File size: 81,967 Bytes
bb9c2bf
 
 
 
022d1be
bb9c2bf
 
 
 
 
 
 
 
 
022d1be
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bb9c2bf
 
 
 
 
 
022d1be
 
 
 
bb9c2bf
 
 
 
 
 
 
022d1be
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bb9c2bf
 
 
 
 
 
022d1be
bb9c2bf
 
 
 
 
 
 
 
 
 
 
 
 
022d1be
bb9c2bf
022d1be
bb9c2bf
 
022d1be
bb9c2bf
 
 
 
 
 
022d1be
 
bb9c2bf
 
 
 
 
 
022d1be
bb9c2bf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
022d1be
 
 
 
 
bb9c2bf
 
 
 
 
 
 
 
022d1be
 
bb9c2bf
 
 
022d1be
bb9c2bf
 
 
 
 
 
022d1be
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bb9c2bf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
022d1be
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bb9c2bf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
022d1be
bb9c2bf
 
 
 
 
 
 
 
 
 
 
022d1be
bb9c2bf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
022d1be
bb9c2bf
 
022d1be
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bb9c2bf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "405bc169-e0b7-48e6-84b8-4e4a791cf61a",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO 06-07 22:36:36 [__init__.py:243] Automatically detected platform cuda.\n",
      "INFO 06-07 22:36:40 [__init__.py:31] Available plugins for group vllm.general_plugins:\n",
      "INFO 06-07 22:36:40 [__init__.py:33] - lora_filesystem_resolver -> vllm.plugins.lora_resolvers.filesystem_resolver:register_filesystem_resolver\n",
      "INFO 06-07 22:36:40 [__init__.py:36] All plugins in this group will be loaded. Set `VLLM_PLUGINS` to control which plugins to load.\n",
      "INFO 06-07 22:36:41 [api_server.py:1289] vLLM API server version 0.9.0.1\n",
      "INFO 06-07 22:36:41 [cli_args.py:300] non-default args: {'host': '0.0.0.0', 'task': 'embed', 'trust_remote_code': True, 'enforce_eager': True, 'served_model_name': ['local'], 'tensor_parallel_size': 2, 'gpu_memory_utilization': 0.4}\n",
      "WARNING 06-07 22:36:43 [config.py:3096] Your Quadro RTX 8000 device (with compute capability 7.5) doesn't support torch.bfloat16. Falling back to torch.float16 for compatibility.\n",
      "WARNING 06-07 22:36:43 [config.py:3135] Casting torch.bfloat16 to torch.float16.\n",
      "INFO 06-07 22:36:51 [config.py:473] Found sentence-transformers modules configuration.\n",
      "INFO 06-07 22:36:52 [config.py:493] Found pooling configuration.\n",
      "WARNING 06-07 22:36:52 [arg_utils.py:1583] Compute Capability < 8.0 is not supported by the V1 Engine. Falling back to V0. \n",
      "WARNING 06-07 22:36:52 [arg_utils.py:1431] The model has a long context length (40960). This may causeOOM during the initial memory profiling phase, or result in low performance due to small KV cache size. Consider setting --max-model-len to a smaller value.\n",
      "INFO 06-07 22:36:52 [config.py:1875] Defaulting to use mp for distributed inference\n",
      "WARNING 06-07 22:36:52 [cuda.py:87] To see benefits of async output processing, enable CUDA graph. Since, enforce-eager is enabled, async output processor cannot be used\n",
      "INFO 06-07 22:36:52 [api_server.py:257] Started engine process with PID 16420\n",
      "INFO 06-07 22:36:56 [__init__.py:243] Automatically detected platform cuda.\n",
      "INFO 06-07 22:36:59 [__init__.py:31] Available plugins for group vllm.general_plugins:\n",
      "INFO 06-07 22:36:59 [__init__.py:33] - lora_filesystem_resolver -> vllm.plugins.lora_resolvers.filesystem_resolver:register_filesystem_resolver\n",
      "INFO 06-07 22:36:59 [__init__.py:36] All plugins in this group will be loaded. Set `VLLM_PLUGINS` to control which plugins to load.\n",
      "INFO 06-07 22:36:59 [llm_engine.py:230] Initializing a V0 LLM engine (v0.9.0.1) with config: model='Qwen/Qwen3-Embedding-4B', speculative_config=None, tokenizer='Qwen/Qwen3-Embedding-4B', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, override_neuron_config={}, tokenizer_revision=None, trust_remote_code=True, dtype=torch.float16, max_seq_len=40960, download_dir=None, load_format=LoadFormat.AUTO, tensor_parallel_size=2, pipeline_parallel_size=1, disable_custom_all_reduce=False, quantization=None, enforce_eager=True, kv_cache_dtype=auto,  device_config=cuda, decoding_config=DecodingConfig(backend='auto', disable_fallback=False, disable_any_whitespace=False, disable_additional_properties=False, reasoning_backend=''), observability_config=ObservabilityConfig(show_hidden_metrics_for_version=None, otlp_traces_endpoint=None, collect_detailed_traces=None), seed=0, served_model_name=local, num_scheduler_steps=1, multi_step_stream_outputs=True, enable_prefix_caching=None, chunked_prefill_enabled=False, use_async_output_proc=False, pooler_config=PoolerConfig(pooling_type='LAST', normalize=True, softmax=None, step_tag_id=None, returned_token_ids=None), compilation_config={\"compile_sizes\": [], \"inductor_compile_config\": {\"enable_auto_functionalized_v2\": false}, \"cudagraph_capture_sizes\": [], \"max_capture_size\": 0}, use_cached_outputs=True, \n",
      "WARNING 06-07 22:37:00 [multiproc_worker_utils.py:306] Reducing Torch parallelism from 64 threads to 1 to avoid unnecessary CPU contention. Set OMP_NUM_THREADS in the external environment to tune this value as needed.\n",
      "INFO 06-07 22:37:00 [cuda.py:240] Cannot use FlashAttention-2 backend for Volta and Turing GPUs.\n",
      "INFO 06-07 22:37:00 [cuda.py:289] Using XFormers backend.\n",
      "INFO 06-07 22:37:04 [__init__.py:243] Automatically detected platform cuda.\n",
      "\u001b[1;36m(VllmWorkerProcess pid=16602)\u001b[0;0m INFO 06-07 22:37:07 [multiproc_worker_utils.py:225] Worker ready; awaiting tasks\n",
      "\u001b[1;36m(VllmWorkerProcess pid=16602)\u001b[0;0m INFO 06-07 22:37:07 [__init__.py:31] Available plugins for group vllm.general_plugins:\n",
      "\u001b[1;36m(VllmWorkerProcess pid=16602)\u001b[0;0m INFO 06-07 22:37:07 [__init__.py:33] - lora_filesystem_resolver -> vllm.plugins.lora_resolvers.filesystem_resolver:register_filesystem_resolver\n",
      "\u001b[1;36m(VllmWorkerProcess pid=16602)\u001b[0;0m INFO 06-07 22:37:07 [__init__.py:36] All plugins in this group will be loaded. Set `VLLM_PLUGINS` to control which plugins to load.\n",
      "\u001b[1;36m(VllmWorkerProcess pid=16602)\u001b[0;0m INFO 06-07 22:37:07 [cuda.py:240] Cannot use FlashAttention-2 backend for Volta and Turing GPUs.\n",
      "\u001b[1;36m(VllmWorkerProcess pid=16602)\u001b[0;0m INFO 06-07 22:37:07 [cuda.py:289] Using XFormers backend.\n",
      "INFO 06-07 22:37:08 [utils.py:1077] Found nccl from library libnccl.so.2\n",
      "INFO 06-07 22:37:08 [pynccl.py:69] vLLM is using nccl==2.26.2\n",
      "\u001b[1;36m(VllmWorkerProcess pid=16602)\u001b[0;0m INFO 06-07 22:37:08 [utils.py:1077] Found nccl from library libnccl.so.2\n",
      "\u001b[1;36m(VllmWorkerProcess pid=16602)\u001b[0;0m INFO 06-07 22:37:08 [pynccl.py:69] vLLM is using nccl==2.26.2\n",
      "\u001b[1;36m(VllmWorkerProcess pid=16602)\u001b[0;0m INFO 06-07 22:37:09 [custom_all_reduce_utils.py:245] reading GPU P2P access cache from /home/jovyan/.cache/vllm/gpu_p2p_access_cache_for_0,1.json\n",
      "INFO 06-07 22:37:09 [custom_all_reduce_utils.py:245] reading GPU P2P access cache from /home/jovyan/.cache/vllm/gpu_p2p_access_cache_for_0,1.json\n",
      "WARNING 06-07 22:37:09 [custom_all_reduce.py:146] Custom allreduce is disabled because your platform lacks GPU P2P capability or P2P test failed. To silence this warning, specify disable_custom_all_reduce=True explicitly.\n",
      "\u001b[1;36m(VllmWorkerProcess pid=16602)\u001b[0;0m WARNING 06-07 22:37:09 [custom_all_reduce.py:146] Custom allreduce is disabled because your platform lacks GPU P2P capability or P2P test failed. To silence this warning, specify disable_custom_all_reduce=True explicitly.\n",
      "INFO 06-07 22:37:09 [shm_broadcast.py:250] vLLM message queue communication handle: Handle(local_reader_ranks=[1], buffer_handle=(1, 4194304, 6, 'psm_f9ac8311'), local_subscribe_addr='ipc:///tmp/718ad6af-61c7-4d9f-8044-b415ab240a60', remote_subscribe_addr=None, remote_addr_ipv6=False)\n",
      "\u001b[1;36m(VllmWorkerProcess pid=16602)\u001b[0;0m INFO 06-07 22:37:09 [parallel_state.py:1064] rank 1 in world size 2 is assigned as DP rank 0, PP rank 0, TP rank 1, EP rank 1\n",
      "INFO 06-07 22:37:09 [parallel_state.py:1064] rank 0 in world size 2 is assigned as DP rank 0, PP rank 0, TP rank 0, EP rank 0\n",
      "INFO 06-07 22:37:09 [model_runner.py:1170] Starting to load model Qwen/Qwen3-Embedding-4B...\n",
      "\u001b[1;36m(VllmWorkerProcess pid=16602)\u001b[0;0m INFO 06-07 22:37:09 [model_runner.py:1170] Starting to load model Qwen/Qwen3-Embedding-4B...\n",
      "INFO 06-07 22:37:09 [weight_utils.py:291] Using model weights format ['*.safetensors']\n",
      "\u001b[1;36m(VllmWorkerProcess pid=16602)\u001b[0;0m INFO 06-07 22:37:09 [weight_utils.py:291] Using model weights format ['*.safetensors']\n",
      "INFO 06-07 22:38:15 [weight_utils.py:307] Time spent downloading weights for Qwen/Qwen3-Embedding-4B: 65.320092 seconds\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Loading safetensors checkpoint shards:   0% Completed | 0/2 [00:00<?, ?it/s]\n",
      "Loading safetensors checkpoint shards:  50% Completed | 1/2 [00:02<00:02,  2.96s/it]\n",
      "Loading safetensors checkpoint shards: 100% Completed | 2/2 [00:09<00:00,  4.92s/it]\n",
      "Loading safetensors checkpoint shards: 100% Completed | 2/2 [00:09<00:00,  4.63s/it]\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO 06-07 22:38:24 [default_loader.py:280] Loading weights took 9.43 seconds\n",
      "\u001b[1;36m(VllmWorkerProcess pid=16602)\u001b[0;0m INFO 06-07 22:38:24 [default_loader.py:280] Loading weights took 9.14 seconds\n",
      "INFO 06-07 22:38:24 [model_runner.py:1202] Model loading took 3.8162 GiB and 75.582441 seconds\n",
      "\u001b[1;36m(VllmWorkerProcess pid=16602)\u001b[0;0m INFO 06-07 22:38:25 [model_runner.py:1202] Model loading took 3.8162 GiB and 75.610664 seconds\n",
      "INFO 06-07 22:38:25 [api_server.py:1336] Starting vLLM API server on http://0.0.0.0:8000\n",
      "INFO 06-07 22:38:25 [launcher.py:28] Available routes are:\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /openapi.json, Methods: GET, HEAD\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /docs, Methods: GET, HEAD\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /docs/oauth2-redirect, Methods: GET, HEAD\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /redoc, Methods: GET, HEAD\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /health, Methods: GET\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /load, Methods: GET\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /ping, Methods: POST\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /ping, Methods: GET\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /tokenize, Methods: POST\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /detokenize, Methods: POST\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /v1/models, Methods: GET\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /version, Methods: GET\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /v1/chat/completions, Methods: POST\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /v1/completions, Methods: POST\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /v1/embeddings, Methods: POST\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /pooling, Methods: POST\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /classify, Methods: POST\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /score, Methods: POST\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /v1/score, Methods: POST\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /v1/audio/transcriptions, Methods: POST\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /rerank, Methods: POST\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /v1/rerank, Methods: POST\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /v2/rerank, Methods: POST\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /invocations, Methods: POST\n",
      "INFO 06-07 22:38:25 [launcher.py:36] Route: /metrics, Methods: GET\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:     Started server process [16093]\n",
      "INFO:     Waiting for application startup.\n",
      "INFO:     Application startup complete.\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import subprocess\n",
    "import threading\n",
    "import time\n",
    "\n",
    "# Set environment variable we need to support dual-GPU on Cirrus\n",
    "os.environ[\"NCCL_P2P_LEVEL\"] = \"NVL\"\n",
    "os.environ[\"VLLM_API_KEY\"] = os.getenv(\"OPENAI_API_KEY\") # set same key for simplicity\n",
    "\n",
    "# https://huggingface.co/spaces/mteb/leaderboard \n",
    "def run_vllm_server():\n",
    "    subprocess.run([\n",
    "        \"vllm\", \"serve\", \"Qwen/Qwen3-Embedding-4B\",\n",
    "        \"--host\", \"0.0.0.0\",\n",
    "        \"--port\", \"8000\",\n",
    "        \"--tensor-parallel-size\", \"2\",\n",
    "        \"--trust-remote-code\",\n",
    "        \"--gpu-memory-utilization\", \"0.4\",\n",
    "        \"--enforce-eager\",\n",
    "        \"--served-model-name\", \"local\",\n",
    "        \"--task\", \"embed\" # Run in embed mode!  (default is 'generate')\n",
    "    ])\n",
    "\n",
    "# Start server in daemon thread\n",
    "server_thread = threading.Thread(target=run_vllm_server, daemon=True)\n",
    "server_thread.start()\n",
    "\n",
    "## give server time to start up:\n",
    "import time\n",
    "# Pause execution for 100 seconds\n",
    "time.sleep(200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9a8397fa-6896-40a5-97d9-1d0c98797b35",
   "metadata": {},
   "outputs": [],
   "source": [
    "## wait for output above to print routes, ending with: \n",
    "## INFO:     Application startup complete.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "24b64902-1305-43e7-9da8-e4d82d097cb5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO 06-07 22:39:51 [logger.py:42] Received request embd-9059fd2aadf84d8288c45ca3ecc8cd3c-0: prompt: ' product down', params: PoolingParams(dimensions=None, additional_metadata=None), prompt_token_ids: [1985, 1495], prompt_embeds shape: None, lora_request: None, prompt_adapter_request: None.\n",
      "INFO 06-07 22:39:51 [engine.py:316] Added request embd-9059fd2aadf84d8288c45ca3ecc8cd3c-0.\n",
      "INFO 06-07 22:39:53 [metrics.py:486] Avg prompt throughput: 0.2 tokens/s, Avg generation throughput: 0.1 tokens/s, Running: 0 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 0.0%, CPU KV cache usage: 0.0%.\n",
      "INFO:     127.0.0.1:36040 - \"POST /v1/embeddings HTTP/1.1\" 200 OK\n",
      "INFO 06-07 22:40:03 [metrics.py:486] Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 0 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 0.0%, CPU KV cache usage: 0.0%.\n"
     ]
    }
   ],
   "source": [
    "## NOTE!  You must wait until the log above finishes and not just the cell.\n",
    "## Connect to the local model\n",
    "from langchain_openai import OpenAIEmbeddings\n",
    "embedding = OpenAIEmbeddings(\n",
    "                 model = \"local\", ## served model name\n",
    "                 api_key = os.getenv(\"OPENAI_API_KEY\"),\n",
    "                 base_url = \"http://localhost:8000/v1\",\n",
    ")\n",
    "\n",
    "## test that the model can do embeddings\n",
    "from langchain_core.vectorstores import InMemoryVectorStore\n",
    "vectorstore = InMemoryVectorStore.from_texts([\"test text\"], embedding=embedding)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "6181a644-e419-4986-a900-44f1d569d244",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO 06-07 23:16:55 [logger.py:42] Received request embd-60cc555da14743da90337ba88edfe7cf-0: prompt: ' Stainless\">\\r\\n', params: PoolingParams(dimensions=None, additional_metadata=None), prompt_token_ids: [32490, 4424], prompt_embeds shape: None, lora_request: None, prompt_adapter_request: None.\n",
      "INFO 06-07 23:16:55 [engine.py:316] Added request embd-60cc555da14743da90337ba88edfe7cf-0.\n",
      "INFO:     127.0.0.1:53594 - \"POST /v1/embeddings HTTP/1.1\" 200 OK\n",
      "INFO 06-07 23:17:05 [metrics.py:486] Avg prompt throughput: 0.1 tokens/s, Avg generation throughput: 0.1 tokens/s, Running: 0 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 0.0%, CPU KV cache usage: 0.0%.\n",
      "INFO 06-07 23:17:15 [metrics.py:486] Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 0 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 0.0%, CPU KV cache usage: 0.0%.\n"
     ]
    }
   ],
   "source": [
    "from langchain_qdrant import QdrantVectorStore\n",
    "from qdrant_client import QdrantClient\n",
    "from qdrant_client.http.models import Distance, VectorParams\n",
    "\n",
    "client = QdrantClient(path = \"hwc_qdrant.db\")\n",
    "\n",
    "client.create_collection(\n",
    "    collection_name=\"demo_collection\",\n",
    "    vectors_config=VectorParams(size=2560, distance=Distance.COSINE),\n",
    ")\n",
    "\n",
    "vector_store = QdrantVectorStore(\n",
    "    client=client,\n",
    "    collection_name=\"demo_collection\",\n",
    "    embedding=embedding\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "ec8d7936-d6b9-4487-9146-c42f855523ec",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO 06-07 23:19:58 [logger.py:42] Received request embd-d18cf8a824104f36a07c970cfa3e999f-0: prompt: 'Iys landscape SL defends and-fired Philadelphia for Utils wh whose.', params: PoolingParams(dimensions=None, additional_metadata=None), prompt_token_ids: [40, 1047, 18414, 16797, 80960, 323, 71578, 19335, 369, 17954, 420, 6693, 13], prompt_embeds shape: None, lora_request: None, prompt_adapter_request: None.\n",
      "INFO 06-07 23:19:58 [logger.py:42] Received request embd-d18cf8a824104f36a07c970cfa3e999f-1: prompt: 'eg observ.price for guitar is=! andshional,iz aik of  deep(long\\\\Category.', params: PoolingParams(dimensions=None, additional_metadata=None), prompt_token_ids: [791, 9282, 18057, 369, 16986, 374, 74649, 323, 927, 3914, 11, 449, 264, 1579, 315, 220, 5538, 12628, 69823, 13], prompt_embeds shape: None, lora_request: None, prompt_adapter_request: None.\n",
      "INFO 06-07 23:19:58 [logger.py:42] Received request embd-d18cf8a824104f36a07c970cfa3e999f-2: prompt: '.setPropertyata dram j.Iiz��shareine *) eachessso!', params: PoolingParams(dimensions=None, additional_metadata=None), prompt_token_ids: [31233, 459, 13548, 502, 2447, 449, 23272, 19368, 482, 2586, 1817, 433, 704, 0], prompt_embeds shape: None, lora_request: None, prompt_adapter_request: None.\n",
      "INFO 06-07 23:19:58 [logger.py:42] Received request embd-d18cf8a824104f36a07c970cfa3e999f-3: prompt: ' Wasook.offset_T the means());\\r\\n and.after $1 +\\n in Alex.', params: PoolingParams(dimensions=None, additional_metadata=None), prompt_token_ids: [14804, 1941, 14760, 1139, 279, 3363, 6201, 323, 40606, 400, 16, 3610, 304, 8515, 13], prompt_embeds shape: None, lora_request: None, prompt_adapter_request: None.\n",
      "INFO 06-07 23:19:58 [logger.py:42] Received request embd-d18cf8a824104f36a07c970cfa3e999f-4: prompt: ' Traffic! });\\n\\nthisata station=[. Iprra games to================ess ed.', params: PoolingParams(dimensions=None, additional_metadata=None), prompt_token_ids: [36981, 0, 3011, 574, 459, 8056, 5818, 13, 358, 649, 956, 3868, 311, 1518, 433, 1578, 13], prompt_embeds shape: None, lora_request: None, prompt_adapter_request: None.\n",
      "INFO 06-07 23:19:58 [logger.py:42] Received request embd-d18cf8a824104f36a07c970cfa3e999f-5: prompt: '.Control the jPairutable theBe?\\t\\t\\t\\n whReg to (!so.', params: PoolingParams(dimensions=None, additional_metadata=None), prompt_token_ids: [3957, 279, 502, 12443, 5922, 279, 3430, 30, 4557, 420, 3477, 311, 1505, 704, 13], prompt_embeds shape: None, lora_request: None, prompt_adapter_request: None.\n",
      "INFO 06-07 23:19:58 [logger.py:42] Received request embd-d18cf8a824104f36a07c970cfa3e999f-6: prompt: 'eg between ==== Marine City in the actionutil][.', params: PoolingParams(dimensions=None, additional_metadata=None), prompt_token_ids: [791, 1948, 220, 605, 22963, 4311, 304, 279, 1917, 1314, 1457, 13], prompt_embeds shape: None, lora_request: None, prompt_adapter_request: None.\n",
      "INFO 06-07 23:19:58 [logger.py:42] Received request embd-d18cf8a824104f36a07c970cfa3e999f-7: prompt: ' parad Bad is theideo JavaScript for(key modelRO,ault-group Russian!', params: PoolingParams(dimensions=None, additional_metadata=None), prompt_token_ids: [27317, 11461, 374, 279, 1888, 12914, 369, 4857, 1614, 1285, 11, 945, 4351, 8522, 0], prompt_embeds shape: None, lora_request: None, prompt_adapter_request: None.\n",
      "INFO 06-07 23:19:58 [logger.py:42] Received request embd-d18cf8a824104f36a07c970cfa3e999f-8: prompt: 'eg sysyear is_h ehAGE having.io toReceiveMemoryWarning of a strangers.', params: PoolingParams(dimensions=None, additional_metadata=None), prompt_token_ids: [791, 5708, 3157, 374, 1523, 220, 2636, 3585, 3432, 4245, 311, 24087, 315, 264, 39621, 13], prompt_embeds shape: None, lora_request: None, prompt_adapter_request: None.\n",
      "INFO 06-07 23:19:58 [logger.py:42] Received request embd-d18cf8a824104f36a07c970cfa3e999f-9: prompt: 'Iull a éSec Irefoute to �Touchacency', params: PoolingParams(dimensions=None, additional_metadata=None), prompt_token_ids: [40, 617, 264, 3958, 8430, 358, 1097, 2133, 311, 636, 11309, 40624], prompt_embeds shape: None, lora_request: None, prompt_adapter_request: None.\n",
      "INFO 06-07 23:19:58 [engine.py:316] Added request embd-d18cf8a824104f36a07c970cfa3e999f-0.\n",
      "INFO 06-07 23:19:58 [engine.py:316] Added request embd-d18cf8a824104f36a07c970cfa3e999f-1.\n",
      "INFO 06-07 23:19:58 [engine.py:316] Added request embd-d18cf8a824104f36a07c970cfa3e999f-2.\n",
      "INFO 06-07 23:19:58 [engine.py:316] Added request embd-d18cf8a824104f36a07c970cfa3e999f-3.\n",
      "INFO 06-07 23:19:58 [engine.py:316] Added request embd-d18cf8a824104f36a07c970cfa3e999f-4.\n",
      "INFO 06-07 23:19:58 [engine.py:316] Added request embd-d18cf8a824104f36a07c970cfa3e999f-5.\n",
      "INFO 06-07 23:19:58 [engine.py:316] Added request embd-d18cf8a824104f36a07c970cfa3e999f-6.\n",
      "INFO 06-07 23:19:58 [engine.py:316] Added request embd-d18cf8a824104f36a07c970cfa3e999f-7.\n",
      "INFO 06-07 23:19:58 [engine.py:316] Added request embd-d18cf8a824104f36a07c970cfa3e999f-8.\n",
      "INFO 06-07 23:19:58 [engine.py:316] Added request embd-d18cf8a824104f36a07c970cfa3e999f-9.\n",
      "INFO:     127.0.0.1:35840 - \"POST /v1/embeddings HTTP/1.1\" 200 OK\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "['878ac59d-e276-4d29-b715-6b20cd0f76f8',\n",
       " '5e5c960e-7c02-446d-9ad4-ca930ec3b57a',\n",
       " '8f478874-a730-4e19-b5f2-ef3caeb96a9d',\n",
       " 'ed8dd908-a4da-4b20-b3d3-b2778846298d',\n",
       " 'cc487e15-f7f3-46b6-a0a4-d0be9955cdaa',\n",
       " '5f380da0-9b78-4051-beea-6aa98421df33',\n",
       " 'd5fa20d1-cc94-4613-ac64-b5ba3830589c',\n",
       " 'cb0cbe18-5118-45f6-ae9c-1452b43cb92e',\n",
       " '15181e21-fa5d-45d0-b97b-3e206d671101',\n",
       " '1e769b7d-d327-465e-9b5b-3274b159ede3']"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO 06-07 23:20:08 [metrics.py:486] Avg prompt throughput: 12.1 tokens/s, Avg generation throughput: 0.8 tokens/s, Running: 0 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 0.0%, CPU KV cache usage: 0.0%.\n",
      "INFO 06-07 23:20:19 [metrics.py:486] Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 0 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 0.0%, CPU KV cache usage: 0.0%.\n"
     ]
    }
   ],
   "source": [
    "from uuid import uuid4\n",
    "\n",
    "from langchain_core.documents import Document\n",
    "\n",
    "document_1 = Document(\n",
    "    page_content=\"I had chocolate chip pancakes and scrambled eggs for breakfast this morning.\",\n",
    "    metadata={\"source\": \"tweet\"},\n",
    ")\n",
    "\n",
    "document_2 = Document(\n",
    "    page_content=\"The weather forecast for tomorrow is cloudy and overcast, with a high of 62 degrees Fahrenheit.\",\n",
    "    metadata={\"source\": \"news\"},\n",
    ")\n",
    "\n",
    "document_3 = Document(\n",
    "    page_content=\"Building an exciting new project with LangChain - come check it out!\",\n",
    "    metadata={\"source\": \"tweet\"},\n",
    ")\n",
    "\n",
    "document_4 = Document(\n",
    "    page_content=\"Robbers broke into the city bank and stole $1 million in cash.\",\n",
    "    metadata={\"source\": \"news\"},\n",
    ")\n",
    "\n",
    "document_5 = Document(\n",
    "    page_content=\"Wow! That was an amazing movie. I can't wait to see it again.\",\n",
    "    metadata={\"source\": \"tweet\"},\n",
    ")\n",
    "\n",
    "document_6 = Document(\n",
    "    page_content=\"Is the new iPhone worth the price? Read this review to find out.\",\n",
    "    metadata={\"source\": \"website\"},\n",
    ")\n",
    "\n",
    "document_7 = Document(\n",
    "    page_content=\"The top 10 soccer players in the world right now.\",\n",
    "    metadata={\"source\": \"website\"},\n",
    ")\n",
    "\n",
    "document_8 = Document(\n",
    "    page_content=\"LangGraph is the best framework for building stateful, agentic applications!\",\n",
    "    metadata={\"source\": \"tweet\"},\n",
    ")\n",
    "\n",
    "document_9 = Document(\n",
    "    page_content=\"The stock market is down 500 points today due to fears of a recession.\",\n",
    "    metadata={\"source\": \"news\"},\n",
    ")\n",
    "\n",
    "document_10 = Document(\n",
    "    page_content=\"I have a bad feeling I am going to get deleted :(\",\n",
    "    metadata={\"source\": \"tweet\"},\n",
    ")\n",
    "\n",
    "documents = [\n",
    "    document_1,\n",
    "    document_2,\n",
    "    document_3,\n",
    "    document_4,\n",
    "    document_5,\n",
    "    document_6,\n",
    "    document_7,\n",
    "    document_8,\n",
    "    document_9,\n",
    "    document_10,\n",
    "]\n",
    "uuids = [str(uuid4()) for _ in range(len(documents))]\n",
    "vector_store.add_documents(documents=documents, ids=uuids)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "95ed10f3-5339-40cd-bf16-b0854f8b4b91",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import requests\n",
    "import zipfile\n",
    "import pathlib\n",
    "from langchain_community.document_loaders import PyPDFLoader\n",
    "\n",
    "def download_and_unzip(url, output_dir):\n",
    "    response = requests.get(url)\n",
    "    zip_file_path = os.path.basename(url)\n",
    "    with open(zip_file_path, 'wb') as f:\n",
    "        f.write(response.content)\n",
    "    with zipfile.ZipFile(zip_file_path, 'r') as zip_ref:\n",
    "        zip_ref.extractall(output_dir)\n",
    "    os.remove(zip_file_path)\n",
    "\n",
    "def pdf_loader(path):\n",
    "    all_documents = []\n",
    "    docs_dir = pathlib.Path(path)\n",
    "    for file in docs_dir.iterdir():\n",
    "        loader = PyPDFLoader(file)\n",
    "        documents = loader.load()\n",
    "        all_documents.extend(documents)\n",
    "    return all_documents\n",
    "\n",
    "\n",
    "download_and_unzip(\"https://minio.carlboettiger.info/public-data/hwc.zip\", 'hwc')\n",
    "docs = pdf_loader('hwc/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "95d3e9a3-7334-44ba-a4bc-e7bfc4076358",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Build a retrival agent\n",
    "from langchain_core.vectorstores import InMemoryVectorStore\n",
    "from langchain_text_splitters import RecursiveCharacterTextSplitter\n",
    "text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)\n",
    "splits = text_splitter.split_documents(docs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "fd8bcc13-d06d-43dd-9e06-4f29da803133",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "ename": "ResponseHandlingException",
     "evalue": "[Errno 111] Connection refused",
     "output_type": "error",
     "traceback": [
      "\u001b[31m---------------------------------------------------------------------------\u001b[39m",
      "\u001b[31mConnectError\u001b[39m                              Traceback (most recent call last)",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/httpx/_transports/default.py:101\u001b[39m, in \u001b[36mmap_httpcore_exceptions\u001b[39m\u001b[34m()\u001b[39m\n\u001b[32m    100\u001b[39m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[32m--> \u001b[39m\u001b[32m101\u001b[39m     \u001b[38;5;28;01myield\u001b[39;00m\n\u001b[32m    102\u001b[39m \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mException\u001b[39;00m \u001b[38;5;28;01mas\u001b[39;00m exc:\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/httpx/_transports/default.py:250\u001b[39m, in \u001b[36mHTTPTransport.handle_request\u001b[39m\u001b[34m(self, request)\u001b[39m\n\u001b[32m    249\u001b[39m \u001b[38;5;28;01mwith\u001b[39;00m map_httpcore_exceptions():\n\u001b[32m--> \u001b[39m\u001b[32m250\u001b[39m     resp = \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43m_pool\u001b[49m\u001b[43m.\u001b[49m\u001b[43mhandle_request\u001b[49m\u001b[43m(\u001b[49m\u001b[43mreq\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m    252\u001b[39m \u001b[38;5;28;01massert\u001b[39;00m \u001b[38;5;28misinstance\u001b[39m(resp.stream, typing.Iterable)\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/httpcore/_sync/connection_pool.py:256\u001b[39m, in \u001b[36mConnectionPool.handle_request\u001b[39m\u001b[34m(self, request)\u001b[39m\n\u001b[32m    255\u001b[39m     \u001b[38;5;28mself\u001b[39m._close_connections(closing)\n\u001b[32m--> \u001b[39m\u001b[32m256\u001b[39m     \u001b[38;5;28;01mraise\u001b[39;00m exc \u001b[38;5;28;01mfrom\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[38;5;28;01mNone\u001b[39;00m\n\u001b[32m    258\u001b[39m \u001b[38;5;66;03m# Return the response. Note that in this case we still have to manage\u001b[39;00m\n\u001b[32m    259\u001b[39m \u001b[38;5;66;03m# the point at which the response is closed.\u001b[39;00m\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/httpcore/_sync/connection_pool.py:236\u001b[39m, in \u001b[36mConnectionPool.handle_request\u001b[39m\u001b[34m(self, request)\u001b[39m\n\u001b[32m    234\u001b[39m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[32m    235\u001b[39m     \u001b[38;5;66;03m# Send the request on the assigned connection.\u001b[39;00m\n\u001b[32m--> \u001b[39m\u001b[32m236\u001b[39m     response = \u001b[43mconnection\u001b[49m\u001b[43m.\u001b[49m\u001b[43mhandle_request\u001b[49m\u001b[43m(\u001b[49m\n\u001b[32m    237\u001b[39m \u001b[43m        \u001b[49m\u001b[43mpool_request\u001b[49m\u001b[43m.\u001b[49m\u001b[43mrequest\u001b[49m\n\u001b[32m    238\u001b[39m \u001b[43m    \u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m    239\u001b[39m \u001b[38;5;28;01mexcept\u001b[39;00m ConnectionNotAvailable:\n\u001b[32m    240\u001b[39m     \u001b[38;5;66;03m# In some cases a connection may initially be available to\u001b[39;00m\n\u001b[32m    241\u001b[39m     \u001b[38;5;66;03m# handle a request, but then become unavailable.\u001b[39;00m\n\u001b[32m    242\u001b[39m     \u001b[38;5;66;03m#\u001b[39;00m\n\u001b[32m    243\u001b[39m     \u001b[38;5;66;03m# In this case we clear the connection and try again.\u001b[39;00m\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/httpcore/_sync/connection.py:101\u001b[39m, in \u001b[36mHTTPConnection.handle_request\u001b[39m\u001b[34m(self, request)\u001b[39m\n\u001b[32m    100\u001b[39m     \u001b[38;5;28mself\u001b[39m._connect_failed = \u001b[38;5;28;01mTrue\u001b[39;00m\n\u001b[32m--> \u001b[39m\u001b[32m101\u001b[39m     \u001b[38;5;28;01mraise\u001b[39;00m exc\n\u001b[32m    103\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m._connection.handle_request(request)\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/httpcore/_sync/connection.py:78\u001b[39m, in \u001b[36mHTTPConnection.handle_request\u001b[39m\u001b[34m(self, request)\u001b[39m\n\u001b[32m     77\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m._connection \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n\u001b[32m---> \u001b[39m\u001b[32m78\u001b[39m     stream = \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43m_connect\u001b[49m\u001b[43m(\u001b[49m\u001b[43mrequest\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m     80\u001b[39m     ssl_object = stream.get_extra_info(\u001b[33m\"\u001b[39m\u001b[33mssl_object\u001b[39m\u001b[33m\"\u001b[39m)\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/httpcore/_sync/connection.py:124\u001b[39m, in \u001b[36mHTTPConnection._connect\u001b[39m\u001b[34m(self, request)\u001b[39m\n\u001b[32m    123\u001b[39m \u001b[38;5;28;01mwith\u001b[39;00m Trace(\u001b[33m\"\u001b[39m\u001b[33mconnect_tcp\u001b[39m\u001b[33m\"\u001b[39m, logger, request, kwargs) \u001b[38;5;28;01mas\u001b[39;00m trace:\n\u001b[32m--> \u001b[39m\u001b[32m124\u001b[39m     stream = \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43m_network_backend\u001b[49m\u001b[43m.\u001b[49m\u001b[43mconnect_tcp\u001b[49m\u001b[43m(\u001b[49m\u001b[43m*\u001b[49m\u001b[43m*\u001b[49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m    125\u001b[39m     trace.return_value = stream\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/httpcore/_backends/sync.py:207\u001b[39m, in \u001b[36mSyncBackend.connect_tcp\u001b[39m\u001b[34m(self, host, port, timeout, local_address, socket_options)\u001b[39m\n\u001b[32m    202\u001b[39m exc_map: ExceptionMapping = {\n\u001b[32m    203\u001b[39m     socket.timeout: ConnectTimeout,\n\u001b[32m    204\u001b[39m     \u001b[38;5;167;01mOSError\u001b[39;00m: ConnectError,\n\u001b[32m    205\u001b[39m }\n\u001b[32m--> \u001b[39m\u001b[32m207\u001b[39m \u001b[38;5;28;01mwith\u001b[39;00m map_exceptions(exc_map):\n\u001b[32m    208\u001b[39m     sock = socket.create_connection(\n\u001b[32m    209\u001b[39m         address,\n\u001b[32m    210\u001b[39m         timeout,\n\u001b[32m    211\u001b[39m         source_address=source_address,\n\u001b[32m    212\u001b[39m     )\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/contextlib.py:158\u001b[39m, in \u001b[36m_GeneratorContextManager.__exit__\u001b[39m\u001b[34m(self, typ, value, traceback)\u001b[39m\n\u001b[32m    157\u001b[39m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[32m--> \u001b[39m\u001b[32m158\u001b[39m     \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43mgen\u001b[49m\u001b[43m.\u001b[49m\u001b[43mthrow\u001b[49m\u001b[43m(\u001b[49m\u001b[43mvalue\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m    159\u001b[39m \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mStopIteration\u001b[39;00m \u001b[38;5;28;01mas\u001b[39;00m exc:\n\u001b[32m    160\u001b[39m     \u001b[38;5;66;03m# Suppress StopIteration *unless* it's the same exception that\u001b[39;00m\n\u001b[32m    161\u001b[39m     \u001b[38;5;66;03m# was passed to throw().  This prevents a StopIteration\u001b[39;00m\n\u001b[32m    162\u001b[39m     \u001b[38;5;66;03m# raised inside the \"with\" statement from being suppressed.\u001b[39;00m\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/httpcore/_exceptions.py:14\u001b[39m, in \u001b[36mmap_exceptions\u001b[39m\u001b[34m(map)\u001b[39m\n\u001b[32m     13\u001b[39m     \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28misinstance\u001b[39m(exc, from_exc):\n\u001b[32m---> \u001b[39m\u001b[32m14\u001b[39m         \u001b[38;5;28;01mraise\u001b[39;00m to_exc(exc) \u001b[38;5;28;01mfrom\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[34;01mexc\u001b[39;00m\n\u001b[32m     15\u001b[39m \u001b[38;5;28;01mraise\u001b[39;00m\n",
      "\u001b[31mConnectError\u001b[39m: [Errno 111] Connection refused",
      "\nThe above exception was the direct cause of the following exception:\n",
      "\u001b[31mConnectError\u001b[39m                              Traceback (most recent call last)",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/qdrant_client/http/api_client.py:129\u001b[39m, in \u001b[36mApiClient.send_inner\u001b[39m\u001b[34m(self, request)\u001b[39m\n\u001b[32m    128\u001b[39m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[32m--> \u001b[39m\u001b[32m129\u001b[39m     response = \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43m_client\u001b[49m\u001b[43m.\u001b[49m\u001b[43msend\u001b[49m\u001b[43m(\u001b[49m\u001b[43mrequest\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m    130\u001b[39m \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mException\u001b[39;00m \u001b[38;5;28;01mas\u001b[39;00m e:\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/httpx/_client.py:914\u001b[39m, in \u001b[36mClient.send\u001b[39m\u001b[34m(self, request, stream, auth, follow_redirects)\u001b[39m\n\u001b[32m    912\u001b[39m auth = \u001b[38;5;28mself\u001b[39m._build_request_auth(request, auth)\n\u001b[32m--> \u001b[39m\u001b[32m914\u001b[39m response = \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43m_send_handling_auth\u001b[49m\u001b[43m(\u001b[49m\n\u001b[32m    915\u001b[39m \u001b[43m    \u001b[49m\u001b[43mrequest\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    916\u001b[39m \u001b[43m    \u001b[49m\u001b[43mauth\u001b[49m\u001b[43m=\u001b[49m\u001b[43mauth\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    917\u001b[39m \u001b[43m    \u001b[49m\u001b[43mfollow_redirects\u001b[49m\u001b[43m=\u001b[49m\u001b[43mfollow_redirects\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    918\u001b[39m \u001b[43m    \u001b[49m\u001b[43mhistory\u001b[49m\u001b[43m=\u001b[49m\u001b[43m[\u001b[49m\u001b[43m]\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    919\u001b[39m \u001b[43m\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m    920\u001b[39m \u001b[38;5;28;01mtry\u001b[39;00m:\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/httpx/_client.py:942\u001b[39m, in \u001b[36mClient._send_handling_auth\u001b[39m\u001b[34m(self, request, auth, follow_redirects, history)\u001b[39m\n\u001b[32m    941\u001b[39m \u001b[38;5;28;01mwhile\u001b[39;00m \u001b[38;5;28;01mTrue\u001b[39;00m:\n\u001b[32m--> \u001b[39m\u001b[32m942\u001b[39m     response = \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43m_send_handling_redirects\u001b[49m\u001b[43m(\u001b[49m\n\u001b[32m    943\u001b[39m \u001b[43m        \u001b[49m\u001b[43mrequest\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    944\u001b[39m \u001b[43m        \u001b[49m\u001b[43mfollow_redirects\u001b[49m\u001b[43m=\u001b[49m\u001b[43mfollow_redirects\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    945\u001b[39m \u001b[43m        \u001b[49m\u001b[43mhistory\u001b[49m\u001b[43m=\u001b[49m\u001b[43mhistory\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    946\u001b[39m \u001b[43m    \u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m    947\u001b[39m     \u001b[38;5;28;01mtry\u001b[39;00m:\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/httpx/_client.py:979\u001b[39m, in \u001b[36mClient._send_handling_redirects\u001b[39m\u001b[34m(self, request, follow_redirects, history)\u001b[39m\n\u001b[32m    977\u001b[39m     hook(request)\n\u001b[32m--> \u001b[39m\u001b[32m979\u001b[39m response = \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43m_send_single_request\u001b[49m\u001b[43m(\u001b[49m\u001b[43mrequest\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m    980\u001b[39m \u001b[38;5;28;01mtry\u001b[39;00m:\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/httpx/_client.py:1014\u001b[39m, in \u001b[36mClient._send_single_request\u001b[39m\u001b[34m(self, request)\u001b[39m\n\u001b[32m   1013\u001b[39m \u001b[38;5;28;01mwith\u001b[39;00m request_context(request=request):\n\u001b[32m-> \u001b[39m\u001b[32m1014\u001b[39m     response = \u001b[43mtransport\u001b[49m\u001b[43m.\u001b[49m\u001b[43mhandle_request\u001b[49m\u001b[43m(\u001b[49m\u001b[43mrequest\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m   1016\u001b[39m \u001b[38;5;28;01massert\u001b[39;00m \u001b[38;5;28misinstance\u001b[39m(response.stream, SyncByteStream)\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/httpx/_transports/default.py:249\u001b[39m, in \u001b[36mHTTPTransport.handle_request\u001b[39m\u001b[34m(self, request)\u001b[39m\n\u001b[32m    237\u001b[39m req = httpcore.Request(\n\u001b[32m    238\u001b[39m     method=request.method,\n\u001b[32m    239\u001b[39m     url=httpcore.URL(\n\u001b[32m   (...)\u001b[39m\u001b[32m    247\u001b[39m     extensions=request.extensions,\n\u001b[32m    248\u001b[39m )\n\u001b[32m--> \u001b[39m\u001b[32m249\u001b[39m \u001b[38;5;28;01mwith\u001b[39;00m map_httpcore_exceptions():\n\u001b[32m    250\u001b[39m     resp = \u001b[38;5;28mself\u001b[39m._pool.handle_request(req)\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/contextlib.py:158\u001b[39m, in \u001b[36m_GeneratorContextManager.__exit__\u001b[39m\u001b[34m(self, typ, value, traceback)\u001b[39m\n\u001b[32m    157\u001b[39m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[32m--> \u001b[39m\u001b[32m158\u001b[39m     \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43mgen\u001b[49m\u001b[43m.\u001b[49m\u001b[43mthrow\u001b[49m\u001b[43m(\u001b[49m\u001b[43mvalue\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m    159\u001b[39m \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mStopIteration\u001b[39;00m \u001b[38;5;28;01mas\u001b[39;00m exc:\n\u001b[32m    160\u001b[39m     \u001b[38;5;66;03m# Suppress StopIteration *unless* it's the same exception that\u001b[39;00m\n\u001b[32m    161\u001b[39m     \u001b[38;5;66;03m# was passed to throw().  This prevents a StopIteration\u001b[39;00m\n\u001b[32m    162\u001b[39m     \u001b[38;5;66;03m# raised inside the \"with\" statement from being suppressed.\u001b[39;00m\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/httpx/_transports/default.py:118\u001b[39m, in \u001b[36mmap_httpcore_exceptions\u001b[39m\u001b[34m()\u001b[39m\n\u001b[32m    117\u001b[39m message = \u001b[38;5;28mstr\u001b[39m(exc)\n\u001b[32m--> \u001b[39m\u001b[32m118\u001b[39m \u001b[38;5;28;01mraise\u001b[39;00m mapped_exc(message) \u001b[38;5;28;01mfrom\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[34;01mexc\u001b[39;00m\n",
      "\u001b[31mConnectError\u001b[39m: [Errno 111] Connection refused",
      "\nDuring handling of the above exception, another exception occurred:\n",
      "\u001b[31mResponseHandlingException\u001b[39m                 Traceback (most recent call last)",
      "\u001b[36mCell\u001b[39m\u001b[36m \u001b[39m\u001b[32mIn[27]\u001b[39m\u001b[32m, line 2\u001b[39m\n\u001b[32m      1\u001b[39m \u001b[38;5;66;03m# slow part here, runs on remote GPU\u001b[39;00m\n\u001b[32m----> \u001b[39m\u001b[32m2\u001b[39m vectorstore = \u001b[43mvector_store\u001b[49m\u001b[43m.\u001b[49m\u001b[43mfrom_documents\u001b[49m\u001b[43m(\u001b[49m\u001b[43mdocuments\u001b[49m\u001b[43m=\u001b[49m\u001b[43msplits\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43membedding\u001b[49m\u001b[43m \u001b[49m\u001b[43m=\u001b[49m\u001b[43m \u001b[49m\u001b[43membedding\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m      3\u001b[39m retriever = vectorstore.as_retriever()\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/langchain_core/vectorstores/base.py:848\u001b[39m, in \u001b[36mVectorStore.from_documents\u001b[39m\u001b[34m(cls, documents, embedding, **kwargs)\u001b[39m\n\u001b[32m    845\u001b[39m     \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28many\u001b[39m(ids):\n\u001b[32m    846\u001b[39m         kwargs[\u001b[33m\"\u001b[39m\u001b[33mids\u001b[39m\u001b[33m\"\u001b[39m] = ids\n\u001b[32m--> \u001b[39m\u001b[32m848\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28;43mcls\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43mfrom_texts\u001b[49m\u001b[43m(\u001b[49m\u001b[43mtexts\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43membedding\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mmetadatas\u001b[49m\u001b[43m=\u001b[49m\u001b[43mmetadatas\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43m*\u001b[49m\u001b[43m*\u001b[49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/langchain_qdrant/qdrant.py:343\u001b[39m, in \u001b[36mQdrantVectorStore.from_texts\u001b[39m\u001b[34m(cls, texts, embedding, metadatas, ids, collection_name, location, url, port, grpc_port, prefer_grpc, https, api_key, prefix, timeout, host, path, distance, content_payload_key, metadata_payload_key, vector_name, retrieval_mode, sparse_embedding, sparse_vector_name, collection_create_options, vector_params, sparse_vector_params, batch_size, force_recreate, validate_embeddings, validate_collection_config, **kwargs)\u001b[39m\n\u001b[32m    311\u001b[39m \u001b[38;5;250m\u001b[39m\u001b[33;03m\"\"\"Construct an instance of `QdrantVectorStore` from a list of texts.\u001b[39;00m\n\u001b[32m    312\u001b[39m \n\u001b[32m    313\u001b[39m \u001b[33;03mThis is a user-friendly interface that:\u001b[39;00m\n\u001b[32m   (...)\u001b[39m\u001b[32m    326\u001b[39m \u001b[33;03m    qdrant = Qdrant.from_texts(texts, embeddings, url=\"http://localhost:6333\")\u001b[39;00m\n\u001b[32m    327\u001b[39m \u001b[33;03m\"\"\"\u001b[39;00m\n\u001b[32m    328\u001b[39m client_options = {\n\u001b[32m    329\u001b[39m     \u001b[33m\"\u001b[39m\u001b[33mlocation\u001b[39m\u001b[33m\"\u001b[39m: location,\n\u001b[32m    330\u001b[39m     \u001b[33m\"\u001b[39m\u001b[33murl\u001b[39m\u001b[33m\"\u001b[39m: url,\n\u001b[32m   (...)\u001b[39m\u001b[32m    340\u001b[39m     **kwargs,\n\u001b[32m    341\u001b[39m }\n\u001b[32m--> \u001b[39m\u001b[32m343\u001b[39m qdrant = \u001b[38;5;28;43mcls\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43mconstruct_instance\u001b[49m\u001b[43m(\u001b[49m\n\u001b[32m    344\u001b[39m \u001b[43m    \u001b[49m\u001b[43membedding\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    345\u001b[39m \u001b[43m    \u001b[49m\u001b[43mretrieval_mode\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    346\u001b[39m \u001b[43m    \u001b[49m\u001b[43msparse_embedding\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    347\u001b[39m \u001b[43m    \u001b[49m\u001b[43mclient_options\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    348\u001b[39m \u001b[43m    \u001b[49m\u001b[43mcollection_name\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    349\u001b[39m \u001b[43m    \u001b[49m\u001b[43mdistance\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    350\u001b[39m \u001b[43m    \u001b[49m\u001b[43mcontent_payload_key\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    351\u001b[39m \u001b[43m    \u001b[49m\u001b[43mmetadata_payload_key\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    352\u001b[39m \u001b[43m    \u001b[49m\u001b[43mvector_name\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    353\u001b[39m \u001b[43m    \u001b[49m\u001b[43msparse_vector_name\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    354\u001b[39m \u001b[43m    \u001b[49m\u001b[43mforce_recreate\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    355\u001b[39m \u001b[43m    \u001b[49m\u001b[43mcollection_create_options\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    356\u001b[39m \u001b[43m    \u001b[49m\u001b[43mvector_params\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    357\u001b[39m \u001b[43m    \u001b[49m\u001b[43msparse_vector_params\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    358\u001b[39m \u001b[43m    \u001b[49m\u001b[43mvalidate_embeddings\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    359\u001b[39m \u001b[43m    \u001b[49m\u001b[43mvalidate_collection_config\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    360\u001b[39m \u001b[43m\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m    361\u001b[39m qdrant.add_texts(texts, metadatas, ids, batch_size)\n\u001b[32m    362\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m qdrant\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/langchain_qdrant/qdrant.py:810\u001b[39m, in \u001b[36mQdrantVectorStore.construct_instance\u001b[39m\u001b[34m(cls, embedding, retrieval_mode, sparse_embedding, client_options, collection_name, distance, content_payload_key, metadata_payload_key, vector_name, sparse_vector_name, force_recreate, collection_create_options, vector_params, sparse_vector_params, validate_embeddings, validate_collection_config)\u001b[39m\n\u001b[32m    807\u001b[39m collection_name = collection_name \u001b[38;5;129;01mor\u001b[39;00m uuid.uuid4().hex\n\u001b[32m    808\u001b[39m client = QdrantClient(**client_options)\n\u001b[32m--> \u001b[39m\u001b[32m810\u001b[39m collection_exists = \u001b[43mclient\u001b[49m\u001b[43m.\u001b[49m\u001b[43mcollection_exists\u001b[49m\u001b[43m(\u001b[49m\u001b[43mcollection_name\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m    812\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m collection_exists \u001b[38;5;129;01mand\u001b[39;00m force_recreate:\n\u001b[32m    813\u001b[39m     client.delete_collection(collection_name)\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/qdrant_client/qdrant_client.py:2240\u001b[39m, in \u001b[36mQdrantClient.collection_exists\u001b[39m\u001b[34m(self, collection_name, **kwargs)\u001b[39m\n\u001b[32m   2230\u001b[39m \u001b[38;5;250m\u001b[39m\u001b[33;03m\"\"\"Check whether collection already exists\u001b[39;00m\n\u001b[32m   2231\u001b[39m \n\u001b[32m   2232\u001b[39m \u001b[33;03mArgs:\u001b[39;00m\n\u001b[32m   (...)\u001b[39m\u001b[32m   2236\u001b[39m \u001b[33;03m    True if collection exists, False if not\u001b[39;00m\n\u001b[32m   2237\u001b[39m \u001b[33;03m\"\"\"\u001b[39;00m\n\u001b[32m   2238\u001b[39m \u001b[38;5;28;01massert\u001b[39;00m \u001b[38;5;28mlen\u001b[39m(kwargs) == \u001b[32m0\u001b[39m, \u001b[33mf\u001b[39m\u001b[33m\"\u001b[39m\u001b[33mUnknown arguments: \u001b[39m\u001b[38;5;132;01m{\u001b[39;00m\u001b[38;5;28mlist\u001b[39m(kwargs.keys())\u001b[38;5;132;01m}\u001b[39;00m\u001b[33m\"\u001b[39m\n\u001b[32m-> \u001b[39m\u001b[32m2240\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43m_client\u001b[49m\u001b[43m.\u001b[49m\u001b[43mcollection_exists\u001b[49m\u001b[43m(\u001b[49m\u001b[43mcollection_name\u001b[49m\u001b[43m=\u001b[49m\u001b[43mcollection_name\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43m*\u001b[49m\u001b[43m*\u001b[49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/qdrant_client/qdrant_remote.py:2597\u001b[39m, in \u001b[36mQdrantRemote.collection_exists\u001b[39m\u001b[34m(self, collection_name, **kwargs)\u001b[39m\n\u001b[32m   2591\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m._prefer_grpc:\n\u001b[32m   2592\u001b[39m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m.grpc_collections.CollectionExists(\n\u001b[32m   2593\u001b[39m         grpc.CollectionExistsRequest(collection_name=collection_name),\n\u001b[32m   2594\u001b[39m         timeout=\u001b[38;5;28mself\u001b[39m._timeout,\n\u001b[32m   2595\u001b[39m     ).result.exists\n\u001b[32m-> \u001b[39m\u001b[32m2597\u001b[39m result: Optional[models.CollectionExistence] = \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43mhttp\u001b[49m\u001b[43m.\u001b[49m\u001b[43mcollections_api\u001b[49m\u001b[43m.\u001b[49m\u001b[43mcollection_exists\u001b[49m\u001b[43m(\u001b[49m\n\u001b[32m   2598\u001b[39m \u001b[43m    \u001b[49m\u001b[43mcollection_name\u001b[49m\u001b[43m=\u001b[49m\u001b[43mcollection_name\u001b[49m\n\u001b[32m   2599\u001b[39m \u001b[43m\u001b[49m\u001b[43m)\u001b[49m.result\n\u001b[32m   2600\u001b[39m \u001b[38;5;28;01massert\u001b[39;00m result \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m, \u001b[33m\"\u001b[39m\u001b[33mCollection exists returned None\u001b[39m\u001b[33m\"\u001b[39m\n\u001b[32m   2601\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m result.exists\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/qdrant_client/http/api/collections_api.py:281\u001b[39m, in \u001b[36mSyncCollectionsApi.collection_exists\u001b[39m\u001b[34m(self, collection_name)\u001b[39m\n\u001b[32m    274\u001b[39m \u001b[38;5;28;01mdef\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[34mcollection_exists\u001b[39m(\n\u001b[32m    275\u001b[39m     \u001b[38;5;28mself\u001b[39m,\n\u001b[32m    276\u001b[39m     collection_name: \u001b[38;5;28mstr\u001b[39m,\n\u001b[32m    277\u001b[39m ) -> m.InlineResponse2007:\n\u001b[32m    278\u001b[39m \u001b[38;5;250m    \u001b[39m\u001b[33;03m\"\"\"\u001b[39;00m\n\u001b[32m    279\u001b[39m \u001b[33;03m    Returns \\\"true\\\" if the given collection name exists, and \\\"false\\\" otherwise\u001b[39;00m\n\u001b[32m    280\u001b[39m \u001b[33;03m    \"\"\"\u001b[39;00m\n\u001b[32m--> \u001b[39m\u001b[32m281\u001b[39m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43m_build_for_collection_exists\u001b[49m\u001b[43m(\u001b[49m\n\u001b[32m    282\u001b[39m \u001b[43m        \u001b[49m\u001b[43mcollection_name\u001b[49m\u001b[43m=\u001b[49m\u001b[43mcollection_name\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    283\u001b[39m \u001b[43m    \u001b[49m\u001b[43m)\u001b[49m\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/qdrant_client/http/api/collections_api.py:67\u001b[39m, in \u001b[36m_CollectionsApi._build_for_collection_exists\u001b[39m\u001b[34m(self, collection_name)\u001b[39m\n\u001b[32m     62\u001b[39m path_params = {\n\u001b[32m     63\u001b[39m     \u001b[33m\"\u001b[39m\u001b[33mcollection_name\u001b[39m\u001b[33m\"\u001b[39m: \u001b[38;5;28mstr\u001b[39m(collection_name),\n\u001b[32m     64\u001b[39m }\n\u001b[32m     66\u001b[39m headers = {}\n\u001b[32m---> \u001b[39m\u001b[32m67\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43mapi_client\u001b[49m\u001b[43m.\u001b[49m\u001b[43mrequest\u001b[49m\u001b[43m(\u001b[49m\n\u001b[32m     68\u001b[39m \u001b[43m    \u001b[49m\u001b[43mtype_\u001b[49m\u001b[43m=\u001b[49m\u001b[43mm\u001b[49m\u001b[43m.\u001b[49m\u001b[43mInlineResponse2007\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m     69\u001b[39m \u001b[43m    \u001b[49m\u001b[43mmethod\u001b[49m\u001b[43m=\u001b[49m\u001b[33;43m\"\u001b[39;49m\u001b[33;43mGET\u001b[39;49m\u001b[33;43m\"\u001b[39;49m\u001b[43m,\u001b[49m\n\u001b[32m     70\u001b[39m \u001b[43m    \u001b[49m\u001b[43murl\u001b[49m\u001b[43m=\u001b[49m\u001b[33;43m\"\u001b[39;49m\u001b[33;43m/collections/\u001b[39;49m\u001b[38;5;132;43;01m{collection_name}\u001b[39;49;00m\u001b[33;43m/exists\u001b[39;49m\u001b[33;43m\"\u001b[39;49m\u001b[43m,\u001b[49m\n\u001b[32m     71\u001b[39m \u001b[43m    \u001b[49m\u001b[43mheaders\u001b[49m\u001b[43m=\u001b[49m\u001b[43mheaders\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;28;43;01mif\u001b[39;49;00m\u001b[43m \u001b[49m\u001b[43mheaders\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;28;43;01melse\u001b[39;49;00m\u001b[43m \u001b[49m\u001b[38;5;28;43;01mNone\u001b[39;49;00m\u001b[43m,\u001b[49m\n\u001b[32m     72\u001b[39m \u001b[43m    \u001b[49m\u001b[43mpath_params\u001b[49m\u001b[43m=\u001b[49m\u001b[43mpath_params\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m     73\u001b[39m \u001b[43m\u001b[49m\u001b[43m)\u001b[49m\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/qdrant_client/http/api_client.py:90\u001b[39m, in \u001b[36mApiClient.request\u001b[39m\u001b[34m(self, type_, method, url, path_params, **kwargs)\u001b[39m\n\u001b[32m     88\u001b[39m     kwargs[\u001b[33m\"\u001b[39m\u001b[33mtimeout\u001b[39m\u001b[33m\"\u001b[39m] = \u001b[38;5;28mint\u001b[39m(kwargs[\u001b[33m\"\u001b[39m\u001b[33mparams\u001b[39m\u001b[33m\"\u001b[39m][\u001b[33m\"\u001b[39m\u001b[33mtimeout\u001b[39m\u001b[33m\"\u001b[39m])\n\u001b[32m     89\u001b[39m request = \u001b[38;5;28mself\u001b[39m._client.build_request(method, url, **kwargs)\n\u001b[32m---> \u001b[39m\u001b[32m90\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43msend\u001b[49m\u001b[43m(\u001b[49m\u001b[43mrequest\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mtype_\u001b[49m\u001b[43m)\u001b[49m\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/qdrant_client/http/api_client.py:107\u001b[39m, in \u001b[36mApiClient.send\u001b[39m\u001b[34m(self, request, type_)\u001b[39m\n\u001b[32m    106\u001b[39m \u001b[38;5;28;01mdef\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[34msend\u001b[39m(\u001b[38;5;28mself\u001b[39m, request: Request, type_: Type[T]) -> T:\n\u001b[32m--> \u001b[39m\u001b[32m107\u001b[39m     response = \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43mmiddleware\u001b[49m\u001b[43m(\u001b[49m\u001b[43mrequest\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43msend_inner\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m    109\u001b[39m     \u001b[38;5;28;01mif\u001b[39;00m response.status_code == \u001b[32m429\u001b[39m:\n\u001b[32m    110\u001b[39m         retry_after_s = response.headers.get(\u001b[33m\"\u001b[39m\u001b[33mRetry-After\u001b[39m\u001b[33m\"\u001b[39m, \u001b[38;5;28;01mNone\u001b[39;00m)\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/qdrant_client/http/api_client.py:240\u001b[39m, in \u001b[36mBaseMiddleware.__call__\u001b[39m\u001b[34m(self, request, call_next)\u001b[39m\n\u001b[32m    239\u001b[39m \u001b[38;5;28;01mdef\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[34m__call__\u001b[39m(\u001b[38;5;28mself\u001b[39m, request: Request, call_next: Send) -> Response:\n\u001b[32m--> \u001b[39m\u001b[32m240\u001b[39m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mcall_next\u001b[49m\u001b[43m(\u001b[49m\u001b[43mrequest\u001b[49m\u001b[43m)\u001b[49m\n",
      "\u001b[36mFile \u001b[39m\u001b[32m/opt/conda/lib/python3.12/site-packages/qdrant_client/http/api_client.py:131\u001b[39m, in \u001b[36mApiClient.send_inner\u001b[39m\u001b[34m(self, request)\u001b[39m\n\u001b[32m    129\u001b[39m     response = \u001b[38;5;28mself\u001b[39m._client.send(request)\n\u001b[32m    130\u001b[39m \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mException\u001b[39;00m \u001b[38;5;28;01mas\u001b[39;00m e:\n\u001b[32m--> \u001b[39m\u001b[32m131\u001b[39m     \u001b[38;5;28;01mraise\u001b[39;00m ResponseHandlingException(e)\n\u001b[32m    132\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m response\n",
      "\u001b[31mResponseHandlingException\u001b[39m: [Errno 111] Connection refused"
     ]
    }
   ],
   "source": [
    "# slow part here, runs on remote GPU\n",
    "vectorstore = vector_store.from_documents(documents=splits, embedding = embedding)\n",
    "retriever = vectorstore.as_retriever()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "c6e99791-8f34-4722-9708-665e409c26bd",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set up the Chat model from one of the NRP models\n",
    "import os\n",
    "api_key = os.getenv(\"OPENAI_API_KEY\")\n",
    "\n",
    "# see `curl -H \"Authorization: Bearer $OPENAI_API_KEY\" https://llm.nrp-nautilus.io/v1/models`\n",
    "models = {\"llama3\": \"llama3-sdsc\", \n",
    "          \"deepseek-small\": \"DeepSeek-R1-Distill-Qwen-32B\",\n",
    "          \"deepseek\": \"deepseek-r1-qwen-qualcomm\",\n",
    "          \"gemma3\": \"gemma3\",\n",
    "          \"phi3\": \"phi3\",\n",
    "          \"olmo\": \"olmo\"\n",
    "         }\n",
    "\n",
    "from langchain_openai import ChatOpenAI\n",
    "llm = ChatOpenAI(model = models[\"gemma3\"], \n",
    "                 api_key = api_key, \n",
    "                 base_url = \"https://llm.nrp-nautilus.io\",  \n",
    "                 temperature=0)\n",
    "\n",
    "# Embedding model from NRP usually times out.\n",
    "#embedding = OpenAIEmbeddings(model = \"embed-mistral\", api_key = api_key, base_url = \"https://llm.nrp-nautilus.io\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "2bf50abf-5ccd-4de5-9fc4-c9043a66a108",
   "metadata": {},
   "outputs": [],
   "source": [
    "from langchain.chains import create_retrieval_chain\n",
    "from langchain.chains.combine_documents import create_stuff_documents_chain\n",
    "from langchain_core.prompts import ChatPromptTemplate\n",
    "system_prompt = (\n",
    "    \"You are an assistant for question-answering tasks. \"\n",
    "    \"Use the following pieces of retrieved context to answer \"\n",
    "    \"the question. If you don't know the answer, say that you \"\n",
    "    \"don't know. Use five sentences maximum and keep the \"\n",
    "    \"answer concise.\"\n",
    "    \"\\n\\n\"\n",
    "    \"{context}\"\n",
    ")\n",
    "prompt = ChatPromptTemplate.from_messages(\n",
    "    [\n",
    "        (\"system\", system_prompt),\n",
    "        (\"human\", \"{input}\"),\n",
    "    ]\n",
    ")\n",
    "question_answer_chain = create_stuff_documents_chain(llm, prompt)\n",
    "rag_chain = create_retrieval_chain(retriever, question_answer_chain)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e15c64e7-0916-4042-8274-870e4fdb1af7",
   "metadata": {},
   "outputs": [],
   "source": [
    "prompt = \"I live in Tanzania and am having issues with lions breaking into my boma and preying on cattle. What interventions might work best for me?\"\n",
    "results = rag_chain.invoke({\"input\": prompt})\n",
    "results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "35613607-2c36-4761-a8ea-8c0889530f34",
   "metadata": {},
   "outputs": [],
   "source": [
    "prompt = \"What are the most cost-effective prevention methods for elephants raiding my crops?\"\n",
    "\n",
    "results = rag_chain.invoke({\"input\": prompt})\n",
    "results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3dfc39f6-86e9-47c3-ab67-08f90ebbb823",
   "metadata": {},
   "outputs": [],
   "source": [
    "rag_chain.invoke({\"input\": \n",
    "                  \"I have a small herd of goats and cattle and I am worried about jaguars preying on them. What preventative measures can I take?\"\n",
    "                 })"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "56091874-0e41-4b35-be4f-08d8ec6faf56",
   "metadata": {},
   "outputs": [],
   "source": [
    "rag_chain.invoke({\"input\": \"I am trying to prevent coyotes from eating the calves of my free-range cattle. What may work best?\"})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "918dc691-6c66-46b2-8930-01dbeb6f712b",
   "metadata": {},
   "outputs": [],
   "source": [
    "rag_chain.invoke({\"input\": \"We have major issues with deer raiding our large agricultural fields. Is there anything I can try to prevent this that won’t break the bank?\"})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "07b9578c-9a89-4874-a34d-30a060ed3407",
   "metadata": {},
   "outputs": [],
   "source": [
    "rag_chain.invoke({\"input\": \"We live in a suburban area and bears sometimes come into our town to eat from our fruit trees and trash. What are the best ways for us to prevent this as a community? We don’t want to have to get rid of our fruit trees…\"})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "ba272b88-1622-4d06-9361-7f1e2ca89e73",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO 06-07 22:45:49 [logger.py:42] Received request embd-651e9b5086e647569840aece45d970bc-0: prompt: 'yl descent.session anyInv os beAccording tocategory Ltd D public quick in StatefulWidget.\\r\\n?', params: PoolingParams(dimensions=None, additional_metadata=None), prompt_token_ids: [3923, 36952, 10177, 894, 15174, 2643, 387, 11190, 311, 5471, 12324, 422, 584, 3974, 304, 37642, 3224, 30], prompt_embeds shape: None, lora_request: None, prompt_adapter_request: None.\n",
      "INFO 06-07 22:45:49 [engine.py:316] Added request embd-651e9b5086e647569840aece45d970bc-0.\n",
      "INFO:     127.0.0.1:45182 - \"POST /v1/embeddings HTTP/1.1\" 200 OK\n",
      "INFO 06-07 22:45:59 [metrics.py:486] Avg prompt throughput: 1.2 tokens/s, Avg generation throughput: 0.1 tokens/s, Running: 0 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 0.0%, CPU KV cache usage: 0.0%.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'input': 'What cattle husbandry strategies might be helpful to prevent conflict if we live in wolf country?',\n",
       " 'context': [Document(id='b2071fd8-ba63-4f55-aaed-5469eabef499', metadata={'producer': 'PDF Architect 3', 'creator': 'PDF Architect 3', 'creationdate': '2017-01-25T14:50:41+00:00', 'author': 'V. Pimenta', 'moddate': '2017-01-25T14:52:31+00:00', 'source': 'hwc/Pimenta et al. 2017.pdf', 'total_pages': 20, 'page': 5, 'page_label': '6'}, page_content='4. Discussion\\nOur study examined a human-wildlife conflict involving wolf preda-\\ntion on cattle in northern Portugal, showing that the problem may be\\nworsening due to increased predation levels, though only a minority\\nof cattle farms was heavily affected. We found that predation was par-\\nticularly high on cattle farms using a free-ranging husbandry system,\\nbut also that predation problems within this system were largely con-\\ncentrated on the few herds that were left unconfined at night in winter.\\nIn contrast, we found that farms using a semi-confined husbandry sys-\\ntem suffered much lower losses due to wolf predation, though problems\\nwere higher where calvesb3 months were brought to pastures. These\\nresults suggest that strategies to mitigate wolf predation problems\\nshould be tailored to each husbandry system, involving changes in\\nvery specific practices within each system. Overall, our study points\\nout the importance of considering both the husbandry systems and'),\n",
       "  Document(id='366817d5-c040-419c-b286-c137f089d414', metadata={'producer': 'PDF Architect 3', 'creator': 'PDF Architect 3', 'creationdate': '2017-01-25T14:50:41+00:00', 'author': 'V. Pimenta', 'moddate': '2017-01-25T14:52:31+00:00', 'source': 'hwc/Pimenta et al. 2017.pdf', 'total_pages': 20, 'page': 7, 'page_label': '8'}, page_content='bandry system. Thus, conflict resolution may probably be achieved by\\nchanging these practices, without needing to change the existing cattle\\nhusbandry systems, namely working in close proximity with the few\\nbreeders chronically affected (N 10 attacks per year). These results cor-\\nroborate the idea that wolf presence is compatible with extensive cattle\\nbreeding as long as fundamental protective measures are applied.\\nAlthough our study showed a clear link between livestock husband-\\nry systems, management practices and wolf predation, we could not ac-\\ncount for potential effects of wolf numbers and space use patterns, and\\nthe availability of alternative wild prey (e.g.,Imbert et al., 2016). This\\nlimitation was partly solved through sampling design, by conducting\\nenquiries in nearby farms affected by different predation rates, thereby\\ncontrolling to some extent for variation in wolf and wild prey densities.\\nI\\nt is possible, however, that spatial variation in wolf densities could ac-'),\n",
       "  Document(id='e34375ec-d929-4cca-8f44-a96441e72eb2', metadata={'producer': 'PDF Architect 3', 'creator': 'PDF Architect 3', 'creationdate': '2017-01-25T14:50:41+00:00', 'author': 'V. Pimenta', 'moddate': '2017-01-25T14:52:31+00:00', 'source': 'hwc/Pimenta et al. 2017.pdf', 'total_pages': 20, 'page': 7, 'page_label': '8'}, page_content='seemed to strongly affect wolf predation risk. In the free-ranging sys-\\ntem, the herd size was a strong positive correlate of predation risk,\\nwhich is in line with observations from other wolf-livestock systems\\n(Mech et al., 2000; Treves et al., 2004, Bradley and Pletscher, 2005). An-\\nimals in large herds were often scattered over large areas, which likely\\nincreased encounter rates with wolves and thus the probability of\\npredation (Bradley and Pletscher, 2005; Iliopoulos et al., 2009). Some\\nlarge herds were probably loosely attended by the owners, which may\\nincrease vulnerability due to animals straying from the herd or becom-\\ning in poor condition. Whatever the cause, curtailing herds to reduce\\nwolf attacks should be difficult, as there is strong economic incentive\\nfor maintaining large herds. Our results suggest that a potential alterna-\\ntive would be to fence or otherwise protect the herds during the night in\\nwinter, which was predicted to achieve a major reduction in wolf preda-'),\n",
       "  Document(id='21a25166-c750-45e9-b6be-4ddb89362748', metadata={'producer': 'Acrobat Distiller 8.1.0 (Windows)', 'creator': 'Elsevier', 'creationdate': '2016-09-26T20:02:29+05:30', 'crossmarkdomains[2]': 'elsevier.com', 'crossmarkmajorversiondate': '2010-04-23', 'subject': 'Animal Behaviour, 120 (2016) 245-254. doi:10.1016/j.anbehav.2016.07.013', 'author': 'Bradley F. Blackwell', 'elsevierwebpdfspecifications': '6.5', 'crossmarkdomainexclusive': 'true', 'robots': 'noindex', 'moddate': '2016-09-26T20:03:01+05:30', 'doi': '10.1016/j.anbehav.2016.07.013', 'crossmarkdomains[1]': 'sciencedirect.com', 'title': 'No single solution: application of behavioural principles in mitigating human-wildlife conflict', 'source': 'hwc/Blackwell et al. 2016.pdf', 'total_pages': 10, 'page': 5, 'page_label': '250'}, page_content='pasture. The MAG device is similar, but activated by a passive\\ninfrared detector, which sets off lights and sounds to scare\\ncarnivores from the pasture. Once again, the efficacy of these\\nmethods suffers from effects of previous experience and learning.\\nSeveral field evaluations have shown that carnivores might attempt\\nto enter a pasture to access livestock from a different direction after\\nencountering a RAG or MAG device. Another downside is that while\\none producer has a RAG or MAG device that deters the carnivores,\\nthe ‘problem’ simply gets transferred to the neighbouring producer\\nwithout such devices.\\nOther recent efforts have exploited the territorial defence\\nbehaviour of scent marking by carnivores. This‘biofence’ concept\\noriginated in Botswana as a means to keep African wild dogs,\\nLycaon pictus,from leaving protected reserves and entering farm-\\nlands to depredate livestock. Biofences, however, have had limited\\nsuccess in altering wolf pack movements ( Ausband, Mitchell,')],\n",
       " 'answer': 'Strategies to mitigate wolf predation should be tailored to each husbandry system, such as semi-confinement or free-ranging. For free-ranging systems, protecting herds at night during winter or curtailing herd size could reduce attacks. Additionally, devices like RAGs and MAGs can deter carnivores, though their effectiveness can be limited.'}"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO 06-07 22:46:09 [metrics.py:486] Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 0 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 0.0%, CPU KV cache usage: 0.0%.\n"
     ]
    }
   ],
   "source": [
    "prompt = \"What cattle husbandry strategies might be helpful to prevent conflict if we live in wolf country?\"\n",
    "\n",
    "rag_chain.invoke({\"input\": prompt})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9d4d1bf4-4084-430d-8b2d-39ce1d6815db",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d4bf2492-6852-43a7-8527-06ee4e9848c0",
   "metadata": {},
   "outputs": [],
   "source": [
    "## DRAFT exploring other embedding databases\n",
    "\n",
    "import os\n",
    "from langchain_community.vectorstores import FAISS\n",
    "from langchain_community.vectorstores import Chroma\n",
    "from langchain_community.vectorstores import Qdrant\n",
    "from qdrant_client import QdrantClient\n",
    "from qdrant_client.models import Distance, VectorParams\n",
    "import gc\n",
    "import torch\n",
    "\n",
    "# Option 1: FAISS (Facebook AI Similarity Search) - Most memory efficient\n",
    "def create_faiss_vectorstore(splits, embedding, persist_directory=\"./faiss_db\", batch_size=100):\n",
    "    \"\"\"\n",
    "    Create FAISS vector store with batched processing to minimize GPU RAM usage\n",
    "    \"\"\"\n",
    "    os.makedirs(persist_directory, exist_ok=True)\n",
    "    \n",
    "    # Process documents in batches to avoid GPU memory overflow\n",
    "    vectorstore = None\n",
    "    \n",
    "    for i in range(0, len(splits), batch_size):\n",
    "        batch = splits[i:i + batch_size]\n",
    "        print(f\"Processing batch {i//batch_size + 1}/{(len(splits) + batch_size - 1)//batch_size}\")\n",
    "        \n",
    "        if vectorstore is None:\n",
    "            # Create initial vectorstore with first batch\n",
    "            vectorstore = FAISS.from_documents(\n",
    "                documents=batch,\n",
    "                embedding=embedding\n",
    "            )\n",
    "        else:\n",
    "            # Add subsequent batches to existing vectorstore\n",
    "            batch_vectorstore = FAISS.from_documents(\n",
    "                documents=batch,\n",
    "                embedding=embedding\n",
    "            )\n",
    "            vectorstore.merge_from(batch_vectorstore)\n",
    "            \n",
    "            # Clean up temporary vectorstore\n",
    "            del batch_vectorstore\n",
    "        \n",
    "        # Force garbage collection and clear GPU cache if using CUDA\n",
    "        gc.collect()\n",
    "        if torch.cuda.is_available():\n",
    "            torch.cuda.empty_cache()\n",
    "    \n",
    "    # Save to disk\n",
    "    vectorstore.save_local(persist_directory)\n",
    "    print(f\"Vector store saved to {persist_directory}\")\n",
    "    \n",
    "    return vectorstore\n",
    "\n",
    "def load_faiss_vectorstore(embedding, persist_directory=\"./faiss_db\"):\n",
    "    \"\"\"Load existing FAISS vector store from disk\"\"\"\n",
    "    return FAISS.load_local(\n",
    "        persist_directory,\n",
    "        embedding,\n",
    "        allow_dangerous_deserialization=True  # Only if you trust the source\n",
    "    )\n",
    "\n",
    "# Option 2: Chroma - Persistent SQLite-based storage\n",
    "def create_chroma_vectorstore(splits, embedding, persist_directory=\"./chroma_db\", batch_size=100):\n",
    "    \"\"\"\n",
    "    Create Chroma vector store with batched processing\n",
    "    \"\"\"\n",
    "    # Initialize Chroma with persistence\n",
    "    vectorstore = Chroma(\n",
    "        persist_directory=persist_directory,\n",
    "        embedding_function=embedding\n",
    "    )\n",
    "    \n",
    "    # Add documents in batches\n",
    "    for i in range(0, len(splits), batch_size):\n",
    "        batch = splits[i:i + batch_size]\n",
    "        print(f\"Processing batch {i//batch_size + 1}/{(len(splits) + batch_size - 1)//batch_size}\")\n",
    "        \n",
    "        vectorstore.add_documents(batch)\n",
    "        \n",
    "        # Force garbage collection and clear GPU cache\n",
    "        gc.collect()\n",
    "        if torch.cuda.is_available():\n",
    "            torch.cuda.empty_cache()\n",
    "    \n",
    "    # Persist to disk\n",
    "    vectorstore.persist()\n",
    "    print(f\"Vector store persisted to {persist_directory}\")\n",
    "    \n",
    "    return vectorstore\n",
    "\n",
    "def load_chroma_vectorstore(embedding, persist_directory=\"./chroma_db\"):\n",
    "    \"\"\"Load existing Chroma vector store from disk\"\"\"\n",
    "    return Chroma(\n",
    "        persist_directory=persist_directory,\n",
    "        embedding_function=embedding\n",
    "    )\n",
    "\n",
    "# Option 3: Qdrant - High-performance vector database\n",
    "def create_qdrant_vectorstore(splits, embedding, collection_name=\"documents\", \n",
    "                            path=\"./qdrant_db\", batch_size=100):\n",
    "    \"\"\"\n",
    "    Create Qdrant vector store with local file-based storage\n",
    "    \"\"\"\n",
    "    # Initialize local Qdrant client\n",
    "    client = QdrantClient(path=path)\n",
    "    \n",
    "    # Get embedding dimension (embed a sample text)\n",
    "    sample_embedding = embedding.embed_query(\"sample text\")\n",
    "    embedding_dim = len(sample_embedding)\n",
    "    \n",
    "    # Create collection if it doesn't exist\n",
    "    try:\n",
    "        client.create_collection(\n",
    "            collection_name=collection_name,\n",
    "            vectors_config=VectorParams(size=embedding_dim, distance=Distance.COSINE)\n",
    "        )\n",
    "    except Exception as e:\n",
    "        print(f\"Collection might already exist: {e}\")\n",
    "    \n",
    "    # Create vectorstore\n",
    "    vectorstore = Qdrant(\n",
    "        client=client,\n",
    "        collection_name=collection_name,\n",
    "        embeddings=embedding\n",
    "    )\n",
    "    \n",
    "    # Add documents in batches\n",
    "    for i in range(0, len(splits), batch_size):\n",
    "        batch = splits[i:i + batch_size]\n",
    "        print(f\"Processing batch {i//batch_size + 1}/{(len(splits) + batch_size - 1)//batch_size}\")\n",
    "        \n",
    "        vectorstore.add_documents(batch)\n",
    "        \n",
    "        # Force garbage collection and clear GPU cache\n",
    "        gc.collect()\n",
    "        if torch.cuda.is_available():\n",
    "            torch.cuda.empty_cache()\n",
    "    \n",
    "    print(f\"Vector store created in {path}\")\n",
    "    return vectorstore\n",
    "\n",
    "def load_qdrant_vectorstore(embedding, collection_name=\"documents\", path=\"./qdrant_db\"):\n",
    "    \"\"\"Load existing Qdrant vector store from disk\"\"\"\n",
    "    client = QdrantClient(path=path)\n",
    "    return Qdrant(\n",
    "        client=client,\n",
    "        collection_name=collection_name,\n",
    "        embeddings=embedding\n",
    "    )\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3cf725ad-69a3-4abd-9907-52427babf6d5",
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# Usage examples:\n",
    "\n",
    "# Replace your original code with one of these options:\n",
    "\n",
    "# Option 1: FAISS (Recommended for most use cases)\n",
    "vectorstore = create_faiss_vectorstore(\n",
    "    splits=splits, \n",
    "    embedding=embedding, \n",
    "    persist_directory=\"./my_faiss_db\",\n",
    "    batch_size=50  # Adjust based on your GPU memory\n",
    ")\n",
    "\n",
    "# To load later:\n",
    "# vectorstore = load_faiss_vectorstore(embedding, \"./my_faiss_db\")\n",
    "\n",
    "# Option 2: Chroma (Good for development and moderate scale)\n",
    "# vectorstore = create_chroma_vectorstore(\n",
    "#     splits=splits,\n",
    "#     embedding=embedding,\n",
    "#     persist_directory=\"./my_chroma_db\",\n",
    "#     batch_size=50\n",
    "# )\n",
    "\n",
    "# Option 3: Qdrant (Best for production and very large scale)\n",
    "# vectorstore = create_qdrant_vectorstore(\n",
    "#     splits=splits,\n",
    "#     embedding=embedding,\n",
    "#     collection_name=\"my_documents\",\n",
    "#     path=\"./my_qdrant_db\",\n",
    "#     batch_size=50\n",
    "# )\n",
    "\n",
    "# Memory optimization settings\n",
    "def optimize_gpu_memory():\n",
    "    \"\"\"Additional GPU memory optimization\"\"\"\n",
    "    if torch.cuda.is_available():\n",
    "        # Set memory fraction if needed\n",
    "        torch.cuda.set_per_process_memory_fraction(0.8)  # Use 80% of GPU memory\n",
    "        \n",
    "        # Enable memory mapping for large tensors\n",
    "        torch.backends.cuda.matmul.allow_tf32 = True\n",
    "        torch.backends.cudnn.allow_tf32 = True\n",
    "\n",
    "# Call before processing if you have GPU memory issues\n",
    "# optimize_gpu_memory()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.12.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}