ai-saas / saas-search

Commit 76e1f08890671791eb5c646c9f1e1c47c9f1e9dd

Authored by tangwang
1 parent a73a751f

1. 减少一列sell points。有时候大模型输出会将这最后两列混淆，因此干脆去掉一个 

2. 优化缓存，缓存粒度为商品级，每次只对batch中未cache的重新计算；key使用每个商品输入的hash
Showing 6 changed files with 119 additions and 72 deletions   Show diff stats
Inline Side-by-side
indexer/ANCHORS_AND_SEMANTIC_ATTRIBUTES.md
  Show/Hide comments   View file @76e1f08
... ... @@ -135,7 +135,6 @@ SUPPORTED_LANGS = set(LANG_LABELS.keys())
135 135 "key_attributes": "<逗号分隔的关键属性>",
136 136 "material": "<逗号分隔的材质说明>",
137 137 "features": "<逗号分隔的功能特点>",
138   - "selling_points": "<一句话卖点>",
139 138 "anchor_text": "<逗号分隔的锚文本短语>",
140 139 # 若发生错误，还会附带:
141 140 # "error": "<异常信息>"
... ...
indexer/product_enrich.py
  Show/Hide comments   View file @76e1f08
... ... @@ -425,8 +425,7 @@ def parse_markdown_table(markdown_content: str) -&gt; List[Dict[str, str]]:
425 425 "key_attributes": parts[7] if len(parts) > 7 else "", # 关键属性
426 426 "material": parts[8] if len(parts) > 8 else "", # 材质说明
427 427 "features": parts[9] if len(parts) > 9 else "", # 功能特点
428   - "selling_points": parts[10] if len(parts) > 10 else "", # 商品卖点
429   - "anchor_text": parts[11] if len(parts) > 11 else "", # 锚文本
  428 + "anchor_text": parts[10] if len(parts) > 10 else "", # 锚文本
430 429 }
431 430 data.append(row)
432 431  
... ... @@ -503,7 +502,6 @@ def process_batch(
503 502 "key_attributes": "",
504 503 "material": "",
505 504 "features": "",
506   - "selling_points": "",
507 505 "anchor_text": "",
508 506 "error": f"prompt_creation_failed: unsupported target_lang={target_lang}",
509 507 }
... ... @@ -544,7 +542,6 @@ def process_batch(
544 542 "key_attributes": parsed_item.get("key_attributes", ""), # 关键属性
545 543 "material": parsed_item.get("material", ""), # 材质说明
546 544 "features": parsed_item.get("features", ""), # 功能特点
547   - "selling_points": parsed_item.get("selling_points", ""), # 商品卖点
548 545 "anchor_text": parsed_item.get("anchor_text", ""), # 锚文本
549 546 }
550 547 results_with_ids.append(result)
... ... @@ -586,7 +583,6 @@ def process_batch(
586 583 "key_attributes": "",
587 584 "material": "",
588 585 "features": "",
589   - "selling_points": "",
590 586 "anchor_text": "",
591 587 "error": str(e),
592 588 }
... ... @@ -614,39 +610,48 @@ def analyze_products(
614 610 if not products:
615 611 return []
616 612  
617   - # 简单路径：索引阶段通常 batch_size=1，这里优先做单条缓存命中
618   - if len(products) == 1:
619   - p = products[0]
620   - title = str(p.get("title") or "").strip()
621   - if title:
622   - cached = _get_cached_anchor_result(title, target_lang, tenant_id=tenant_id)
623   - if cached:
624   - logger.info(
625   - f"[analyze_products] Cache hit for title='{title[:50]}...', "
626   - f"lang={target_lang}, tenant_id={tenant_id or 'global'}"
627   - )
628   - return [cached]
  613 + results_by_index: List[Optional[Dict[str, Any]]] = [None] * len(products)
  614 + uncached_items: List[Tuple[int, Dict[str, str]]] = []
  615 +
  616 + for idx, product in enumerate(products):
  617 + title = str(product.get("title") or "").strip()
  618 + if not title:
  619 + uncached_items.append((idx, product))
  620 + continue
  621 +
  622 + cached = _get_cached_anchor_result(title, target_lang, tenant_id=tenant_id)
  623 + if cached:
  624 + logger.info(
  625 + f"[analyze_products] Cache hit for title='{title[:50]}...', "
  626 + f"lang={target_lang}, tenant_id={tenant_id or 'global'}"
  627 + )
  628 + results_by_index[idx] = cached
  629 + continue
  630 +
  631 + uncached_items.append((idx, product))
  632 +
  633 + if not uncached_items:
  634 + return [item for item in results_by_index if item is not None]
629 635  
630 636 # call_llm 一次处理上限固定为 BATCH_SIZE（默认 20）：
631 637 # - 尽可能攒批处理；
632 638 # - 即便调用方传入更大的 batch_size，也会自动按上限拆批。
633 639 req_bs = BATCH_SIZE if batch_size is None else int(batch_size)
634 640 bs = max(1, min(req_bs, BATCH_SIZE))
635   - all_results: List[Dict[str, Any]] = []
636   - total_batches = (len(products) + bs - 1) // bs
  641 + total_batches = (len(uncached_items) + bs - 1) // bs
637 642  
638   - for i in range(0, len(products), bs):
  643 + for i in range(0, len(uncached_items), bs):
639 644 batch_num = i // bs + 1
640   - batch = products[i : i + bs]
  645 + batch_slice = uncached_items[i : i + bs]
  646 + batch = [item for _, item in batch_slice]
641 647 logger.info(
642 648 f"[analyze_products] Processing batch {batch_num}/{total_batches}, "
643 649 f"size={len(batch)}, target_lang={target_lang}"
644 650 )
645 651 batch_results = process_batch(batch, batch_num=batch_num, target_lang=target_lang)
646   - all_results.extend(batch_results)
647 652  
648   - # 写入缓存
649   - for item in batch_results:
  653 + for (original_idx, _), item in zip(batch_slice, batch_results):
  654 + results_by_index[original_idx] = item
650 655 title_input = str(item.get("title_input") or "").strip()
651 656 if not title_input:
652 657 continue
... ... @@ -659,4 +664,4 @@ def analyze_products(
659 664 # 已在内部记录 warning
660 665 pass
661 666  
662   - return all_results
  667 + return [item for item in results_by_index if item is not None]
... ...
indexer/product_enrich_prompts.py
  Show/Hide comments   View file @76e1f08
... ... @@ -19,8 +19,7 @@ SHARED_ANALYSIS_INSTRUCTION = &quot;&quot;&quot;Analyze each input product title and fill these
19 19 7. Key attributes
20 20 8. Material description
21 21 9. Functional features
22   -10. Selling point: one concise core selling phrase
23   -11. Anchor text: a set of search-oriented words or phrases covering category, attributes, scenes, and demand
  22 +10. Anchor text: a search-focused set of keywords, selling points, and phrases covering categories, attributes, usage scenarios, and user intent
24 23  
25 24 Rules:
26 25 - Keep the input order and row count exactly the same.
... ... @@ -46,7 +45,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
46 45 "Key attributes",
47 46 "Material",
48 47 "Features",
49   - "Selling point",
50 48 "Anchor text"
51 49 ],
52 50 "zh": [
... ... @@ -60,7 +58,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
60 58 "关键属性",
61 59 "材质说明",
62 60 "功能特点",
63   - "商品卖点",
64 61 "锚文本"
65 62 ],
66 63 "zh_tw": [
... ... @@ -74,7 +71,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
74 71 "關鍵屬性",
75 72 "材質說明",
76 73 "功能特點",
77   - "商品賣點",
78 74 "錨文本"
79 75 ],
80 76 "ru": [
... ... @@ -88,7 +84,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
88 84 "Ключевые атрибуты",
89 85 "Материал",
90 86 "Особенности",
91   - "Преимущество товара",
92 87 "Анкорный текст"
93 88 ],
94 89 "ja": [
... ... @@ -102,7 +97,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
102 97 "主要属性",
103 98 "素材",
104 99 "機能特徴",
105   - "商品の訴求点",
106 100 "アンカーテキスト"
107 101 ],
108 102 "ko": [
... ... @@ -116,7 +110,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
116 110 "핵심 속성",
117 111 "소재",
118 112 "기능 특징",
119   - "상품 포인트",
120 113 "앵커 텍스트"
121 114 ],
122 115 "es": [
... ... @@ -130,7 +123,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
130 123 "Atributos clave",
131 124 "Material",
132 125 "Caracteristicas",
133   - "Punto de venta",
134 126 "Texto ancla"
135 127 ],
136 128 "fr": [
... ... @@ -144,7 +136,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
144 136 "Attributs cles",
145 137 "Matiere",
146 138 "Caracteristiques",
147   - "Argument de vente",
148 139 "Texte d'ancrage"
149 140 ],
150 141 "pt": [
... ... @@ -158,7 +149,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
158 149 "Atributos principais",
159 150 "Material",
160 151 "Caracteristicas",
161   - "Ponto de venda",
162 152 "Texto ancora"
163 153 ],
164 154 "de": [
... ... @@ -172,7 +162,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
172 162 "Wichtige Attribute",
173 163 "Material",
174 164 "Funktionen",
175   - "Verkaufsargument",
176 165 "Ankertext"
177 166 ],
178 167 "it": [
... ... @@ -186,7 +175,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
186 175 "Attributi chiave",
187 176 "Materiale",
188 177 "Caratteristiche",
189   - "Punto di forza",
190 178 "Testo ancora"
191 179 ],
192 180 "th": [
... ... @@ -200,7 +188,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
200 188 "คุณสมบัติสำคัญ",
201 189 "วัสดุ",
202 190 "คุณสมบัติการใช้งาน",
203   - "จุดขายสินค้า",
204 191 "แองเคอร์เท็กซ์"
205 192 ],
206 193 "vi": [
... ... @@ -214,7 +201,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
214 201 "Thuoc tinh chinh",
215 202 "Chat lieu",
216 203 "Tinh nang",
217   - "Diem ban hang",
218 204 "Van ban neo"
219 205 ],
220 206 "id": [
... ... @@ -228,7 +214,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
228 214 "Atribut utama",
229 215 "Bahan",
230 216 "Fitur",
231   - "Nilai jual",
232 217 "Teks jangkar"
233 218 ],
234 219 "ms": [
... ... @@ -242,7 +227,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
242 227 "Atribut utama",
243 228 "Bahan",
244 229 "Ciri-ciri",
245   - "Nilai jual",
246 230 "Teks sauh"
247 231 ],
248 232 "ar": [
... ... @@ -256,7 +240,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
256 240 "السمات الرئيسية",
257 241 "المادة",
258 242 "الميزات",
259   - "نقطة البيع",
260 243 "نص الربط"
261 244 ],
262 245 "hi": [
... ... @@ -270,7 +253,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
270 253 "मुख्य गुण",
271 254 "सामग्री",
272 255 "विशेषताएं",
273   - "बिक्री बिंदु",
274 256 "एंकर टेक्स्ट"
275 257 ],
276 258 "he": [
... ... @@ -284,7 +266,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
284 266 "מאפיינים מרכזיים",
285 267 "חומר",
286 268 "תכונות",
287   - "נקודת מכירה",
288 269 "טקסט עוגן"
289 270 ],
290 271 "my": [
... ... @@ -298,7 +279,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
298 279 "အဓိကဂုဏ်သတ္တိများ",
299 280 "ပစ္စည်း",
300 281 "လုပ်ဆောင်ချက်များ",
301   - "အရောင်းထူးခြားချက်",
302 282 "အန်ကာစာသား"
303 283 ],
304 284 "ta": [
... ... @@ -312,7 +292,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
312 292 "முக்கிய பண்புகள்",
313 293 "பொருள்",
314 294 "அம்சங்கள்",
315   - "விற்பனை அம்சம்",
316 295 "ஆங்கர் உரை"
317 296 ],
318 297 "ur": [
... ... @@ -326,7 +305,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
326 305 "کلیدی خصوصیات",
327 306 "مواد",
328 307 "فیچرز",
329   - "فروختی نقطہ",
330 308 "اینکر ٹیکسٹ"
331 309 ],
332 310 "bn": [
... ... @@ -340,7 +318,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
340 318 "মূল বৈশিষ্ট্য",
341 319 "উপাদান",
342 320 "ফিচার",
343   - "বিক্রয় পয়েন্ট",
344 321 "অ্যাঙ্কর টেক্সট"
345 322 ],
346 323 "pl": [
... ... @@ -354,7 +331,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
354 331 "Kluczowe atrybuty",
355 332 "Material",
356 333 "Cechy",
357   - "Atut sprzedazowy",
358 334 "Tekst kotwicy"
359 335 ],
360 336 "nl": [
... ... @@ -368,7 +344,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
368 344 "Belangrijke kenmerken",
369 345 "Materiaal",
370 346 "Functies",
371   - "Verkooppunt",
372 347 "Ankertekst"
373 348 ],
374 349 "ro": [
... ... @@ -382,7 +357,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
382 357 "Atribute cheie",
383 358 "Material",
384 359 "Caracteristici",
385   - "Punct de vanzare",
386 360 "Text ancora"
387 361 ],
388 362 "tr": [
... ... @@ -396,7 +370,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
396 370 "Temel ozellikler",
397 371 "Malzeme",
398 372 "Ozellikler",
399   - "Satis noktasi",
400 373 "Capa metni"
401 374 ],
402 375 "km": [
... ... @@ -410,7 +383,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
410 383 "លក្ខណៈសម្បត្តិសំខាន់",
411 384 "សម្ភារៈ",
412 385 "មុខងារ",
413   - "ចំណុចលក់",
414 386 "អត្ថបទអង់ក័រ"
415 387 ],
416 388 "lo": [
... ... @@ -424,7 +396,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
424 396 "ຄຸນລັກສະນະສຳຄັນ",
425 397 "ວັດສະດຸ",
426 398 "ຄຸນສົມບັດ",
427   - "ຈຸດຂາຍ",
428 399 "ຂໍ້ຄວາມອັງເຄີ"
429 400 ],
430 401 "yue": [
... ... @@ -438,7 +409,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
438 409 "關鍵屬性",
439 410 "材質說明",
440 411 "功能特點",
441   - "商品賣點",
442 412 "錨文本"
443 413 ],
444 414 "cs": [
... ... @@ -452,7 +422,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
452 422 "Klicove atributy",
453 423 "Material",
454 424 "Vlastnosti",
455   - "Prodejni argument",
456 425 "Kotvici text"
457 426 ],
458 427 "el": [
... ... @@ -466,7 +435,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
466 435 "Βασικά χαρακτηριστικά",
467 436 "Υλικό",
468 437 "Λειτουργίες",
469   - "Σημείο πώλησης",
470 438 "Κείμενο άγκυρας"
471 439 ],
472 440 "sv": [
... ... @@ -480,7 +448,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
480 448 "Viktiga attribut",
481 449 "Material",
482 450 "Funktioner",
483   - "Saljpunkt",
484 451 "Ankartext"
485 452 ],
486 453 "hu": [
... ... @@ -494,7 +461,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
494 461 "Fo jellemzok",
495 462 "Anyag",
496 463 "Funkciok",
497   - "Ertekesitesi elony",
498 464 "Horgonyszoveg"
499 465 ],
500 466 "da": [
... ... @@ -508,7 +474,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
508 474 "Nogleattributter",
509 475 "Materiale",
510 476 "Funktioner",
511   - "Salgsargument",
512 477 "Ankertekst"
513 478 ],
514 479 "fi": [
... ... @@ -522,7 +487,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
522 487 "Keskeiset ominaisuudet",
523 488 "Materiaali",
524 489 "Ominaisuudet",
525   - "Myyntivaltti",
526 490 "Ankkuriteksti"
527 491 ],
528 492 "uk": [
... ... @@ -536,7 +500,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
536 500 "Ключові атрибути",
537 501 "Матеріал",
538 502 "Особливості",
539   - "Продаюча перевага",
540 503 "Анкорний текст"
541 504 ],
542 505 "bg": [
... ... @@ -550,7 +513,6 @@ LANGUAGE_MARKDOWN_TABLE_HEADERS: Dict[str, Dict[str, Any]] = {
550 513 "Ключови атрибути",
551 514 "Материал",
552 515 "Характеристики",
553   - "Търговско предимство",
554 516 "Анкор текст"
555 517 ]
556 518 }
557 519 \ No newline at end of file
... ...
tests/ci/test_service_api_contracts.py
  Show/Hide comments   View file @76e1f08
... ... @@ -366,7 +366,6 @@ def test_indexer_enrich_content_contract(indexer_client: TestClient, monkeypatch
366 366 "key_attributes": "",
367 367 "material": "",
368 368 "features": "",
369   - "selling_points": "",
370 369 "anchor_text": f"{target_lang}-anchor-{p['id']}",
371 370 }
372 371 for p in products
... ...
tests/test_process_products_batching.py
  Show/Hide comments   View file @76e1f08
... ... @@ -29,7 +29,6 @@ def test_analyze_products_caps_batch_size_to_20(monkeypatch):
29 29 "key_attributes": "",
30 30 "material": "",
31 31 "features": "",
32   - "selling_points": "",
33 32 "anchor_text": "",
34 33 }
35 34 for item in batch_data
... ... @@ -69,7 +68,6 @@ def test_analyze_products_uses_min_batch_size_1(monkeypatch):
69 68 "key_attributes": "",
70 69 "material": "",
71 70 "features": "",
72   - "selling_points": "",
73 71 "anchor_text": "",
74 72 }
75 73 for item in batch_data
... ...
tests/test_product_enrich_partial_mode.py
  Show/Hide comments   View file @76e1f08
... ... @@ -195,9 +195,9 @@ def test_call_llm_logs_shared_context_once_and_verbose_contains_full_requests():
195 195  
196 196  
197 197 def test_process_batch_reads_result_and_validates_expected_fields():
198   - merged_markdown = """| 序号 | 商品标题 | 品类路径 | 细分标签 | 适用人群 | 使用场景 | 适用季节 | 关键属性 | 材质说明 | 功能特点 | 商品卖点 | 锚文本 |
199   -|----|----|----|----|----|----|----|----|----|----|----|----|
200   -| 1 | 法式连衣裙 | 女装>连衣裙 | 法式,收腰 | 年轻女性 | 通勤,约会 | 春季,夏季 | 中长款 | 聚酯纤维 | 透气 | 修身显瘦 | 法式收腰连衣裙 |
  198 + merged_markdown = """| 序号 | 商品标题 | 品类路径 | 细分标签 | 适用人群 | 使用场景 | 适用季节 | 关键属性 | 材质说明 | 功能特点 | 锚文本 |
  199 +|----|----|----|----|----|----|----|----|----|----|----|
  200 +| 1 | 法式连衣裙 | 女装>连衣裙 | 法式,收腰 | 年轻女性 | 通勤,约会 | 春季,夏季 | 中长款 | 聚酯纤维 | 透气 | 法式收腰连衣裙 |
201 201 """
202 202  
203 203 with mock.patch.object(
... ... @@ -225,5 +225,89 @@ def test_process_batch_reads_result_and_validates_expected_fields():
225 225 assert row["key_attributes"] == "中长款"
226 226 assert row["material"] == "聚酯纤维"
227 227 assert row["features"] == "透气"
228   - assert row["selling_points"] == "修身显瘦"
229 228 assert row["anchor_text"] == "法式收腰连衣裙"
  229 +
  230 +
  231 +def test_analyze_products_uses_product_level_cache_across_batch_requests():
  232 + cache_store = {}
  233 + process_calls = []
  234 +
  235 + def fake_get_cached_anchor_result(title, target_lang, tenant_id=None):
  236 + return cache_store.get((tenant_id, target_lang, title))
  237 +
  238 + def fake_set_cached_anchor_result(title, target_lang, result, tenant_id=None):
  239 + cache_store[(tenant_id, target_lang, title)] = result
  240 +
  241 + def fake_process_batch(batch_data, batch_num, target_lang="zh"):
  242 + process_calls.append(
  243 + {
  244 + "batch_num": batch_num,
  245 + "target_lang": target_lang,
  246 + "titles": [item["title"] for item in batch_data],
  247 + }
  248 + )
  249 + return [
  250 + {
  251 + "id": item["id"],
  252 + "lang": target_lang,
  253 + "title_input": item["title"],
  254 + "title": f"normalized:{item['title']}",
  255 + "category_path": "cat",
  256 + "tags": "tags",
  257 + "target_audience": "audience",
  258 + "usage_scene": "scene",
  259 + "season": "season",
  260 + "key_attributes": "attrs",
  261 + "material": "material",
  262 + "features": "features",
  263 + "anchor_text": f"anchor:{item['title']}",
  264 + }
  265 + for item in batch_data
  266 + ]
  267 +
  268 + products = [
  269 + {"id": "1", "title": "dress"},
  270 + {"id": "2", "title": "shirt"},
  271 + ]
  272 +
  273 + with mock.patch.object(product_enrich, "API_KEY", "fake-key"), mock.patch.object(
  274 + product_enrich,
  275 + "_get_cached_anchor_result",
  276 + side_effect=fake_get_cached_anchor_result,
  277 + ), mock.patch.object(
  278 + product_enrich,
  279 + "_set_cached_anchor_result",
  280 + side_effect=fake_set_cached_anchor_result,
  281 + ), mock.patch.object(
  282 + product_enrich,
  283 + "process_batch",
  284 + side_effect=fake_process_batch,
  285 + ):
  286 + first = product_enrich.analyze_products(
  287 + [products[0]],
  288 + target_lang="zh",
  289 + tenant_id="170",
  290 + )
  291 + second = product_enrich.analyze_products(
  292 + products,
  293 + target_lang="zh",
  294 + tenant_id="170",
  295 + )
  296 + third = product_enrich.analyze_products(
  297 + products,
  298 + target_lang="zh",
  299 + tenant_id="170",
  300 + )
  301 +
  302 + assert [row["title_input"] for row in first] == ["dress"]
  303 + assert [row["title_input"] for row in second] == ["dress", "shirt"]
  304 + assert [row["title_input"] for row in third] == ["dress", "shirt"]
  305 +
  306 + assert process_calls == [
  307 + {"batch_num": 1, "target_lang": "zh", "titles": ["dress"]},
  308 + {"batch_num": 1, "target_lang": "zh", "titles": ["shirt"]},
  309 + ]
  310 + assert second[0]["anchor_text"] == "anchor:dress"
  311 + assert second[1]["anchor_text"] == "anchor:shirt"
  312 + assert third[0]["anchor_text"] == "anchor:dress"
  313 + assert third[1]["anchor_text"] == "anchor:shirt"
... ...