if variant_descriptors else "" model_phrases = [ phrase for phrase in core_phrases[1:] @@ -1665,7 +1735,11 @@ def build_search_terms(name: str, max_terms: int = 3) -> list[str]: primary_with_model = " ".join( part for part in (core_primary, model_phrases[0] if model_phrases else "") if part ) + variant_sensitive = any(keyword in identity.searchable_name for keyword in VARIANT_SENSITIVE_KEYWORDS) for value in ( + " ".join(part for part in (brand_part, core_primary, variant_primary, spec_part) if part) + if variant_sensitive and variant_primary + else "", " ".join(part for part in (brand_part, primary_with_model, spec_part) if part), " ".join(part for part in (brand_part, core_short, spec_part) if part), " ".join(part for part in (brand_part, core_short) if part), diff --git a/services/openclaw_strategist_service.py b/services/openclaw_strategist_service.py index 8f2551b..84559d0 100644 --- a/services/openclaw_strategist_service.py +++ b/services/openclaw_strategist_service.py @@ -589,7 +589,7 @@ def _fetch_competitor_summary() -> Dict[str, Any]: ) SELECT SUM(CASE WHEN attempt_status IN ('unit_comparable', 'refresh_unit_comparable') THEN 1 ELSE 0 END) AS unit_comparable_count, - SUM(CASE WHEN attempt_status IN ('unit_comparable', 'refresh_unit_comparable', 'identity_veto', 'low_score', 'expired_match', 'no_result', 'refresh_no_result') THEN 1 ELSE 0 END) AS review_queue_count + SUM(CASE WHEN attempt_status IN ('unit_comparable', 'refresh_unit_comparable', 'identity_veto', 'low_score', 'refresh_low_score', 'recoverable_low_score', 'true_low_confidence', 'protected_existing_match', 'expired_match', 'no_result', 'refresh_no_result') THEN 1 ELSE 0 END) AS review_queue_count FROM latest_attempt """)).fetchone() return { diff --git a/tests/test_competitor_match_attempts_persistence.py b/tests/test_competitor_match_attempts_persistence.py index 03c6639..083df6a 100644 --- a/tests/test_competitor_match_attempts_persistence.py +++ b/tests/test_competitor_match_attempts_persistence.py @@ -18,14 +18,16 @@ def test_competitor_feeder_persists_all_match_attempt_outcomes(): assert "INSERT INTO competitor_match_attempts" in source assert "CAST(:search_terms AS jsonb)" in source assert 'attempt_status="matched"' in source - assert '"low_score"' in source + assert '"recoverable_low_score"' in source + assert '"true_low_confidence"' in source assert '"identity_veto"' in source assert 'attempt_status="no_result"' in source assert 'attempt_status="no_match"' in source assert 'attempt_status="error"' in source assert "_search_pchome_candidates(crawler, momo_name, search_terms, momo_price=momo_price)" in source - assert 'attempt_status="needs_review"' in source + assert 'attempt_status="protected_existing_match"' in source assert "_should_upsert_competitor_price" in source + assert "_classify_low_score_attempt" in source assert "replace_legacy_unverified" in source assert "identity_v2" in source assert "_fetch_expired_identity_skus" in source @@ -35,8 +37,7 @@ def test_competitor_feeder_persists_all_match_attempt_outcomes(): retryable_source = source.split("def _fetch_retryable_candidate_skus", 1)[1].split( "def _fetch_expired_identity_skus", 1 )[0] - assert "la.attempt_status = 'low_score'" in retryable_source - assert "refresh_low_score')" not in retryable_source + assert "la.attempt_status IN ('low_score', 'refresh_low_score', 'recoverable_low_score')" in retryable_source latest_attempt_source = retryable_source.split("latest_attempt AS", 1)[1].split( "SELECT\n lm.product_id", 1 )[0] @@ -144,7 +145,7 @@ def test_reject_review_expires_current_formal_price(): best_competitor_price, best_match_score, error_message, attempted_at) VALUES ('A005', 'pchome', 1, '舒特膚 AD 乳液 200ml', 980, - '[]', 1, 'needs_review', + '[]', 1, 'protected_existing_match', 'DDAB01-REJECT', '舒特膚 AD 乳液 200ml', 899, 0.84, 'score=0.84', '2026-05-20 09:10:00') """)) @@ -408,6 +409,134 @@ def test_competitor_feeder_splits_hard_veto_from_low_score(monkeypatch): assert attempts[0]["diagnostics"].hard_veto is True +def test_competitor_feeder_marks_near_threshold_same_line_as_recoverable(monkeypatch): + from services.competitor_price_feeder import CompetitorPriceFeeder + from services.pchome_crawler import PChomeProduct + + product = PChomeProduct( + product_id="DDAB01-RECOVERABLE", + name="Recipe Box 韓兔 兒童防曬氣墊粉餅", + price=699, + original_price=799, + discount=12, + image_url="", + product_url="https://24h.pchome.com.tw/prod/DDAB01-RECOVERABLE", + stock=20, + store="24h", + rating=4.7, + review_count=8, + is_on_sale=True, + crawled_at=datetime.now(), + ) + + class FakeCrawler: + def __init__(self, *_args, **_kwargs): + pass + + def search_products(self, *_args, **_kwargs): + return True, "ok", [product] + + def fake_score(*_args, **_kwargs): + return SimpleNamespace( + score=0.754, + brand_score=1.0, + token_score=0.59, + spec_score=0.55, + sequence_score=0.53, + type_score=1.0, + price_penalty=0.0, + hard_veto=False, + reasons=("strong_product_line_match",), + comparison_mode="exact_identity", + tags=["identity_v2", "comparison_exact_identity", "brand_match"], + ) + + monkeypatch.setattr("services.pchome_crawler.PChomeCrawler", FakeCrawler) + monkeypatch.setattr("services.marketplace_product_matcher.score_marketplace_match", fake_score) + feeder = CompetitorPriceFeeder(engine=object()) + attempts = [] + monkeypatch.setattr( + feeder, + "_record_match_attempt", + lambda *args, **kwargs: attempts.append(kwargs), + ) + + result = feeder._run_sku_items([{ + "sku": "RB001", + "name": "【Recipebox】Recipe Box兒童防曬氣墊粉餅(兒童化妝品/無毒防曬粉餅/天然彩妝)", + "product_id": 8, + "momo_price": 699, + }]) + + assert result.matched == 0 + assert result.skipped_low_score == 1 + assert attempts[0]["attempt_status"] == "recoverable_low_score" + + +def test_competitor_feeder_marks_weak_identity_as_true_low_confidence(monkeypatch): + from services.competitor_price_feeder import CompetitorPriceFeeder + from services.pchome_crawler import PChomeProduct + + product = PChomeProduct( + product_id="DDAB01-WEAK", + name="韓系彩妝 十色眼影盤", + price=499, + original_price=699, + discount=28, + image_url="", + product_url="https://24h.pchome.com.tw/prod/DDAB01-WEAK", + stock=20, + store="24h", + rating=4.2, + review_count=8, + is_on_sale=True, + crawled_at=datetime.now(), + ) + + class FakeCrawler: + def __init__(self, *_args, **_kwargs): + pass + + def search_products(self, *_args, **_kwargs): + return True, "ok", [product] + + def fake_score(*_args, **_kwargs): + return SimpleNamespace( + score=0.733, + brand_score=0.95, + token_score=0.51, + spec_score=0.45, + sequence_score=0.44, + type_score=0.55, + price_penalty=0.0, + hard_veto=False, + reasons=(), + comparison_mode="exact_identity", + tags=["identity_v2", "comparison_exact_identity"], + ) + + monkeypatch.setattr("services.pchome_crawler.PChomeCrawler", FakeCrawler) + monkeypatch.setattr("services.marketplace_product_matcher.score_marketplace_match", fake_score) + feeder = CompetitorPriceFeeder(engine=object()) + attempts = [] + monkeypatch.setattr( + feeder, + "_record_match_attempt", + lambda *args, **kwargs: attempts.append(kwargs), + ) + + result = feeder._run_sku_items([{ + "sku": "RM001", + "name": "【rom&nd】勝過眼皮十色眼影盤", + "product_id": 9, + "momo_price": 499, + }]) + + assert result.matched == 0 + assert result.skipped_low_score == 1 + assert attempts[0]["attempt_status"] == "true_low_confidence" + + def test_should_upsert_allows_same_identity_candidate_to_replace_lower_score(): from sqlalchemy import create_engine, text @@ -455,6 +584,83 @@ def test_should_upsert_allows_same_identity_candidate_to_replace_lower_score(): assert reason.startswith("replace_same_identity_better_score=0.788->0.811") +def test_competitor_feeder_marks_existing_stronger_match_as_protected(monkeypatch): + from services.competitor_price_feeder import CompetitorPriceFeeder + from services.pchome_crawler import PChomeProduct + + product = PChomeProduct( + product_id="DDAB01-NEW", + name="PONY EFFECT 絕對持久定妝噴霧", + price=599, + original_price=699, + discount=14, + image_url="", + product_url="https://24h.pchome.com.tw/prod/DDAB01-NEW", + stock=20, + store="24h", + rating=4.7, + review_count=8, + is_on_sale=True, + crawled_at=datetime.now(), + ) + + class FakeCrawler: + def __init__(self, *_args, **_kwargs): + pass + + def search_products(self, *_args, **_kwargs): + return True, "ok", [product] + + def fake_score(*_args, **_kwargs): + return SimpleNamespace( + score=0.781, + brand_score=1.0, + token_score=0.79, + spec_score=0.55, + sequence_score=0.68, + type_score=0.55, + price_penalty=0.0, + hard_veto=False, + reasons=("shared_identity_anchor_packaging_variant",), + comparison_mode="exact_identity", + tags=["identity_v2", "comparison_exact_identity", "brand_match"], + ) + + monkeypatch.setattr("services.pchome_crawler.PChomeCrawler", FakeCrawler) + monkeypatch.setattr("services.marketplace_product_matcher.score_marketplace_match", fake_score) + feeder = CompetitorPriceFeeder(engine=object()) + attempts = [] + writes = [] + monkeypatch.setattr( + feeder, + "_record_match_attempt", + lambda *args, **kwargs: attempts.append(kwargs), + ) + monkeypatch.setattr( + feeder, + "_should_upsert_competitor_price", + lambda *_args, **_kwargs: (False, "existing_match_conflict;existing_score=0.948;incoming_score=0.781"), + ) + monkeypatch.setattr( + feeder, + "_upsert_competitor_price", + lambda *args, **kwargs: writes.append((args, kwargs)), + ) + + result = feeder._run_sku_items([{ + "sku": "14133077", + "name": "【PONY EFFECT】絕對持久定妝噴霧", + "product_id": 10, + "momo_price": 599, + }]) + + assert result.matched == 0 + assert result.skipped_low_score == 1 + assert writes == [] + assert attempts[0]["attempt_status"] == "protected_existing_match" + assert "existing_match_conflict" in attempts[0]["error_message"] + + def test_search_candidates_does_not_stop_on_merely_acceptable_match(monkeypatch): from services.competitor_price_feeder import _search_pchome_candidates from services.pchome_crawler import PChomeProduct diff --git a/tests/test_marketplace_product_matcher.py b/tests/test_marketplace_product_matcher.py index 3e95f3c..f375bb8 100644 --- a/tests/test_marketplace_product_matcher.py +++ b/tests/test_marketplace_product_matcher.py @@ -573,6 +573,22 @@ def test_marketplace_matcher_does_not_promote_different_option_without_spec(): assert diagnostics.score < 0.76 assert "strong_exact_spec_match" not in diagnostics.reasons + assert "variant_descriptor_conflict" in diagnostics.reasons + + +def test_marketplace_matcher_promotes_variant_safe_exact_option(): + from services.marketplace_product_matcher import score_marketplace_match + + diagnostics = score_marketplace_match( + "【DASHING DIVA】MAGICPRESS時尚潮流美甲片_極光之藍", + "Dashing Diva/F 時尚潮流美甲片-極光之藍 MDF5F001AG", + momo_price=331, + competitor_price=420, + ) + + assert diagnostics.score >= 0.76 + assert diagnostics.hard_veto is False + assert "shared_identity_anchor_variant_safe" in diagnostics.reasons def test_marketplace_matcher_promotes_shared_identity_anchor_near_threshold(): @@ -649,6 +665,36 @@ def test_marketplace_matcher_promotes_shared_anchor_without_spec_conflict(): assert "shared_identity_anchor_no_spec" in diagnostics.reasons +def test_marketplace_matcher_promotes_recipe_box_near_threshold_with_variant_safe_anchor(): + from services.marketplace_product_matcher import score_marketplace_match + + diagnostics = score_marketplace_match( + "【Recipebox】Recipe Box兒童防曬氣墊粉餅(兒童化妝品/無毒防曬粉餅/天然彩妝)", + "Recipe Box 韓兔 兒童防曬氣墊粉餅", + momo_price=699, + competitor_price=699, + ) + + assert diagnostics.score >= 0.76 + assert diagnostics.hard_veto is False + assert "shared_identity_anchor_variant_safe" in diagnostics.reasons + + +def test_marketplace_matcher_promotes_romand_palette_exact_line(): + from services.marketplace_product_matcher import score_marketplace_match + + diagnostics = score_marketplace_match( + "【rom&nd】勝過眼皮十色眼影盤", + "rom&nd X ZO&FRIENDS 勝過眼皮十色眼影盤 8g/7g", + momo_price=499, + competitor_price=499, + ) + + assert diagnostics.score >= 0.76 + assert diagnostics.hard_veto is False + assert "shared_identity_anchor_variant_safe" in diagnostics.reasons + + def test_marketplace_matcher_promotes_shared_model_token_for_exact_model(): from services.marketplace_product_matcher import score_marketplace_match @@ -735,6 +781,16 @@ def test_marketplace_search_terms_keep_professional_product_phrase(): assert not any("卸除防曬" in term or "外出清潔" in term for term in mustela_terms) +def test_marketplace_search_terms_keep_variant_descriptor_for_sensitive_lines(): + from services.marketplace_product_matcher import build_search_terms + + dashing_terms = build_search_terms("【DASHING DIVA】MAGICPRESS時尚潮流美甲片_極光之藍", max_terms=5) + romand_terms = build_search_terms("【rom&nd】勝過眼皮十色眼影盤", max_terms=5) + + assert dashing_terms[0] == "dashing diva 時尚潮流美甲片 極光之藍" + assert romand_terms[0] == "romand 勝過眼皮十色眼影盤" + + def test_marketplace_search_terms_prefer_specific_line_over_generic_usage_words(): from services.marketplace_product_matcher import build_search_terms