File size: 31,385 Bytes
1a08523
bd9fae2
e375940
 
 
94a67ea
8cd1f1e
94a67ea
 
 
bd9fae2
c5e4524
9c49e99
bd9fae2
6a79fd2
 
c5e4524
5482130
c5e4524
 
244a3e0
 
c5e4524
a7fc504
bd9fae2
c5e4524
 
1532bd6
bd9fae2
 
 
c5e4524
 
 
 
 
1a08523
8b61059
c5e4524
 
6627aee
 
0175cb6
6627aee
0175cb6
6627aee
 
1a08523
 
4b19adc
1a08523
 
9c49e99
c5e4524
 
 
f9da573
e375940
f9da573
 
1a08523
bd9fae2
f9da573
c5e4524
 
 
 
 
 
8cd1f1e
ff5dcc7
a7b0635
9975133
b19bb41
 
 
8cd1f1e
9975133
 
9c49e99
0175cb6
bd9fae2
 
0f10e25
1a08523
 
 
4b19adc
 
 
 
 
0175cb6
5ef1f60
bd9fae2
 
 
244a3e0
0175cb6
9c49e99
a7fc504
 
9975133
 
1a08523
 
 
 
 
6627aee
4b19adc
 
 
 
 
 
 
 
 
 
c5f41e6
a7fc504
bd1cd80
 
 
 
 
 
 
 
 
 
a7fc504
 
 
1a08523
 
c5f41e6
 
 
 
 
 
 
 
 
 
 
 
 
1a08523
 
 
 
 
 
 
 
6627aee
1a08523
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4b19adc
 
 
 
 
 
 
 
 
 
 
1a08523
4b19adc
 
 
 
1a08523
4b19adc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1a08523
4b19adc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1a08523
4b19adc
 
 
 
 
 
 
 
 
 
 
1a08523
 
 
 
 
c5f41e6
9975133
 
 
1a08523
 
 
 
 
 
4b19adc
1a08523
8cd1f1e
 
 
 
1532bd6
 
 
 
 
 
 
9975133
 
8cd1f1e
 
 
 
1a08523
 
 
 
 
9975133
 
8cd1f1e
 
 
 
e375940
 
 
8cd1f1e
 
 
 
 
 
e375940
 
 
8cd1f1e
 
 
 
aa91fc5
 
 
1532bd6
 
aa91fc5
 
 
bd9fae2
1532bd6
 
 
 
 
 
 
 
 
 
 
 
bd9fae2
1532bd6
 
 
 
 
 
 
aa91fc5
e375940
 
 
 
 
 
 
 
 
 
 
 
8cd1f1e
9975133
1a08523
 
 
 
 
 
 
 
 
 
 
 
4b19adc
1a08523
 
 
 
 
 
 
 
9975133
e514fa8
5ef1f60
1a08523
1532bd6
 
 
 
 
 
 
 
 
1a08523
 
 
 
18468cb
1a08523
bd9fae2
1a08523
 
 
 
 
 
 
 
 
a7fc504
bd9fae2
1a08523
 
e375940
1a08523
bb343a5
 
 
 
 
 
 
 
1a08523
 
 
 
 
 
 
 
176e1ef
bd9fae2
1a08523
 
e375940
1a08523
 
 
 
e514fa8
 
1a08523
4b19adc
 
1532bd6
 
 
 
 
 
 
 
 
 
 
 
4b19adc
 
 
 
18468cb
4b19adc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
176e1ef
bd9fae2
4b19adc
 
 
 
 
 
1a08523
4b19adc
bb343a5
 
 
 
 
 
 
 
4b19adc
 
 
 
 
 
 
 
 
 
 
 
 
 
176e1ef
bd9fae2
4b19adc
 
 
 
 
 
1a08523
4b19adc
 
1a08523
1532bd6
 
 
 
 
 
 
 
 
 
 
 
4b19adc
 
 
 
18468cb
4b19adc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
176e1ef
bd9fae2
4b19adc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
176e1ef
bd9fae2
4b19adc
 
 
 
 
 
 
 
1a08523
4b19adc
bb343a5
 
 
 
 
 
 
 
4b19adc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
176e1ef
bd9fae2
4b19adc
 
 
 
 
 
 
 
8cd1f1e
4b19adc
 
 
 
 
 
 
 
 
 
 
176e1ef
bd9fae2
4b19adc
 
 
 
 
 
 
 
1a08523
4b19adc
 
 
 
 
 
1a08523
 
 
 
 
4b19adc
 
 
 
 
 
 
 
8cd1f1e
9975133
 
e375940
816093e
e375940
9975133
 
 
 
 
 
 
 
 
 
1a08523
9c49e99
1a08523
 
 
 
 
 
 
 
 
 
 
 
9c49e99
8cd1f1e
 
0175cb6
 
 
8cd1f1e
 
9975133
9c49e99
 
816093e
9c49e99
 
 
 
 
 
 
 
 
 
 
8cd1f1e
 
6627aee
8cd1f1e
9975133
6627aee
0175cb6
 
6627aee
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9c49e99
 
816093e
9c49e99
 
 
6627aee
0175cb6
 
 
6627aee
 
 
 
0175cb6
 
 
6627aee
0175cb6
 
 
 
6627aee
 
 
 
 
 
0175cb6
 
 
 
 
6627aee
0175cb6
 
 
 
6627aee
 
 
 
 
 
 
 
 
 
 
 
 
 
 
816093e
6627aee
 
 
 
 
9c49e99
bd9fae2
8cd1f1e
f9da573
1a08523
 
 
 
 
 
 
 
 
 
 
 
 
4b19adc
 
 
 
 
1a08523
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4b19adc
 
 
 
1532bd6
 
 
4b19adc
1532bd6
 
 
 
4b19adc
 
 
1532bd6
 
 
4b19adc
1532bd6
 
 
4b19adc
1532bd6
 
 
 
4b19adc
 
1532bd6
 
 
4b19adc
1532bd6
 
 
4b19adc
1532bd6
 
 
 
4b19adc
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
import re
import numpy as np
import openai
import streamlit_scrollable_textbox as stx

import pinecone
import streamlit as st

st.set_page_config(layout="wide")  # isort: split

from utils import nltkmodules
from utils.entity_extraction import (
    clean_entities,
    extract_keywords,
    extract_quarter_year,
    extract_ticker_spacy,
    format_entities_flan_alpaca,
    generate_alpaca_ner_prompt,
)
from utils.models import (
    generate_entities_flan_alpaca_checkpoint,
    generate_entities_flan_alpaca_inference_api,
    generate_text_flan_t5,
    get_alpaca_model,
    get_data,
    get_flan_alpaca_xl_model,
    get_flan_t5_model,
    get_instructor_embedding_model,
    get_instructor_embedding_model_api,
    get_bm25_model,
    preprocess_text,
    get_mpnet_embedding_model,
    get_sgpt_embedding_model,
    get_spacy_model,
    get_splade_sparse_embedding_model,
    get_t5_model,
    gpt_turbo_model,
    save_key,
)
from utils.prompts import (
    generate_flant5_prompt_instruct_chunk_context,
    generate_flant5_prompt_instruct_chunk_context_single,
    generate_flant5_prompt_instruct_complete_context,
    generate_flant5_prompt_summ_chunk_context,
    generate_flant5_prompt_summ_chunk_context_single,
    generate_gpt_j_two_shot_prompt_1,
    generate_gpt_j_two_shot_prompt_2,
    generate_gpt_prompt_alpaca,
    generate_gpt_prompt_alpaca_multi_doc,
    generate_gpt_prompt_alpaca_multi_doc_multi_company,
    generate_gpt_prompt_original,
    generate_multi_doc_context,
    get_context_list_prompt,
)
from utils.retriever import (
    format_query,
    query_pinecone,
    query_pinecone_sparse,
    sentence_id_combine,
    text_lookup,
    year_quarter_range,
    get_bm25_search_hits,
)
from utils.transcript_retrieval import retrieve_transcript
from utils.vector_index import (
    create_dense_embeddings,
    create_sparse_embeddings,
    hybrid_score_norm,
)

st.title("Question Answering on Earnings Call Transcripts")


st.write(
    "The app uses the quarterly earnings call transcripts for 10 companies (Apple, AMD, Amazon, Cisco, Google, Microsoft, Nvidia, ASML, Intel, Micron) for the years 2016 to 2020."
)

col1, col2 = st.columns([3, 3], gap="medium")


with st.sidebar:
    use_bm25 = st.checkbox("Use BM25 for filtering results")

    ner_choice = st.selectbox("Select NER Model", ["Spacy", "Alpaca"])
    document_type = st.selectbox(
        "Select Query Type", ["Single-Document", "Multi-Document"]
    )
    if document_type == "Multi-Document":
        multi_company_choice = st.selectbox(
            "Select Company Query Type",
            ["Single-Company", "Compare Companies"],
        )

data = get_data()
corpus, bm25 = get_bm25_model(data)


if ner_choice == "Spacy":
    ner_model = get_spacy_model()

alpaca_model = get_alpaca_model()

with col1:
    st.subheader("Question")
    if document_type == "Single-Document":
        query_text = st.text_area(
            "Input Query",
            value="What was discussed regarding Wearables revenue performance?",
        )
    else:
        if multi_company_choice == "Single-Company":
            query_text = st.text_area(
                "Input Query",
                value="What was the reported revenue for Wearables over the last 2 years?",
            )
        else:
            query_text = st.text_area(
                "Input Query",
                value="How was AAPL's capex spend compared to GOOGL?",
            )


tokenized_query = preprocess_text(query_text).split()
sparse_scores = np.argsort(bm25.get_scores(tokenized_query), axis=0)[::-1]
indices_hits = get_bm25_search_hits(corpus, sparse_scores, 50)

if use_bm25 == True:
    indices = indices_hits
else:
    indices = None


# Extract keywords from query
keywords = extract_keywords(query_text, alpaca_model)

years_choice = ["2020", "2019", "2018", "2017", "2016", "All"]
quarters_choice = ["Q1", "Q2", "Q3", "Q4", "All"]
ticker_choice = [
    "AAPL",
    "CSCO",
    "MSFT",
    "ASML",
    "NVDA",
    "GOOGL",
    "MU",
    "INTC",
    "AMZN",
    "AMD",
]


if document_type == "Single-Document":
    if ner_choice == "Alpaca":
        ner_prompt = generate_alpaca_ner_prompt(query_text)
        entity_text = generate_entities_flan_alpaca_inference_api(ner_prompt)
        company_ent, quarter_ent, year_ent = format_entities_flan_alpaca(
            entity_text
        )
    else:
        company_ent = extract_ticker_spacy(query_text, ner_model)
        quarter_ent, year_ent = extract_quarter_year(query_text)

    ticker_index, quarter_index, year_index = clean_entities(
        company_ent, quarter_ent, year_ent
    )

    with col1:
        # Hardcoding the defaults for a question without metadata
        if (
            query_text
            == "What was discussed regarding Wearables revenue performance?"
        ):
            year = st.selectbox("Year", years_choice)
            quarter = st.selectbox("Quarter", quarters_choice)
            ticker = st.selectbox("Company", ticker_choice)
        else:
            year = st.selectbox("Year", years_choice, index=year_index)
            quarter = st.selectbox(
                "Quarter", quarters_choice, index=quarter_index
            )
            ticker = st.selectbox("Company", ticker_choice, ticker_index)

        participant_type = st.selectbox(
            "Speaker", ["Company Speaker", "Analyst"]
        )

else:
    # Multi-Document Case
    with col1:
        # Single Company Summary
        if multi_company_choice == "Single-Company":
            # Hardcoding the defaults for a question without metadata
            if (
                query_text
                == "What was the reported revenue for Wearables over the last 2 years?"
            ):
                start_year = st.selectbox("Start Year", years_choice, index=2)
                start_quarter = st.selectbox(
                    "Start Quarter", quarters_choice, index=0
                )

                end_year = st.selectbox("End Year", years_choice, index=0)
                end_quarter = st.selectbox(
                    "End Quarter", quarters_choice, index=0
                )

                ticker = st.selectbox("Company", ticker_choice, index=0)
            else:
                start_year = st.selectbox("Start Year", years_choice, index=2)
                start_quarter = st.selectbox(
                    "Start Quarter", quarters_choice, index=0
                )

                end_year = st.selectbox("End Year", years_choice, index=0)
                end_quarter = st.selectbox(
                    "End Quarter", quarters_choice, index=0
                )

                ticker = st.selectbox("Company", ticker_choice, index=0)

        # Single Company Summary
        if multi_company_choice == "Compare Companies":
            # Hardcoding the defaults for a question without metadata
            if query_text == "How was AAPL's capex spend compared to GOOGL?":
                start_year = st.selectbox("Start Year", years_choice, index=1)
                start_quarter = st.selectbox(
                    "Start Quarter", quarters_choice, index=0
                )

                end_year = st.selectbox("End Year", years_choice, index=0)
                end_quarter = st.selectbox(
                    "End Quarter", quarters_choice, index=0
                )

                ticker_first = st.selectbox(
                    "First Company", ticker_choice, index=0
                )
                ticker_second = st.selectbox(
                    "Second Company", ticker_choice, index=5
                )

            else:
                start_year = st.selectbox("Start Year", years_choice, index=2)
                start_quarter = st.selectbox(
                    "Start Quarter", quarters_choice, index=0
                )

                end_year = st.selectbox("End Year", years_choice, index=0)
                end_quarter = st.selectbox(
                    "End Quarter", quarters_choice, index=0
                )

                ticker_first = st.selectbox(
                    "First Company", ticker_choice, index=0
                )
                ticker_second = st.selectbox(
                    "Second Company", ticker_choice, index=1
                )

        participant_type = st.selectbox(
            "Speaker", ["Company Speaker", "Analyst"]
        )


with st.sidebar:
    st.subheader("Select Options:")

    if document_type == "Single-Document":
        num_results = int(
            st.number_input("Number of Results to query", 1, 15, value=5)
        )
    else:
        num_results = int(
            st.number_input("Number of Results to query", 1, 15, value=4)
        )


# Choose encoder model

encoder_models_choice = [
    "MPNET",
    "Instructor",
    "Hybrid Instructor - SPLADE",
    "SGPT",
    "Hybrid MPNET - SPLADE",
]
with st.sidebar:
    encoder_model = st.selectbox("Select Encoder Model", encoder_models_choice)


# Choose decoder model

# Restricting multi-document to only GPT-3
if document_type == "Single-Document":
    decoder_models_choice = ["GPT-3.5 Turbo", "T5", "FLAN-T5", "GPT-J"]
else:
    decoder_models_choice = ["GPT-3.5 Turbo"]
with st.sidebar:
    decoder_model = st.selectbox("Select Decoder Model", decoder_models_choice)


if encoder_model == "MPNET":
    # Connect to pinecone environment
    pinecone.init(
        api_key=st.secrets["pinecone_mpnet"], environment="us-east1-gcp"
    )
    pinecone_index_name = "week2-all-mpnet-base"
    pinecone_index = pinecone.Index(pinecone_index_name)
    retriever_model = get_mpnet_embedding_model()

elif encoder_model == "SGPT":
    # Connect to pinecone environment
    pinecone.init(
        api_key=st.secrets["pinecone_sgpt"], environment="us-east1-gcp"
    )
    pinecone_index_name = "week2-sgpt-125m"
    pinecone_index = pinecone.Index(pinecone_index_name)
    retriever_model = get_sgpt_embedding_model()

elif encoder_model == "Instructor":
    # Connect to pinecone environment
    pinecone.init(
        api_key=st.secrets["pinecone_instructor"],
        environment="us-west4-gcp-free",
    )
    pinecone_index_name = "week13-instructor-xl"
    pinecone_index = pinecone.Index(pinecone_index_name)
    retriever_model = get_instructor_embedding_model_api()
    instruction = (
        "Represent the financial question for retrieving supporting documents:"
    )

elif encoder_model == "Hybrid Instructor - SPLADE":
    # Connect to pinecone environment
    pinecone.init(
        api_key=st.secrets["pinecone_instructor_splade"],
        environment="us-west4-gcp-free",
    )
    pinecone_index_name = "week13-splade-instructor-xl"
    pinecone_index = pinecone.Index(pinecone_index_name)
    retriever_model = get_instructor_embedding_model_api()
    (
        sparse_retriever_model,
        sparse_retriever_tokenizer,
    ) = get_splade_sparse_embedding_model()
    instruction = (
        "Represent the financial question for retrieving supporting documents:"
    )

elif encoder_model == "Hybrid MPNET - SPLADE":
    pinecone.init(
        api_key=st.secrets["pinecone_hybrid_splade_mpnet"],
        environment="us-central1-gcp",
    )
    pinecone_index_name = "splade-mpnet"
    pinecone_index = pinecone.Index(pinecone_index_name)
    retriever_model = get_mpnet_embedding_model()
    (
        sparse_retriever_model,
        sparse_retriever_tokenizer,
    ) = get_splade_sparse_embedding_model()

with st.sidebar:
    if document_type == "Single-Document":
        window = int(st.number_input("Sentence Window Size", 0, 10, value=1))

        threshold = float(
            st.number_input(
                label="Similarity Score Threshold",
                step=0.05,
                format="%.2f",
                value=0.25,
            )
        )
    else:
        window = int(st.number_input("Sentence Window Size", 0, 10, value=1))

        threshold = float(
            st.number_input(
                label="Similarity Score Threshold",
                step=0.05,
                format="%.2f",
                value=0.6,
            )
        )


if document_type == "Single-Document":
    if encoder_model in ["Hybrid SGPT - SPLADE", "Hybrid Instructor - SPLADE"]:
        if encoder_model == "Hybrid Instructor - SPLADE":
            dense_query_embedding = create_dense_embeddings(
                query_text, retriever_model, instruction
            )
        else:
            dense_query_embedding = create_dense_embeddings(
                query_text, retriever_model
            )
        sparse_query_embedding = create_sparse_embeddings(
            query_text, sparse_retriever_model, sparse_retriever_tokenizer
        )
        dense_query_embedding, sparse_query_embedding = hybrid_score_norm(
            dense_query_embedding, sparse_query_embedding, 0.3
        )

        query_results = query_pinecone_sparse(
            dense_query_embedding,
            sparse_query_embedding,
            num_results,
            pinecone_index,
            year,
            quarter,
            ticker,
            participant_type,
            keywords,
            indices,
            threshold,
        )

    else:
        if encoder_model == "Instructor":
            dense_query_embedding = create_dense_embeddings(
                query_text, retriever_model, instruction
            )
        else:
            dense_query_embedding = create_dense_embeddings(
                query_text, retriever_model
            )
        query_results = query_pinecone(
            dense_query_embedding,
            num_results,
            pinecone_index,
            year,
            quarter,
            ticker,
            participant_type,
            keywords,
            indices,
            threshold,
        )

    if threshold <= 0.90:
        context_list = sentence_id_combine(data, query_results, lag=window)
    else:
        context_list = format_query(query_results)

else:
    # Multi-Document Retreival
    # Single Company
    if multi_company_choice == "Single-Company":
        if encoder_model in [
            "Hybrid SGPT - SPLADE",
            "Hybrid Instructor - SPLADE",
        ]:
            if encoder_model == "Hybrid Instructor - SPLADE":
                dense_query_embedding = create_dense_embeddings(
                    query_text, retriever_model, instruction
                )
            else:
                dense_query_embedding = create_dense_embeddings(
                    query_text, retriever_model
                )
            sparse_query_embedding = create_sparse_embeddings(
                query_text, sparse_retriever_model, sparse_retriever_tokenizer
            )
            dense_query_embedding, sparse_query_embedding = hybrid_score_norm(
                dense_query_embedding, sparse_query_embedding, 0.3
            )
            year_quarter_list = year_quarter_range(
                start_quarter, start_year, end_quarter, end_year
            )

            context_group = []
            for year, quarter in year_quarter_list:
                query_results = query_pinecone_sparse(
                    dense_query_embedding,
                    sparse_query_embedding,
                    num_results,
                    pinecone_index,
                    year,
                    quarter,
                    ticker,
                    participant_type,
                    keywords,
                    indices,
                    threshold,
                )
                results_list = sentence_id_combine(
                    data, query_results, lag=window
                )
                context_group.append((results_list, year, quarter, ticker))

        else:
            if encoder_model == "Instructor":
                dense_query_embedding = create_dense_embeddings(
                    query_text, retriever_model, instruction
                )
            else:
                dense_query_embedding = create_dense_embeddings(
                    query_text, retriever_model
                )
            year_quarter_list = year_quarter_range(
                start_quarter, start_year, end_quarter, end_year
            )

            context_group = []
            for year, quarter in year_quarter_list:
                query_results = query_pinecone(
                    dense_query_embedding,
                    num_results,
                    pinecone_index,
                    year,
                    quarter,
                    ticker,
                    participant_type,
                    keywords,
                    indices,
                    threshold,
                )
                results_list = sentence_id_combine(
                    data, query_results, lag=window
                )
                context_group.append((results_list, year, quarter, ticker))

        multi_doc_context = generate_multi_doc_context(context_group)
    # Companies Comparison
    else:
        if encoder_model in [
            "Hybrid SGPT - SPLADE",
            "Hybrid Instructor - SPLADE",
        ]:
            if encoder_model == "Hybrid Instructor - SPLADE":
                dense_query_embedding = create_dense_embeddings(
                    query_text, retriever_model, instruction
                )
            else:
                dense_query_embedding = create_dense_embeddings(
                    query_text, retriever_model
                )
            sparse_query_embedding = create_sparse_embeddings(
                query_text, sparse_retriever_model, sparse_retriever_tokenizer
            )
            dense_query_embedding, sparse_query_embedding = hybrid_score_norm(
                dense_query_embedding, sparse_query_embedding, 0.3
            )
            year_quarter_list = year_quarter_range(
                start_quarter, start_year, end_quarter, end_year
            )

            # First Company Context
            context_group_first = []
            for year, quarter in year_quarter_list:
                query_results = query_pinecone_sparse(
                    dense_query_embedding,
                    sparse_query_embedding,
                    num_results,
                    pinecone_index,
                    year,
                    quarter,
                    ticker_first,
                    participant_type,
                    keywords,
                    indices,
                    threshold,
                )
                results_list = sentence_id_combine(
                    data, query_results, lag=window
                )
                context_group_first.append(
                    (results_list, year, quarter, ticker_first)
                )

            # Second Company Context
            context_group_second = []
            for year, quarter in year_quarter_list:
                query_results = query_pinecone_sparse(
                    dense_query_embedding,
                    sparse_query_embedding,
                    num_results,
                    pinecone_index,
                    year,
                    quarter,
                    ticker_second,
                    participant_type,
                    keywords,
                    indices,
                    threshold,
                )
                results_list = sentence_id_combine(
                    data, query_results, lag=window
                )
                context_group_second.append(
                    (results_list, year, quarter, ticker_second)
                )

        else:
            if encoder_model == "Instructor":
                dense_query_embedding = create_dense_embeddings(
                    query_text, retriever_model, instruction
                )
            else:
                dense_query_embedding = create_dense_embeddings(
                    query_text, retriever_model
                )
            year_quarter_list = year_quarter_range(
                start_quarter, start_year, end_quarter, end_year
            )

            # First Company Context
            context_group_first = []
            for year, quarter in year_quarter_list:
                query_results = query_pinecone(
                    dense_query_embedding,
                    num_results,
                    pinecone_index,
                    year,
                    quarter,
                    ticker_first,
                    participant_type,
                    keywords,
                    indices,
                    threshold,
                )
                results_list = sentence_id_combine(
                    data, query_results, lag=window
                )
                context_group_first.append(
                    (results_list, year, quarter, ticker_first)
                )

            # Second Company Context
            context_group_second = []
            for year, quarter in year_quarter_list:
                query_results = query_pinecone(
                    dense_query_embedding,
                    num_results,
                    pinecone_index,
                    year,
                    quarter,
                    ticker_second,
                    participant_type,
                    keywords,
                    indices,
                    threshold,
                )
                results_list = sentence_id_combine(
                    data, query_results, lag=window
                )
                context_group_second.append(
                    (results_list, year, quarter, ticker_second)
                )

        multi_doc_context_first = generate_multi_doc_context(
            context_group_first
        )
        multi_doc_context_second = generate_multi_doc_context(
            context_group_second
        )

if decoder_model == "GPT-3.5 Turbo":
    if document_type == "Single-Document":
        prompt = generate_gpt_prompt_alpaca(query_text, context_list)
    else:
        if multi_company_choice == "Single-Company":
            prompt = generate_gpt_prompt_alpaca_multi_doc(
                query_text, context_group
            )
        else:
            prompt = generate_gpt_prompt_alpaca_multi_doc_multi_company(
                query_text, context_group_first, context_group_second
            )

    with col2:
        with st.form("my_form"):
            edited_prompt = st.text_area(
                label="Model Prompt", value=prompt, height=400
            )

            openai_key = st.text_input(
                "Enter OpenAI key",
                value="",
                type="password",
            )
            submitted = st.form_submit_button("Submit")
            if submitted:
                api_key = save_key(openai_key)
                openai.api_key = api_key
                generated_text = gpt_turbo_model(edited_prompt)
                st.subheader("Answer:")
                regex_pattern_sentences = (
                    "(?<!\w\.\w.)(?<![A-Z][a-z]\.)(?<=\.|\?)\s"
                )
                generated_text_list = re.split(
                    regex_pattern_sentences, generated_text
                )
                for answer_text in generated_text_list:
                    answer_text = f"""{answer_text}"""
                    st.write(
                        f"<ul><li><p>{answer_text}</p></li></ul>",
                        unsafe_allow_html=True,
                    )


elif decoder_model == "T5":
    prompt = generate_flant5_prompt_instruct_complete_context(
        query_text, context_list
    )
    t5_pipeline = get_t5_model()
    output_text = []
    with col2:
        with st.form("my_form"):
            edited_prompt = st.text_area(
                label="Model Prompt", value=prompt, height=400
            )
            context_list = get_context_list_prompt(edited_prompt)
            submitted = st.form_submit_button("Submit")
            if submitted:
                for context_text in context_list:
                    output_text.append(
                        t5_pipeline(context_text)[0]["summary_text"]
                    )
                st.subheader("Answer:")
                for text in output_text:
                    st.markdown(f"- {text}")

elif decoder_model == "FLAN-T5":
    flan_t5_model, flan_t5_tokenizer = get_flan_t5_model()
    output_text = []
    with col2:
        prompt_type = st.selectbox(
            "Select prompt type",
            ["Complete Text QA", "Chunkwise QA", "Chunkwise Summarize"],
        )
        if prompt_type == "Complete Text QA":
            prompt = generate_flant5_prompt_instruct_complete_context(
                query_text, context_list
            )
        elif prompt_type == "Chunkwise QA":
            st.write("The following prompt is not editable.")
            prompt = generate_flant5_prompt_instruct_chunk_context(
                query_text, context_list
            )
        elif prompt_type == "Chunkwise Summarize":
            st.write("The following prompt is not editable.")
            prompt = generate_flant5_prompt_summ_chunk_context(
                query_text, context_list
            )
        else:
            prompt = ""
        with st.form("my_form"):
            edited_prompt = st.text_area(
                label="Model Prompt", value=prompt, height=400
            )
            submitted = st.form_submit_button("Submit")
            if submitted:
                if prompt_type == "Complete Text QA":
                    output_text_string = generate_text_flan_t5(
                        flan_t5_model, flan_t5_tokenizer, prompt
                    )
                    st.subheader("Answer:")
                    st.write(output_text_string)
                elif prompt_type == "Chunkwise QA":
                    for context_text in context_list:
                        model_input = generate_flant5_prompt_instruct_chunk_context_single(
                            query_text, context_text
                        )
                        output_text.append(
                            generate_text_flan_t5(
                                flan_t5_model, flan_t5_tokenizer, model_input
                            )
                        )
                    st.subheader("Answer:")
                    for text in output_text:
                        if "(iii)" not in text:
                            st.markdown(f"- {text}")
                elif prompt_type == "Chunkwise Summarize":
                    for context_text in context_list:
                        model_input = (
                            generate_flant5_prompt_summ_chunk_context_single(
                                query_text, context_text
                            )
                        )
                        output_text.append(
                            generate_text_flan_t5(
                                flan_t5_model, flan_t5_tokenizer, model_input
                            )
                        )
                    st.subheader("Answer:")
                    for text in output_text:
                        if "(iii)" not in text:
                            st.markdown(f"- {text}")

if decoder_model == "GPT-J":
    if ticker in ["AAPL", "AMD"]:
        prompt = generate_gpt_j_two_shot_prompt_1(query_text, context_list)
    elif ticker in ["NVDA", "INTC", "AMZN"]:
        prompt = generate_gpt_j_two_shot_prompt_2(query_text, context_list)
    else:
        prompt = generate_gpt_j_two_shot_prompt_1(query_text, context_list)
    with col2:
        with st.form("my_form"):
            edited_prompt = st.text_area(
                label="Model Prompt", value=prompt, height=400
            )
            st.write(
                "The app currently just shows the prompt. The app does not load the model due to memory limitations."
            )
            submitted = st.form_submit_button("Submit")

tab1, tab2 = st.tabs(["Retrieved Text", "Retrieved Documents"])


with tab1:
    if document_type == "Single-Document":
        with st.expander("See Retrieved Text"):
            st.subheader("Retrieved Text:")
            for context_text in context_list:
                context_text = f"""{context_text}"""
                st.write(
                    f"<ul><li><p>{context_text}</p></li></ul>",
                    unsafe_allow_html=True,
                )
    else:
        with st.expander("See Retrieved Text"):
            st.subheader("Retrieved Text:")
            if multi_company_choice == "Compare Companies":
                multi_doc_context = (
                    multi_doc_context_first + multi_doc_context_second
                )

            sections = [
                s.strip()
                for s in multi_doc_context.split("Document: ")
                if s.strip()
            ]

            # Add "Document: " back to the beginning of each section
            context_list = [
                "Document: " + s[0:7] + "\n" + s[7:] for s in sections
            ]
            for context_text in context_list:
                context_text = f"""{context_text}"""
                st.write(
                    f"<ul><li><p>{context_text}</p></li></ul>",
                    unsafe_allow_html=True,
                )


with tab2:
    if document_type == "Single-Document":
        file_text = retrieve_transcript(data, year, quarter, ticker)
        with st.expander("See Transcript"):
            st.subheader("Earnings Call Transcript:")
            stx.scrollableTextbox(
                file_text, height=700, border=False, fontFamily="Helvetica"
            )
    else:
        if multi_company_choice == "Single-Company":
            for year, quarter in year_quarter_list:
                file_text = retrieve_transcript(data, year, quarter, ticker)
                with st.expander(f"See Transcript - {quarter} {year}"):
                    st.subheader(
                        "Earnings Call Transcript - {quarter} {year}:"
                    )
                    stx.scrollableTextbox(
                        file_text,
                        height=700,
                        border=False,
                        fontFamily="Helvetica",
                    )
        else:
            for year, quarter in year_quarter_list:
                file_text = retrieve_transcript(
                    data, year, quarter, ticker_first
                )
                with st.expander(f"See Transcript - {quarter} {year}"):
                    st.subheader(
                        "Earnings Call Transcript - {quarter} {year}:"
                    )
                    stx.scrollableTextbox(
                        file_text,
                        height=700,
                        border=False,
                        fontFamily="Helvetica",
                    )
            for year, quarter in year_quarter_list:
                file_text = retrieve_transcript(
                    data, year, quarter, ticker_second
                )
                with st.expander(f"See Transcript - {quarter} {year}"):
                    st.subheader(
                        "Earnings Call Transcript - {quarter} {year}:"
                    )
                    stx.scrollableTextbox(
                        file_text,
                        height=700,
                        border=False,
                        fontFamily="Helvetica",
                    )