File size: 36,638 Bytes
c8635ed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b8c7ffb
 
c8635ed
 
b8c7ffb
c8635ed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b8c7ffb
c8635ed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b8c7ffb
c8635ed
 
 
 
 
 
 
 
b8c7ffb
c8635ed
 
 
 
 
 
 
 
 
 
 
 
b8c7ffb
c8635ed
 
 
b8c7ffb
 
c8635ed
 
 
b8c7ffb
c8635ed
 
 
b8c7ffb
c8635ed
 
 
 
 
 
 
 
b8c7ffb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c8635ed
b8c7ffb
 
 
c8635ed
 
 
 
 
 
 
 
 
 
 
 
b8c7ffb
c8635ed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b8c7ffb
c8635ed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b8c7ffb
c8635ed
 
 
b8c7ffb
c8635ed
b8c7ffb
 
c8635ed
 
b8c7ffb
c8635ed
b8c7ffb
 
 
c8635ed
 
 
 
 
 
 
b8c7ffb
 
c8635ed
 
 
 
 
 
 
 
 
 
 
 
 
b8c7ffb
c8635ed
 
 
 
b8c7ffb
c8635ed
b8c7ffb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c8635ed
 
 
 
 
b8c7ffb
 
 
 
 
c8635ed
 
 
b8c7ffb
c8635ed
b8c7ffb
c8635ed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b8c7ffb
 
 
 
 
 
 
 
 
 
c8635ed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b8c7ffb
c8635ed
b8c7ffb
c8635ed
 
 
 
 
 
 
 
b8c7ffb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c8635ed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b8c7ffb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c8635ed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
import os
import torch
import pandas as pd
import logging
import faiss
import numpy as np
import time
import gensim
from fastapi import FastAPI, HTTPException, BackgroundTasks
from pydantic import BaseModel
from datasets import load_dataset
from huggingface_hub import login, hf_hub_download, HfApi, create_repo
from keybert import KeyBERT
from sentence_transformers import SentenceTransformer
from joblib import Parallel, delayed
from tqdm import tqdm
import tempfile
import re
import sys
import asyncio
from concurrent.futures import ThreadPoolExecutor, ProcessPoolExecutor

# โœ… ๋กœ๊ทธ ์„ค์ •
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

# โœ… ์Šค๋ ˆ๋“œ ํ’€ ์„ค์ • (๋น„๋™๊ธฐ ์ž‘์—…์„ ์œ„ํ•œ)
thread_pool = ThreadPoolExecutor(max_workers=min(64, os.cpu_count() * 4))


# โœ… FastAPI ์ธ์Šคํ„ด์Šค ์ƒ์„ฑ
app = FastAPI(title="๐Ÿš€ KeyBERT + Word2Vec ๊ธฐ๋ฐ˜ FAISS ๊ฒ€์ƒ‰ API", version="1.2")

# โœ… GPU ์‚ฌ์šฉ ์—ฌ๋ถ€ ํ™•์ธ
device = "cuda" if torch.cuda.is_available() else "cpu"
logger.info(f"๐Ÿš€ ์‹คํ–‰ ๋””๋ฐ”์ด์Šค: {device.upper()}")

# โœ… Hugging Face ๋กœ๊ทธ์ธ
HF_API_TOKEN = os.getenv("HF_API_TOKEN")
if HF_API_TOKEN:
    logger.info("๐Ÿ”‘ Hugging Face API ๋กœ๊ทธ์ธ ์ค‘...")
    login(token=HF_API_TOKEN)
else:
    logger.error("โŒ HF_API_TOKEN์ด ์„ค์ •๋˜์ง€ ์•Š์•˜์Šต๋‹ˆ๋‹ค. ์ผ๋ถ€ ๊ธฐ๋Šฅ์ด ์ œํ•œ๋  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.")

# โœ… Word2Vec ๋ชจ๋ธ ๋กœ๋“œ
word2vec_model = None
try:
    logger.info("๐Ÿ”„ Word2Vec ๋ชจ๋ธ ๋กœ๋“œ ์ค‘...")
    MODEL_REPO = "aikobay/item-model"
    model_path = hf_hub_download(repo_id=MODEL_REPO, filename="item_vectors.bin", repo_type="dataset")
    word2vec_model = gensim.models.KeyedVectors.load_word2vec_format(model_path, binary=True)
    logger.info(f"โœ… Word2Vec ๋ชจ๋ธ ๋กœ๋“œ ์™„๋ฃŒ! ๋‹จ์–ด ์ˆ˜: {len(word2vec_model.key_to_index)}")
except Exception as e:
    logger.error(f"โŒ Word2Vec ๋ชจ๋ธ ๋กœ๋“œ ์‹คํŒจ: {e}")

# โœ… KeyBERT ๋ชจ๋ธ ๋กœ๋“œ
logger.info("๐Ÿ”„ KeyBERT ๋ชจ๋ธ ๋กœ๋“œ ์ค‘...")
kw_model = KeyBERT("paraphrase-multilingual-MiniLM-L12-v2")
original_embedding_model = SentenceTransformer("paraphrase-multilingual-MiniLM-L12-v2")
logger.info("โœ… KeyBERT ๋ชจ๋ธ ๋กœ๋“œ ์™„๋ฃŒ!")

# โœ… ํ•œ๊ตญ์–ด ํŠนํ™” ์ž„๋ฒ ๋”ฉ ๋ชจ๋ธ๋กœ ๊ต์ฒด
embedding_model = None
try:
    logger.info("๐Ÿ”„ ํ•œ๊ตญ์–ด ํŠนํ™” ์ž„๋ฒ ๋”ฉ ๋ชจ๋ธ๋กœ ๊ต์ฒด ์‹œ๋„...")
    # ํ•œ๊ตญ์–ด ํŠนํ™” ๋ชจ๋ธ ๋กœ๋“œ ์‹œ๋„ (์‹คํŒจ์‹œ ๊ธฐ์กด ๋ชจ๋ธ ์œ ์ง€)
    embedding_model = SentenceTransformer("jhgan/ko-sroberta-multitask")
    logger.info("โœ… ํ•œ๊ตญ์–ด ํŠนํ™” ์ž„๋ฒ ๋”ฉ ๋ชจ๋ธ ๋กœ๋“œ ์™„๋ฃŒ!")
except Exception as e:
    logger.warning(f"โš ๏ธ ํ•œ๊ตญ์–ด ํŠนํ™” ๋ชจ๋ธ ๋กœ๋“œ ์‹คํŒจ, ๊ธฐ์กด ๋ชจ๋ธ ์œ ์ง€: {e}")
    embedding_model = original_embedding_model

# โœ… ์ง„ํ–‰ ์ค‘์ธ ๊ฒฝ๋งค ์ƒํ’ˆ ๋ฐ์ดํ„ฐ ๋กœ๋“œ
async def load_huggingface_jsonl(dataset_name, split="train"):
    """Hugging Face Hub์—์„œ ๋ฐ์ดํ„ฐ์…‹ ๋น„๋™๊ธฐ ๋กœ๋“œ"""
    try:
        # ์Šค๋ ˆ๋“œ ํ’€์—์„œ ์‹คํ–‰ํ•˜์—ฌ ๋น„๋™๊ธฐ ์ฒ˜๋ฆฌ
        loop = asyncio.get_event_loop()
        
        def _load_dataset():
            repo_id = f"aikobay/{dataset_name}"
            dataset = load_dataset(repo_id, split=split)
            return dataset.to_pandas().dropna()
        
        # ์Šค๋ ˆ๋“œ ํ’€์—์„œ ๋น„๋™๊ธฐ๋กœ ์‹คํ–‰
        df = await loop.run_in_executor(thread_pool, _load_dataset)
        return df
    except Exception as e:
        logger.error(f"โŒ ๋ฐ์ดํ„ฐ ๋กœ๋“œ ์ค‘ ์˜ค๋ฅ˜ ๋ฐœ์ƒ: {e}")
        return pd.DataFrame()

# ์ดˆ๊ธฐ ๋ฐ์ดํ„ฐ ๋กœ๋“œ - ๋น„๋™๊ธฐ ํ•จ์ˆ˜๋ฅผ ๋™๊ธฐ์ ์œผ๋กœ ํ˜ธ์ถœํ•˜์—ฌ ์‹œ์ž‘ ์‹œ ๋กœ๋“œ
active_sale_items = None
try:
    # ๋น„๋™๊ธฐ ํ•จ์ˆ˜๋ฅผ ์‹œ์ž‘ ์‹œ ์‹คํ–‰ํ•˜๊ธฐ ์œ„ํ•œ ์ž„์‹œ ์ด๋ฒคํŠธ ๋ฃจํ”„ ์‚ฌ์šฉ
    loop = asyncio.new_event_loop()
    active_sale_items = loop.run_until_complete(load_huggingface_jsonl("initial_saleitem_dataset"))
    loop.close()
    
    if active_sale_items.empty:
        logger.error("โŒ ๋ฐ์ดํ„ฐ์…‹์ด ๋น„์–ด ์žˆ์Šต๋‹ˆ๋‹ค. ํ”„๋กœ๊ทธ๋žจ์„ ์ข…๋ฃŒํ•ฉ๋‹ˆ๋‹ค.")
        exit(1)
    logger.info(f"โœ… ๊ฒฝ๋งค ์ƒํ’ˆ ๋ฐ์ดํ„ฐ ๋กœ๋“œ ์™„๋ฃŒ! ์ด {len(active_sale_items)}๊ฐœ ์ƒํ’ˆ")
except Exception as e:
    logger.error(f"โŒ ์ƒํ’ˆ ๋ฐ์ดํ„ฐ ๋กœ๋“œ ์‹คํŒจ: {e}")
    exit(1)

# โœ… FAISS ์ธ๋ฑ์Šค ์ดˆ๊ธฐํ™”
faiss_index = None
indexed_items = []

# โœ… ๋ฉ€ํ‹ฐ์ฝ”์–ด ๋ฒกํ„ฐํ™” ํ•จ์ˆ˜ 
async def encode_texts_parallel(texts, batch_size=1024):
    """GPU ํ™œ์šฉ + ๋ฐฐ์น˜ ์‚ฌ์ด์ฆˆ ์ตœ์ ํ™” ๋ฒกํ„ฐํ™” (๋Œ€๊ทœ๋ชจ ์„ฑ๋Šฅ ํ–ฅ์ƒ)"""
    if not texts:
        return np.array([]).astype("float32")
        
    # ๋ฐฐ์น˜ ํฌ๊ธฐ ์ฆ๊ฐ€๋กœ ์ฒ˜๋ฆฌ ํšจ์œจ ํ–ฅ์ƒ
    loop = asyncio.get_event_loop()
    
    def _encode_efficiently():
        # ๋ฒกํ„ฐํ™” ์ตœ์ ํ™” - GPU ํ™œ์šฉ + ๋ฐฐ์น˜ ์‚ฌ์ด์ฆˆ ์ตœ์ ํ™”
        return embedding_model.encode(
            texts, 
            batch_size=batch_size,
            convert_to_numpy=True,
            show_progress_bar=False,
            device=device  # GPU ์‚ฌ์šฉ
        )
    
    # ์Šค๋ ˆ๋“œ ํ’€์—์„œ ์‹คํ–‰
    embeddings = await loop.run_in_executor(thread_pool, _encode_efficiently)
    return embeddings.astype("float32")


# โœ… FAISS ์ธ๋ฑ์Šค ์ €์žฅ ํ•จ์ˆ˜ (Hugging Face Hub)
async def save_faiss_index():
    """FAISS ์ธ๋ฑ์Šค๋ฅผ Hugging Face Hub์— ์ €์žฅ (๋น„๋™๊ธฐ ์ง€์›)"""
    global faiss_index, indexed_items
    
    if faiss_index is None or not indexed_items:
        logger.error("โŒ ์ €์žฅํ•  FAISS ์ธ๋ฑ์Šค๊ฐ€ ์—†์Šต๋‹ˆ๋‹ค.")
        return False
    
    try:
        # ๋ ˆํฌ์ง€ํ† ๋ฆฌ ID
        repo_id = os.getenv("HF_INDEX_REPO", "aikobay/saleitem_faiss_index")
        
        # ๋น„๋™๊ธฐ ์ž‘์—…์„ ์œ„ํ•œ ๋ฃจํ”„
        loop = asyncio.get_event_loop()
        
        # ๋น„๋™๊ธฐ ์ž‘์—…์œผ๋กœ ์‹คํ–‰
        def _save_index():
            # HfApi ๊ฐ์ฒด ์ƒ์„ฑ
            api = HfApi()
            
            # ๋ ˆํฌ์ง€ํ† ๋ฆฌ ์กด์žฌ ์—ฌ๋ถ€ ํ™•์ธ ๋ฐ ์ƒ์„ฑ
            try:
                api.repo_info(repo_id=repo_id, repo_type="dataset")
                logger.info(f"โœ… ๊ธฐ์กด ๋ ˆํฌ์ง€ํ† ๋ฆฌ ์‚ฌ์šฉ: {repo_id}")
            except Exception:
                logger.info(f"๐Ÿ”„ ๋ ˆํฌ์ง€ํ† ๋ฆฌ๊ฐ€ ์กด์žฌํ•˜์ง€ ์•Š์•„ ์ƒˆ๋กœ ์ƒ์„ฑํ•ฉ๋‹ˆ๋‹ค: {repo_id}")
                create_repo(
                    repo_id=repo_id,
                    repo_type="dataset",
                    private=True,
                    exist_ok=True
                )
                logger.info(f"โœ… ๋ ˆํฌ์ง€ํ† ๋ฆฌ ์ƒ์„ฑ ์™„๋ฃŒ: {repo_id}")
            
            # ์ž„์‹œ ํŒŒ์ผ๋กœ ๋จผ์ € ๋กœ์ปฌ์— ์ €์žฅ
            with tempfile.TemporaryDirectory() as temp_dir:
                index_path = os.path.join(temp_dir, "faiss_index.bin")
                items_path = os.path.join(temp_dir, "indexed_items.txt")
                
                # FAISS ์ธ๋ฑ์Šค ์ €์žฅ
                faiss.write_index(faiss_index, index_path)
                
                # ์•„์ดํ…œ ๋ชฉ๋ก ์ €์žฅ
                with open(items_path, "w", encoding="utf-8") as f:
                    f.write("\n".join(indexed_items))
                
                # README ํŒŒ์ผ ์ƒ์„ฑ
                readme_path = os.path.join(temp_dir, "README.md")
                with open(readme_path, "w", encoding="utf-8") as f:
                    f.write(f"""# FAISS ์ธ๋ฑ์Šค ์ €์žฅ์†Œ
์ด ์ €์žฅ์†Œ๋Š” ์ƒํ’ˆ ๊ฒ€์ƒ‰์„ ์œ„ํ•œ FAISS ์ธ๋ฑ์Šค์™€ ๊ด€๋ จ ๋ฐ์ดํ„ฐ๋ฅผ ํฌํ•จํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.
- ์ตœ์ข… ์—…๋ฐ์ดํŠธ: {pd.Timestamp.now()}
- ์ธ๋ฑ์Šค ํ•ญ๋ชฉ ์ˆ˜: {len(indexed_items)}
- ๋ชจ๋ธ: KeyBERT + Word2Vec
์ด ์ €์žฅ์†Œ๋Š” 'aikobay/initial_saleitem_dataset'์˜ ์ƒํ’ˆ ๋ฐ์ดํ„ฐ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ์ƒ์„ฑ๋œ ๋ฒกํ„ฐ ์ธ๋ฑ์Šค๋ฅผ ์ €์žฅํ•˜๊ธฐ ์œ„ํ•ด ์ž๋™ ์ƒ์„ฑ๋˜์—ˆ์Šต๋‹ˆ๋‹ค.
""")
                
                # ํŒŒ์ผ ์—…๋กœ๋“œ
                for file_path, file_name in [
                    (index_path, "faiss_index.bin"),
                    (items_path, "indexed_items.txt"),
                    (readme_path, "README.md")
                ]:
                    api.upload_file(
                        path_or_fileobj=file_path,
                        path_in_repo=file_name,
                        repo_id=repo_id,
                        repo_type="dataset"
                    )
                
                logger.info(f"โœ… FAISS ์ธ๋ฑ์Šค๊ฐ€ Hugging Face Hub์— ์ €์žฅ๋˜์—ˆ์Šต๋‹ˆ๋‹ค. ๋ ˆํฌ: {repo_id}")
                return True
            
        # ์Šค๋ ˆ๋“œ ํ’€์—์„œ ๋น„๋™๊ธฐ์ ์œผ๋กœ ์‹คํ–‰
        result = await loop.run_in_executor(thread_pool, _save_index)
        return result
            
    except Exception as e:
        logger.error(f"โŒ FAISS ์ธ๋ฑ์Šค Hub ์ €์žฅ ์ค‘ ์˜ค๋ฅ˜ ๋ฐœ์ƒ: {e}")
        
        # ๋กœ์ปฌ์— ๋ฐฑ์—… ์ €์žฅ ์‹œ๋„
        try:
            loop = asyncio.get_event_loop()
            
            def _local_backup():
                local_path = os.path.join(os.getcwd(), "faiss_index.bin")
                faiss.write_index(faiss_index, local_path)
                with open("indexed_items.txt", "w", encoding="utf-8") as f:
                    f.write("\n".join(indexed_items))
                logger.info(f"โœ… FAISS ์ธ๋ฑ์Šค๊ฐ€ ๋กœ์ปฌ์— ๋ฐฑ์—… ์ €์žฅ๋˜์—ˆ์Šต๋‹ˆ๋‹ค: {local_path}")
                return True
            
            result = await loop.run_in_executor(thread_pool, _local_backup)
            return result
        except Exception as local_err:
            logger.error(f"โŒ ๋กœ์ปฌ ๋ฐฑ์—… ์ €์žฅ๋„ ์‹คํŒจ: {local_err}")
            return False

# โœ… FAISS ์ธ๋ฑ์Šค ๋กœ๋“œ ํ•จ์ˆ˜ (Hugging Face Hub)
async def load_faiss_index():
    """Hugging Face Hub์—์„œ FAISS ์ธ๋ฑ์Šค๋ฅผ ๋กœ๋“œ (๋น„๋™๊ธฐ ์ง€์›)"""
    global faiss_index, indexed_items
    
    # ๋ ˆํฌ์ง€ํ† ๋ฆฌ ID
    repo_id = os.getenv("HF_INDEX_REPO", "aikobay/saleitem_faiss_index")
    
    try:
        # ๋น„๋™๊ธฐ ์ž‘์—…์„ ์œ„ํ•œ ๋ฃจํ”„
        loop = asyncio.get_event_loop()
        
        # ๋น„๋™๊ธฐ ์ž‘์—…์œผ๋กœ ์‹คํ–‰
        def _load_index():
            # ๋ ˆํฌ์ง€ํ† ๋ฆฌ ์กด์žฌ ํ™•์ธ
            api = HfApi()
            try:
                api.repo_info(repo_id=repo_id, repo_type="dataset")
                logger.info(f"โœ… FAISS ์ธ๋ฑ์Šค ๋ ˆํฌ์ง€ํ† ๋ฆฌ ํ™•์ธ: {repo_id}")
            except Exception as repo_err:
                logger.warning(f"โš ๏ธ ๋ ˆํฌ์ง€ํ† ๋ฆฌ๊ฐ€ ์กด์žฌํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค: {repo_err}")
                raise FileNotFoundError("Hub ๋ ˆํฌ์ง€ํ† ๋ฆฌ๊ฐ€ ์กด์žฌํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค")
            
            # Hub์—์„œ ํŒŒ์ผ ๋‹ค์šด๋กœ๋“œ
            index_path = hf_hub_download(
                repo_id=repo_id,
                filename="faiss_index.bin",
                repo_type="dataset"
            )
            
            items_path = hf_hub_download(
                repo_id=repo_id,
                filename="indexed_items.txt",
                repo_type="dataset"
            )
            
            # ํŒŒ์ผ ๋กœ๋“œ
            loaded_index = faiss.read_index(index_path)
            with open(items_path, "r", encoding="utf-8") as f:
                loaded_items = f.read().splitlines()
                
            return loaded_index, loaded_items
        
        # ์Šค๋ ˆ๋“œ ํ’€์—์„œ ๋น„๋™๊ธฐ์ ์œผ๋กœ ์‹คํ–‰
        loaded_index, loaded_items = await loop.run_in_executor(thread_pool, _load_index)
        
        # ์ „์—ญ ๋ณ€์ˆ˜์— ํ• ๋‹น
        faiss_index = loaded_index
        indexed_items = loaded_items
        
        logger.info(f"โœ… FAISS ์ธ๋ฑ์Šค๊ฐ€ Hub์—์„œ ๋กœ๋“œ๋˜์—ˆ์Šต๋‹ˆ๋‹ค. ์ด {len(indexed_items)}๊ฐœ ์ƒํ’ˆ")
        return True
        
    except Exception as e:
        logger.warning(f"โš ๏ธ Hub์—์„œ FAISS ์ธ๋ฑ์Šค ๋กœ๋“œ ์ค‘ ์˜ค๋ฅ˜ ๋ฐœ์ƒ: {e}")
        
        # ๋กœ์ปฌ ํŒŒ์ผ ํ™•์ธ
        try:
            loop = asyncio.get_event_loop()
            
            def _load_local():
                local_index_path = "faiss_index.bin"
                local_items_path = "indexed_items.txt"
                
                if os.path.exists(local_index_path) and os.path.exists(local_items_path):
                    loaded_index = faiss.read_index(local_index_path)
                    with open(local_items_path, "r", encoding="utf-8") as f:
                        loaded_items = f.read().splitlines()
                    return loaded_index, loaded_items
                else:
                    logger.warning("โš ๏ธ ๋กœ์ปฌ FAISS ์ธ๋ฑ์Šค ํŒŒ์ผ์ด ์กด์žฌํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค.")
                    return None, None
            
            # ์Šค๋ ˆ๋“œ ํ’€์—์„œ ๋น„๋™๊ธฐ์ ์œผ๋กœ ์‹คํ–‰
            result = await loop.run_in_executor(thread_pool, _load_local)
            
            if result[0] is not None:
                faiss_index, indexed_items = result
                logger.info(f"โœ… ๋กœ์ปฌ FAISS ์ธ๋ฑ์Šค ๋กœ๋“œ ์„ฑ๊ณต. ์ด {len(indexed_items)}๊ฐœ ์ƒํ’ˆ")
                return True
            else:
                return False
                
        except Exception as local_err:
            logger.error(f"โŒ ๋กœ์ปฌ FAISS ์ธ๋ฑ์Šค ๋กœ๋“œ ์ค‘ ์˜ค๋ฅ˜: {local_err}")
            return False

# โœ… FAISS ์–‘์žํ™” ์ธ๋ฑ์Šค ๊ตฌ์ถ• ํ•จ์ˆ˜ (IVF ๊ธฐ๋ฐ˜์œผ๋กœ ๋ณ€๊ฒฝ)
async def rebuild_faiss_index():
    """FAISS ์ธ๋ฑ์Šค๋ฅผ IVF ๊ธฐ๋ฐ˜์œผ๋กœ ์ƒˆ๋กญ๊ฒŒ ๊ตฌ์ถ• (์†๋„ ์ตœ์ ํ™”)"""
    global faiss_index, indexed_items, active_sale_items

    logger.info("๐Ÿ”„ FAISS ์ธ๋ฑ์Šค๋ฅผ ๊ณ ์† IVF ๊ธฐ๋ฐ˜์œผ๋กœ ์žฌ๊ตฌ์ถ• ์ค‘...")
    
    # ์ตœ์‹  ์ƒํ’ˆ ๋ฐ์ดํ„ฐ ๋กœ๋“œ
    active_sale_items = await load_huggingface_jsonl("initial_saleitem_dataset")
    if active_sale_items.empty:
        logger.error("โŒ ์ƒํ’ˆ ๋ฐ์ดํ„ฐ๋ฅผ ๋กœ๋“œํ•  ์ˆ˜ ์—†์Šต๋‹ˆ๋‹ค.")
        raise RuntimeError("์ƒํ’ˆ ๋ฐ์ดํ„ฐ ๋กœ๋“œ ์‹คํŒจ")

    # ์ƒํ’ˆ๋ช… ๋ชฉ๋ก ์ถ”์ถœ
    item_names = active_sale_items["ITEMNAME"].tolist()
    indexed_items = item_names
    
    # ๊ฐ„์†Œํ™”๋œ ๋กœ๊น…
    total_items = len(item_names)
    logger.info(f"๐Ÿ”น ์ด {total_items}๊ฐœ ์ƒํ’ˆ ๊ณ ์† ๋ฒกํ„ฐํ™” ์‹œ์ž‘...")

    # ๋ฒกํ„ฐํ™” ์ตœ์ ํ™” - ๋ฐฐ์น˜ ์‚ฌ์ด์ฆˆ ์ฆ๊ฐ€
    item_vectors = await encode_texts_parallel(item_names, batch_size=1024)
    
    # ๋ฒกํ„ฐ ์ •๊ทœํ™” (์ฝ”์‚ฌ์ธ ์œ ์‚ฌ๋„๋ฅผ ์œ„ํ•ด)
    norms = np.linalg.norm(item_vectors, axis=1, keepdims=True)
    norms[norms == 0] = 1.0  # 0์œผ๋กœ ๋‚˜๋ˆ” ๋ฐฉ์ง€
    normalized_vectors = item_vectors / norms
    
    # IVF ๊ธฐ๋ฐ˜ ์ธ๋ฑ์Šค ๊ตฌ์ถ• (์†๋„ ๋Œ€ํญ ๊ฐœ์„ )
    loop = asyncio.get_event_loop()
    
    def _build_ivf_index():
        dimension = item_vectors.shape[1]
        # IVF ํด๋Ÿฌ์Šคํ„ฐ ์ˆ˜ - ๋ฐ์ดํ„ฐ ํฌ๊ธฐ์— ๋”ฐ๋ผ ์กฐ์ • (โˆšn ๊ทœ์น™ ์‚ฌ์šฉ)
        nlist = int(np.sqrt(total_items) * 4)  # ํด๋Ÿฌ์Šคํ„ฐ ์ˆ˜ ์ฆ๊ฐ€
        nlist = max(32, min(nlist, 1024))  # ์ตœ์†Œ 32, ์ตœ๋Œ€ 1024๊ฐœ ์ œํ•œ
        
        # ์–‘์žํ™” ํŒŒ๋ผ๋ฏธํ„ฐ - ์ฐจ์› ์ˆ˜์— ๋งž๊ฒŒ ์กฐ์ •
        M = min(64, dimension // 2)  # ์„œ๋ธŒ๋ฒกํ„ฐ ์ˆ˜
        nbits = 8  # ๋น„ํŠธ ์ˆ˜
        
        # ๊ณ ์† IVF ์ธ๋ฑ์Šค ์ƒ์„ฑ
        if total_items > 10000:  # ๋ฒกํ„ฐ๊ฐ€ ๋งŽ์œผ๋ฉด ์••์ถ• ๊ธฐ๋ฒ• ์‚ฌ์šฉ
            # IVF + PQ (Product Quantization) ์กฐํ•ฉ - ๋ฉ”๋ชจ๋ฆฌ ํšจ์œจ์ 
            quantizer = faiss.IndexFlatIP(dimension)
            index = faiss.IndexIVFPQ(quantizer, dimension, nlist, M, nbits)
        else:
            # ์ผ๋ฐ˜ IVF - ์†๋„ ํ–ฅ์ƒ
            quantizer = faiss.IndexFlatIP(dimension)
            index = faiss.IndexIVFFlat(quantizer, dimension, nlist)
        
        # ํ•™์Šต ๋ฐ ์ถ”๊ฐ€
        index.train(normalized_vectors)
        index.add(normalized_vectors)
        
        # ๊ฒ€์ƒ‰ ํ’ˆ์งˆ ํ–ฅ์ƒ์„ ์œ„ํ•œ ์„ค์ •
        # nprobe = ๋ช‡ ๊ฐœ์˜ ํด๋Ÿฌ์Šคํ„ฐ๋ฅผ ๊ฒ€์ƒ‰ํ• ์ง€ (๋†’์„์ˆ˜๋ก ์ •ํ™•๋„ โ†‘, ์†๋„ โ†“)
        index.nprobe = min(32, nlist // 4)  # ํด๋Ÿฌ์Šคํ„ฐ์˜ 25% ๊ฒ€์ƒ‰
        
        logger.info(f"โœ… IVF ์ธ๋ฑ์Šค ๊ตฌ์ถ• ์™„๋ฃŒ: clusters={nlist}, nprobe={index.nprobe}")
        return index
    
    # ์ธ๋ฑ์Šค ๊ตฌ์ถ• ์‹คํ–‰
    faiss_index = await loop.run_in_executor(thread_pool, _build_ivf_index)

    logger.info(f"โœ… ๊ณ ์† FAISS ์ธ๋ฑ์Šค ๊ตฌ์ถ• ์™„๋ฃŒ! ์ด {len(indexed_items)}๊ฐœ ํ•ญ๋ชฉ")
    
    # ๊ตฌ์ถ• ํ›„ Hub์— ์ €์žฅ
    await save_faiss_index()
    return True


# โœ… FAISS ์ธ๋ฑ์Šค ์ƒํƒœ ํ™•์ธ ๋ฐ ํ•„์š”์‹œ์—๋งŒ ๊ตฌ์ถ•
async def check_faiss_index():
    """FAISS ์ธ๋ฑ์Šค๊ฐ€ ์กด์žฌํ•˜๋Š”์ง€ ํ™•์ธํ•˜๊ณ  ์—†์œผ๋ฉด ๊ตฌ์ถ• (๋น„๋™๊ธฐ ์ง€์›)"""
    global faiss_index
    
    if faiss_index is None:
        # Hub์—์„œ ๋กœ๋“œ ์‹œ๋„
        if not await load_faiss_index():
            # ๋กœ๋“œ ์‹คํŒจ ์‹œ ์ƒˆ๋กœ ๊ตฌ์ถ•
            logger.warning("โš ๏ธ ์ €์žฅ๋œ ์ธ๋ฑ์Šค๊ฐ€ ์—†์–ด ์ƒˆ๋กœ ๊ตฌ์ถ•ํ•ฉ๋‹ˆ๋‹ค.")
            await rebuild_faiss_index()
    
    # ๋ชจ๋“  ๊ณผ์ • ํ›„์—๋„ ์ธ๋ฑ์Šค๊ฐ€ None์ด๋ฉด ์˜ค๋ฅ˜
    if faiss_index is None:
        raise RuntimeError("FAISS ์ธ๋ฑ์Šค ์ดˆ๊ธฐํ™”์— ์‹คํŒจํ–ˆ์Šต๋‹ˆ๋‹ค.")

# โœ… ์ตœ์ ํ™”๋œ ํ‚ค์›Œ๋“œ ์ถ”์ถœ ํ•จ์ˆ˜
async def extract_keywords(query: str, top_n: int = 2):  # top_n ๊ฐ์†Œ
    """KeyBERT ์ตœ์ ํ™” ํ‚ค์›Œ๋“œ ์ถ”์ถœ (์„ฑ๋Šฅ ์ค‘์‹ฌ)"""
    # ๋งค์šฐ ์งง์€ ์ฟผ๋ฆฌ๋Š” ๊ทธ๋Œ€๋กœ ๋ฐ˜ํ™˜ (์ฒ˜๋ฆฌ ๋น„์šฉ ์ ˆ๊ฐ)
    if len(query) <= 3:
        return [query]
    
    loop = asyncio.get_event_loop()
    
    def _optimized_extract():
        # ์„ฑ๋Šฅ ์ค‘์‹ฌ ์„ค์ •
        return kw_model.extract_keywords(
            query, 
            keyphrase_ngram_range=(1, 1),  # ๋‹จ์ผ ๋‹จ์–ด๋งŒ ์ถ”์ถœ
            stop_words=["์ด", "๊ทธ", "์ €", "์„", "๋ฅผ", "์—", "์—์„œ", "์€", "๋Š”"],  # ํ•œ๊ตญ์–ด ๋ถˆ์šฉ์–ด
            use_mmr=True,
            diversity=0.5,
            top_n=top_n
        )
    
    try:
        keywords = await loop.run_in_executor(thread_pool, _optimized_extract)
        # ๊ฐ€์ค‘์น˜๊ฐ€ ๋„ˆ๋ฌด ๋‚ฎ์€ ํ‚ค์›Œ๋“œ ์ œ์™ธ
        filtered = [(k, s) for k, s in keywords if s > 0.2]
        return [k[0] for k in filtered]
    except Exception as e:
        logger.error(f"โŒ ํ‚ค์›Œ๋“œ ์ถ”์ถœ ์˜ค๋ฅ˜: {str(e)}")
        # ๋‹จ์–ด ๋ถ„๋ฆฌ๋กœ ํด๋ฐฑ
        return query.split()[:2]


# โœ… ์ตœ์ ํ™”๋œ ํ‚ค์›Œ๋“œ ํ™•์žฅ ํ•จ์ˆ˜
async def expand_keywords_with_word2vec(keywords: list, max_new=2):  # max_new ๊ฐ์†Œ
    """Word2Vec ํ‚ค์›Œ๋“œ ํ™•์žฅ ์ตœ์ ํ™”"""
    global word2vec_model
    
    if word2vec_model is None or not keywords:
        return keywords
    
    # ๊ฒฐ๊ณผ ์ €์žฅ์„ ์œ„ํ•œ ์ง‘ํ•ฉ
    expanded = set(keywords)
    
    loop = asyncio.get_event_loop()
    
    def _expand_keywords():
        for keyword in keywords:
            # ๋‹จ์ผ ๋‹จ์–ด์ธ ๊ฒฝ์šฐ
            if keyword in word2vec_model:
                # ์œ ์‚ฌ๋„๊ฐ€ ๋†’์€ ๋‹จ์–ด๋งŒ ์„ ํƒ (์ž„๊ณ„๊ฐ’ ์ ์šฉ)
                similar_words = word2vec_model.most_similar(keyword, topn=max_new)
                for word, score in similar_words:
                    if score > 0.7:  # ๋†’์€ ์œ ์‚ฌ๋„ ์ž„๊ณ„๊ฐ’ ์ ์šฉ
                        expanded.add(word)
            # ๋ณตํ•ฉ์–ด ์ฒ˜๋ฆฌ (์ฒซ ๋‹จ์–ด๋งŒ)
            elif len(keyword.split()) > 1:
                word = keyword.split()[0]
                if word in word2vec_model and len(word) > 1:
                    similar = word2vec_model.most_similar(word, topn=1)
                    if similar and similar[0][1] > 0.8:  # ๋†’์€ ์ž„๊ณ„๊ฐ’
                        expanded.add(similar[0][0])
        
        # ๊ฒฐ๊ณผ ๋ณ€ํ™˜
        result = list(expanded)
        # ํ‚ค์›Œ๋“œ๊ฐ€ ๋„ˆ๋ฌด ๋งŽ์œผ๋ฉด ์ œํ•œ
        if len(result) > 5:
            return keywords + result[len(keywords):5]
        return result
    
    try:
        # ํ™•์žฅ ์‹คํ–‰
        expanded_keywords = await loop.run_in_executor(thread_pool, _expand_keywords)
        return expanded_keywords
    except Exception as e:
        logger.error(f"โŒ Word2Vec ํ™•์žฅ ์˜ค๋ฅ˜: {str(e)}")
        return keywords  # ์˜ค๋ฅ˜ ์‹œ ์›๋ณธ ํ‚ค์›Œ๋“œ ๋ฐ˜ํ™˜


# โœ… ์ตœ์ ํ™”๋œ search_faiss_with_keywords ํ•จ์ˆ˜
async def search_faiss_with_keywords(query: str, top_k: int = 5, keywords=None, expanded_keywords=None):
    """๊ณ ์† ํ‚ค์›Œ๋“œ ๊ธฐ๋ฐ˜ FAISS ๊ฒ€์ƒ‰ ์ˆ˜ํ–‰"""
    global faiss_index, indexed_items
    
    # FAISS ์ธ๋ฑ์Šค ํ™•์ธ - ํ•œ ๋ฒˆ๋งŒ ์‹คํ–‰
    if faiss_index is None:
        await check_faiss_index()
    
    # ํƒ€์ด๋จธ ์‹œ์ž‘
    start_time = time.time()
    
    # ๋ณ‘๋ ฌ ์‹คํ–‰์„ ์œ„ํ•œ ์ค€๋น„
    loop = asyncio.get_event_loop()
    
    # 1. ํ‚ค์›Œ๋“œ ์ถ”์ถœ ๋ฐ ํ™•์žฅ ์ตœ์ ํ™”
    if keywords is None:
        keywords = await extract_keywords(query)
    
    if expanded_keywords is None:
        expanded_keywords = await expand_keywords_with_word2vec(keywords)
    
    # 2. ๋ฒกํ„ฐ ์ธ์ฝ”๋”ฉ ์ตœ์ ํ™” - ์ฟผ๋ฆฌ์™€ ํ‚ค์›Œ๋“œ ํ•œ ๋ฒˆ์— ์ฒ˜๋ฆฌ
    search_texts = [query] + expanded_keywords
    
    # ๋ฒกํ„ฐ ์ธ์ฝ”๋”ฉ - ์ตœ์ ํ™”๋œ ํ•จ์ˆ˜ ์‚ฌ์šฉ
    all_vectors = await encode_texts_parallel(search_texts)
    
    # ๋ฒกํ„ฐ ์ •๊ทœํ™” - ์ตœ์ ํ™”๋œ ๋ฐฉ์‹
    def normalize_batch(vectors):
        if vectors.size == 0:
            return vectors
        norms = np.linalg.norm(vectors, axis=1, keepdims=True)
        norms[norms == 0] = 1.0  # 0์œผ๋กœ ๋‚˜๋ˆ” ๋ฐฉ์ง€
        return vectors / norms
    
    # ๋ฒกํ„ฐ ์ •๊ทœํ™” ์‹คํ–‰
    all_vectors = await loop.run_in_executor(thread_pool, lambda: normalize_batch(all_vectors))
    
    # ์ฟผ๋ฆฌ ๋ฒกํ„ฐ์™€ ํ‚ค์›Œ๋“œ ๋ฒกํ„ฐ ๋ถ„๋ฆฌ
    if len(all_vectors) > 0:
        query_vector = all_vectors[0:1]
        keyword_vectors = all_vectors[1:] if len(all_vectors) > 1 else np.array([])
    else:
        return []  # ๋ฒกํ„ฐํ™” ์‹คํŒจ ์‹œ ๋นˆ ๊ฒฐ๊ณผ ๋ฐ˜ํ™˜
    
    # 3. FAISS ๊ฒ€์ƒ‰ ์ตœ์ ํ™” - ์ผ๊ด„ ๋ฐฐ์น˜ ์ฒ˜๋ฆฌ
    def _optimized_batch_search():
        all_results = {}
        
        # ์ฟผ๋ฆฌ ๋ฒกํ„ฐ ๊ฒ€์ƒ‰ (๊ฐ€์ค‘์น˜ 3๋ฐฐ๋กœ ์ฆ๊ฐ€)
        if query_vector.shape[0] > 0:
            distances, indices = faiss_index.search(query_vector, top_k * 2)
            # ์ฟผ๋ฆฌ ๊ฒฐ๊ณผ ๊ฐ€์ค‘์น˜ ์ ์šฉ (์ค‘์š”๋„ ์ƒํ–ฅ)
            for idx, dist in zip(indices[0], distances[0]):
                if idx < len(indexed_items):
                    all_results[idx] = dist * 3.0  # ๊ฐ€์ค‘์น˜ 3.0
        
        # ํ‚ค์›Œ๋“œ ๋ฒกํ„ฐ ๋ฐฐ์น˜ ๊ฒ€์ƒ‰
        if keyword_vectors.shape[0] > 0:
            # ๋ฐฐ์น˜ ๊ฒ€์ƒ‰ ํ•œ ๋ฒˆ์— ์ฒ˜๋ฆฌ
            k_distances, k_indices = faiss_index.search(keyword_vectors, top_k)
            
            # ํ‚ค์›Œ๋“œ๋ณ„ ๊ฐ€์ค‘์น˜ ์ ์šฉ ๋ฐ ๊ฒฐ๊ณผ ๋ณ‘ํ•ฉ
            for i in range(keyword_vectors.shape[0]):
                for j, (idx, dist) in enumerate(zip(k_indices[i], k_distances[i])):
                    if idx < len(indexed_items):
                        # ์ˆœ์œ„์— ๋”ฐ๋ผ ๊ฐ€์ค‘์น˜ ์ฐจ๋“ฑ ์ ์šฉ (์ƒ์œ„ ๊ฒฐ๊ณผ ์šฐ๋Œ€)
                        rank_weight = 1.0 / (1 + j * 0.2)  # ์ˆœ์œ„๋ณ„ ๊ฐ€์ค‘์น˜ ๊ฐ์†Œ
                        weight = 0.6 * rank_weight  # ๊ธฐ๋ณธ ๊ฐ€์ค‘์น˜ 0.6
                        
                        # ๊ธฐ์กด ์ ์ˆ˜์— ์ถ”๊ฐ€
                        all_results[idx] = all_results.get(idx, 0) + dist * weight
        
        return all_results
    
    # ์ตœ์ ํ™”๋œ ๋ฐฐ์น˜ ๊ฒ€์ƒ‰ ์‹คํ–‰
    result_scores = await loop.run_in_executor(thread_pool, _optimized_batch_search)
    
    # 4. ๊ฒฐ๊ณผ ์ฒ˜๋ฆฌ ๋ฐ ์ •๋ ฌ ์ตœ์ ํ™”
    def _process_results():
        # ์ž„๊ณ„๊ฐ’ ํ•„ํ„ฐ๋ง ๋ฐ ์ •๋ ฌ
        filtered_items = [(idx, score) for idx, score in result_scores.items() 
                        if score >= 0.3]  # ์ตœ์†Œ ์ ์ˆ˜ ํ•„ํ„ฐ๋ง
        
        # ์ ์ˆ˜ ๊ธฐ์ค€ ๋‚ด๋ฆผ์ฐจ์ˆœ ์ •๋ ฌ
        sorted_items = sorted(filtered_items, key=lambda x: x[1], reverse=True)
        
        # ์ตœ์ข… ๊ฒฐ๊ณผ ๋ณ€ํ™˜
        recommendations = []
        for idx, score in sorted_items[:top_k]:  # top_k๊ฐœ๋งŒ ์ฒ˜๋ฆฌ
            item_name = indexed_items[idx]
            try:
                # ๋ฉ”๋ชจ๋ฆฌ ๋‚ด ์กฐํšŒ ์ตœ์ ํ™”
                mask = active_sale_items["ITEMNAME"] == item_name
                if mask.any():
                    item_seq = active_sale_items.loc[mask, "ITEMSEQ"].values[0]
                    recommendations.append({
                        "ITEMSEQ": item_seq,
                        "ITEMNAME": item_name,
                        "score": float(score)
                    })
            except (IndexError, KeyError):
                continue
                
        return recommendations
    
    # ๊ฒฐ๊ณผ ์ฒ˜๋ฆฌ ์‹คํ–‰
    recommendations = await loop.run_in_executor(thread_pool, _process_results)
    
    # 5. ์ง์ ‘ ๋งค์นญ ์ถ”๊ฐ€ ์ตœ์ ํ™” (ํ•„์š”ํ•œ ๊ฒฝ์šฐ์—๋งŒ)
    if len(recommendations) < top_k:
        direct_matches = await find_direct_matches(query, 
                                                  top_k - len(recommendations),
                                                  [r["ITEMNAME"] for r in recommendations])
        if direct_matches:
            recommendations.extend(direct_matches)
    
    # ์ฒ˜๋ฆฌ ์‹œ๊ฐ„์ด 1์ดˆ ์ด์ƒ์ธ ๊ฒฝ์šฐ์—๋งŒ ๋กœ๊น…
    elapsed = time.time() - start_time
    if elapsed > 1.0:
        logger.info(f"๐Ÿ” ๊ฒ€์ƒ‰ ์™„๋ฃŒ | ์†Œ์š”์‹œ๊ฐ„: {elapsed:.2f}์ดˆ | ๊ฒฐ๊ณผ: {len(recommendations)}๊ฐœ")
    
    return recommendations[:top_k]

# โœ… ์ง์ ‘ ๋งค์นญ ๋ถ„๋ฆฌ (์„ฑ๋Šฅ ์ตœ์ ํ™”)
async def find_direct_matches(query, limit=5, existing_names=None):
    """์ง์ ‘ ํ…์ŠคํŠธ ๋งค์นญ ๊ฒ€์ƒ‰ (๋ถ„๋ฆฌํ•˜์—ฌ ์ตœ์ ํ™”)"""
    loop = asyncio.get_event_loop()
    
    def _find_matches():
        matches = []
        query_lower = query.lower()
        existing = set(existing_names or [])
        
        # ๋ฐ์ดํ„ฐ ์ธ๋ฑ์‹ฑ ์ตœ์ ํ™”
        item_dict = {}
        for idx, item_name in enumerate(indexed_items):
            if len(matches) >= limit:
                break
                
            if item_name in existing:
                continue
                
            if query_lower in item_name.lower():
                item_dict[item_name] = idx
        
        # ํ•œ ๋ฒˆ์— ๋ฐ์ดํ„ฐํ”„๋ ˆ์ž„ ์กฐํšŒ
        if item_dict:
            mask = active_sale_items["ITEMNAME"].isin(item_dict.keys())
            filtered_items = active_sale_items[mask]
            
            for _, row in filtered_items.iterrows():
                if len(matches) >= limit:
                    break
                    
                matches.append({
                    "ITEMSEQ": row["ITEMSEQ"], 
                    "ITEMNAME": row["ITEMNAME"], 
                    "score": 1.0
                })
        
        return matches
    
    # ์Šค๋ ˆ๋“œ ํ’€์—์„œ ์‹คํ–‰
    return await loop.run_in_executor(thread_pool, _find_matches)

# โœ… API ์š”์ฒญ ๋ชจ๋ธ
class RecommendRequest(BaseModel):
    search_query: str
    top_k: int = 5
    use_expansion: bool = True  # ํ‚ค์›Œ๋“œ ํ™•์žฅ ์‚ฌ์šฉ ์—ฌ๋ถ€

# โœ… ์ถ”์ฒœ API ์—”๋“œํฌ์ธํŠธ
# โœ… ์ตœ์ ํ™”๋œ recommend API ์—”๋“œํฌ์ธํŠธ
@app.post("/api/recommend")
async def recommend(request: RecommendRequest, background_tasks: BackgroundTasks):
    """๊ณ ์† ์ถ”์ฒœ API (I/O ๋ณ‘๋ ฌํ™” + ๋ถˆํ•„์š” ์ž‘์—… ์ œ๊ฑฐ)"""
    try:
        # ๋ฒค์น˜๋งˆํฌ์šฉ ํƒ€์ด๋จธ ์‹œ์ž‘
        start_time = time.time()
        
        # ํŒŒ๋ผ๋ฏธํ„ฐ ์ตœ์ ํ™” ๋ฐ ๊ฒ€์ฆ
        search_query = request.search_query.strip()
        if not search_query:
            raise HTTPException(status_code=400, detail="๊ฒ€์ƒ‰์–ด๋ฅผ ์ž…๋ ฅํ•ด์ฃผ์„ธ์š”")
            
        top_k = min(max(1, request.top_k), 20)  # 1~20 ๋ฒ”์œ„๋กœ ์ œํ•œ
        
        # ๋ณ‘๋ ฌ ํ”„๋กœ์„ธ์‹ฑ์„ ์œ„ํ•œ ๋™์‹œ ์‹คํ–‰
        keywords, expanded_keywords = await asyncio.gather(
            extract_keywords(search_query),
            expand_keywords_with_word2vec(
                [search_query.split()[0]] if search_query.split() else [search_query],
                max_new=2
            ) if request.use_expansion else None
        )
        
        # ๊ฒ€์ƒ‰ ์‹คํ–‰ - ๋ณ‘๋ ฌ ์ฒ˜๋ฆฌ๋œ ํ‚ค์›Œ๋“œ ํ™œ์šฉ
        recommendations = await search_faiss_with_keywords(
            search_query, 
            top_k,
            keywords,
            expanded_keywords
        )
        
        # ๊ฒฐ๊ณผ ๋ฐ˜ํ™˜
        result = {
            "query": search_query,
            "recommendations": recommendations,
            "keywords": keywords if len(keywords) > 0 else None,
            "expanded_keywords": expanded_keywords if expanded_keywords and len(expanded_keywords) > 0 else None
        }
        
        # ์‘๋‹ต ์‹œ๊ฐ„ ์ธก์ • (1์ดˆ ์ด์ƒ๋งŒ ๋กœ๊น…)
        elapsed = time.time() - start_time
        if elapsed > 1.0:
            logger.info(f"โฑ๏ธ API ์‘๋‹ต ์‹œ๊ฐ„: {elapsed:.2f}์ดˆ | ์ฟผ๋ฆฌ: '{search_query}'")
        
        return result
        
    except Exception as e:
        logger.error(f"โŒ ์ถ”์ฒœ ์ฒ˜๋ฆฌ ์˜ค๋ฅ˜: {str(e)}")
        raise HTTPException(status_code=500, detail=f"์ถ”์ฒœ ์ฒ˜๋ฆฌ ์ค‘ ์˜ค๋ฅ˜๊ฐ€ ๋ฐœ์ƒํ–ˆ์Šต๋‹ˆ๋‹ค")

# ์ธ๋ฑ์Šค ์ƒํƒœ ํ™•์ธ ํ•จ์ˆ˜ (๋ฐฑ๊ทธ๋ผ์šด๋“œ ํƒœ์Šคํฌ์šฉ)
async def check_index_health():
    """์ธ๋ฑ์Šค ์ƒํƒœ๋ฅผ ์ฃผ๊ธฐ์ ์œผ๋กœ ํ™•์ธํ•˜๋Š” ๋ฐฑ๊ทธ๋ผ์šด๋“œ ํƒœ์Šคํฌ"""
    try:
        # ์ธ๋ฑ์Šค ์‚ฌ์šฉ ์ƒํƒœ ํ™•์ธ
        if faiss_index is None:
            logger.warning("โš ๏ธ ๋ฐฑ๊ทธ๋ผ์šด๋“œ ์ฒดํฌ: FAISS ์ธ๋ฑ์Šค๊ฐ€ ๋กœ๋“œ๋˜์ง€ ์•Š์•˜์Šต๋‹ˆ๋‹ค.")
            await check_faiss_index()
        
        # ์ถ”๊ฐ€์ ์ธ ์ƒํƒœ ํ™•์ธ ๋กœ์ง์„ ์—ฌ๊ธฐ์— ๊ตฌํ˜„ํ•  ์ˆ˜ ์žˆ์Œ
        logger.debug("โœ… ์ธ๋ฑ์Šค ์ƒํƒœ ํ™•์ธ ์™„๋ฃŒ")
    except Exception as e:
        logger.error(f"โŒ ๋ฐฑ๊ทธ๋ผ์šด๋“œ ์ธ๋ฑ์Šค ์ฒดํฌ ์ค‘ ์˜ค๋ฅ˜: {str(e)}")

# โœ… ์œ ์‚ฌ ๋‹จ์–ด ๊ฒ€์ƒ‰ API
@app.post("/api/similar_words")
async def similar_words(word: str, top_k: int = 10):
    """Word2Vec ๋ชจ๋ธ์„ ์‚ฌ์šฉํ•œ ์œ ์‚ฌ ๋‹จ์–ด ๊ฒ€์ƒ‰ API (๋น„๋™๊ธฐ ์ง€์›)"""
    try:
        if word2vec_model is None:
            return {"error": "Word2Vec ๋ชจ๋ธ์ด ๋กœ๋“œ๋˜์ง€ ์•Š์•˜์Šต๋‹ˆ๋‹ค."}
        
        loop = asyncio.get_event_loop()
        
        def _get_similar():
            if word not in word2vec_model:
                return []
            
            similar = word2vec_model.most_similar(word, topn=top_k)
            return [{"word": w, "similarity": float(s)} for w, s in similar]
        
        result = await loop.run_in_executor(thread_pool, _get_similar)
        
        if not result:
            return {"word": word, "similar_words": [], "message": "๋‹จ์–ด๊ฐ€ ๋ชจ๋ธ์— ์—†์Šต๋‹ˆ๋‹ค."}
        
        return {"word": word, "similar_words": result}
    except Exception as e:
        logger.error(f"โŒ ์œ ์‚ฌ ๋‹จ์–ด ๊ฒ€์ƒ‰ ์ค‘ ์˜ค๋ฅ˜: {str(e)}")
        raise HTTPException(status_code=500, detail=f"์œ ์‚ฌ ๋‹จ์–ด ๊ฒ€์ƒ‰ ์˜ค๋ฅ˜: {str(e)}")

# โœ… FAISS ์ธ๋ฑ์Šค ๊ฐฑ์‹  API (๋ช…์‹œ์ ์œผ๋กœ ์š”์ฒญํ•  ๋•Œ๋งŒ ์‹คํ–‰)
@app.post("/api/update_index")
async def update_index(background_tasks: BackgroundTasks):
    """FAISS ์ธ๋ฑ์Šค๋ฅผ ์ƒˆ๋กญ๊ฒŒ ๊ตฌ์ถ• (๋ช…์‹œ์  ์š”์ฒญ ์‹œ์—๋งŒ, ๋น„๋™๊ธฐ ์ฒ˜๋ฆฌ)"""
    try:
        # ์ธ๋ฑ์Šค ์žฌ๊ตฌ์ถ•์„ ๋ฐฑ๊ทธ๋ผ์šด๋“œ ํƒœ์Šคํฌ๋กœ ์‹คํ–‰
        background_tasks.add_task(rebuild_and_log_index)
        return {"message": "โœ… FAISS ์ธ๋ฑ์Šค ์—…๋ฐ์ดํŠธ๊ฐ€ ๋ฐฑ๊ทธ๋ผ์šด๋“œ์—์„œ ์‹œ์ž‘๋˜์—ˆ์Šต๋‹ˆ๋‹ค."}
    except Exception as e:
        logger.exception("โŒ [API] ์ธ๋ฑ์Šค ์—…๋ฐ์ดํŠธ ์ฒ˜๋ฆฌ ์ค‘ ์˜ˆ์™ธ ๋ฐœ์ƒ")
        raise HTTPException(status_code=500, detail=f"์ธ๋ฑ์Šค ์—…๋ฐ์ดํŠธ ์‹คํŒจ: {str(e)}")

# ๋ฐฑ๊ทธ๋ผ์šด๋“œ ์ž‘์—…์šฉ ์ธ๋ฑ์Šค ์žฌ๊ตฌ์ถ• ํ•จ์ˆ˜
async def rebuild_and_log_index():
    """๋ฐฑ๊ทธ๋ผ์šด๋“œ์—์„œ ์ธ๋ฑ์Šค๋ฅผ ์žฌ๊ตฌ์ถ•ํ•˜๊ณ  ๊ฒฐ๊ณผ๋ฅผ ๋กœ๊น…"""
    try:
        logger.info("๐Ÿ”„ ๋ฐฑ๊ทธ๋ผ์šด๋“œ์—์„œ ์ธ๋ฑ์Šค ์žฌ๊ตฌ์ถ• ์‹œ์ž‘")
        start_time = time.time()
        await rebuild_faiss_index()
        elapsed = time.time() - start_time
        logger.info(f"โœ… ๋ฐฑ๊ทธ๋ผ์šด๋“œ ์ธ๋ฑ์Šค ์žฌ๊ตฌ์ถ• ์™„๋ฃŒ! ์†Œ์š” ์‹œ๊ฐ„: {elapsed:.2f}์ดˆ")
    except Exception as e:
        logger.error(f"โŒ ๋ฐฑ๊ทธ๋ผ์šด๋“œ ์ธ๋ฑ์Šค ์žฌ๊ตฌ์ถ• ์ค‘ ์˜ค๋ฅ˜: {str(e)}")

# โœ… ์ธ๋ฑ์Šค ๋””๋ฒ„๊น… API
@app.get("/api/debug_index")
async def debug_index(query: str, top_k: int = 20):
    """์ธ๋ฑ์Šค ๋””๋ฒ„๊น…์„ ์œ„ํ•œ API (๋น„๋™๊ธฐ ์ง€์›)"""
    try:
        await check_faiss_index()
        
        loop = asyncio.get_event_loop()
        
        # ์›๋ณธ ๋ฒกํ„ฐ ์ƒ์„ฑ (๋น„๋™๊ธฐ)
        def _get_vector():
            vector = embedding_model.encode(query, convert_to_numpy=True).astype("float32")
            norm = np.linalg.norm(vector)
            normalized_vector = vector / norm
            return normalized_vector, norm
        
        normalized_vector, norm = await loop.run_in_executor(thread_pool, _get_vector)
        
        # ์›๋ณธ ์ฟผ๋ฆฌ๋กœ ๊ฒ€์ƒ‰ (๋น„๋™๊ธฐ)
        def _search():
            return faiss_index.search(np.array([normalized_vector]), top_k)
        
        distances, indices = await loop.run_in_executor(thread_pool, _search)
        
        # ๊ฒฐ๊ณผ ๋งคํ•‘
        results = []
        for i, (idx, dist) in enumerate(zip(indices[0], distances[0])):
            if idx < len(indexed_items):
                item_name = indexed_items[idx]
                results.append({
                    "rank": i + 1,
                    "index": int(idx),
                    "item_name": item_name,
                    "distance/score": float(dist)
                })
        
        # ๋ฐ์ดํ„ฐ์…‹์— ํ•ด๋‹น ๋‹จ์–ด๊ฐ€ ์žˆ๋Š”์ง€ ํ™•์ธ (๋น„๋™๊ธฐ)
        def _find_matches():
            contains = [item for item in indexed_items if query.lower() in item.lower()][:5]
            exact = [item for item in indexed_items if query.lower() == item.lower()]
            return contains, exact
        
        contains_query, exact_matches = await loop.run_in_executor(thread_pool, _find_matches)
        
        return {
            "query": query,
            "vector_norm": float(norm),
            "contains_query": contains_query,
            "exact_matches": exact_matches,
            "results": results
        }
    except Exception as e:
        logger.error(f"โŒ ์ธ๋ฑ์Šค ๋””๋ฒ„๊น… ์ค‘ ์˜ค๋ฅ˜: {str(e)}")
        raise HTTPException(status_code=500, detail=f"์ธ๋ฑ์Šค ๋””๋ฒ„๊น… ์˜ค๋ฅ˜: {str(e)}")

# โœ… ๋ฌธ์ž์—ด ํฌํ•จ ๊ฒ€์ƒ‰ API
@app.get("/api/text_search")
async def text_search(query: str, top_k: int = 10):
    """๋‹จ์ˆœ ํ…์ŠคํŠธ ํฌํ•จ ๊ฒ€์ƒ‰ API (๋น„๋™๊ธฐ ์ง€์›)"""
    try:
        loop = asyncio.get_event_loop()
        
        # ๋น„๋™๊ธฐ ๊ฒ€์ƒ‰ ํ•จ์ˆ˜
        def _text_search():
            # ๋‹จ์ˆœ ํ…์ŠคํŠธ ํฌํ•จ ๊ฒ€์ƒ‰
            matched_items = []
            for idx, item_name in enumerate(indexed_items):
                if query.lower() in item_name.lower():
                    try:
                        item_seq = active_sale_items.loc[active_sale_items["ITEMNAME"] == item_name, "ITEMSEQ"].values[0]
                        matched_items.append({"ITEMSEQ": item_seq, "ITEMNAME": item_name, "match_type": "contains"})
                    except (IndexError, KeyError):
                        continue

            # ์ •ํ™•ํžˆ ์ผ์น˜ํ•˜๋Š” ํ•ญ๋ชฉ์„ ์•ž์œผ๋กœ
            exact_matches = []
            partial_matches = []
            
            for item in matched_items:
                if query.lower() == item["ITEMNAME"].lower():
                    item["match_type"] = "exact"
                    exact_matches.append(item)
                else:
                    partial_matches.append(item)
            
            # ๊ฒฐํ•ฉ ๋ฐ ์ œํ•œ
            return exact_matches + partial_matches
        
        # ๋น„๋™๊ธฐ์ ์œผ๋กœ ๊ฒ€์ƒ‰ ์‹คํ–‰
        results = await loop.run_in_executor(thread_pool, _text_search)
        
        logger.info(f"๐Ÿ” ํ…์ŠคํŠธ ๊ฒ€์ƒ‰ ๊ฒฐ๊ณผ: {len(results)}๊ฐœ ์ฐพ์Œ, ์ฟผ๋ฆฌ: '{query}'")
        
        return {
            "query": query,
            "total_matches": len(results),
            "results": results[:top_k]
        }
    except Exception as e:
        logger.error(f"โŒ ํ…์ŠคํŠธ ๊ฒ€์ƒ‰ ์ค‘ ์˜ค๋ฅ˜: {str(e)}")
        raise HTTPException(status_code=500, detail=f"ํ…์ŠคํŠธ ๊ฒ€์ƒ‰ ์˜ค๋ฅ˜: {str(e)}")

# โœ… FastAPI ์‹คํ–‰
if __name__ == "__main__":
    # ์„œ๋ฒ„ ์‹œ์ž‘ ์‹œ ์ €์žฅ๋œ ์ธ๋ฑ์Šค ๋กœ๋“œ ์‹œ๋„
    try:
        # ๋น„๋™๊ธฐ ํ•จ์ˆ˜๋ฅผ ๋™๊ธฐ์ ์œผ๋กœ ํ˜ธ์ถœํ•˜๊ธฐ ์œ„ํ•œ ์ž„์‹œ ์ด๋ฒคํŠธ ๋ฃจํ”„ ์‚ฌ์šฉ
        loop = asyncio.new_event_loop()
        if not loop.run_until_complete(load_faiss_index()):
            logger.warning("โš ๏ธ ๊ธฐ์กด ์ธ๋ฑ์Šค ๋กœ๋“œ์— ์‹คํŒจํ–ˆ์Šต๋‹ˆ๋‹ค. ์ฆ‰์‹œ ์ƒˆ ์ธ๋ฑ์Šค๋ฅผ ๊ตฌ์ถ•ํ•ฉ๋‹ˆ๋‹ค.")
            # ์ธ๋ฑ์Šค ์ฆ‰์‹œ ์žฌ๊ตฌ์ถ•
            loop.run_until_complete(rebuild_faiss_index())
            logger.info("โœ… FAISS ์ธ๋ฑ์Šค ์ƒ์„ฑ ์™„๋ฃŒ!")
        else:
            logger.info("โœ… ๊ธฐ์กด ์ธ๋ฑ์Šค๋ฅผ ์„ฑ๊ณต์ ์œผ๋กœ ๋กœ๋“œํ–ˆ์Šต๋‹ˆ๋‹ค.")
        loop.close()
    except Exception as e:
        logger.error(f"โŒ ์ธ๋ฑ์Šค ์ดˆ๊ธฐ ๊ตฌ์ถ• ์‹คํŒจ: {e}")
        logger.warning("โš ๏ธ ์ธ๋ฑ์Šค ์—†์ด ์‹œ์ž‘ํ•ฉ๋‹ˆ๋‹ค. ๊ฒ€์ƒ‰ ๊ธฐ๋Šฅ์ด ์ œํ•œ๋  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.")
    
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=7860)