Spaces:
Running
Running
File size: 45,665 Bytes
c6fc32c 86cf81a 71830dc c6fc32c 70cc7b0 c6fc32c 8e1caa1 c6fc32c 71830dc c6fc32c 17ebb8a c6fc32c 8ebff9f c574b13 8ebff9f c7d38be 60fd07f c7d38be 60fd07f c7d38be 60fd07f c7d38be 60fd07f c7d38be 60fd07f c7d38be 60fd07f 71830dc 60fd07f 924f5e7 60fd07f c7d38be 60fd07f c7d38be 60fd07f c7d38be c6fc32c 71830dc c6fc32c 71830dc c6fc32c 71830dc c6fc32c aad2e16 c6fc32c 86cf81a 71830dc 86cf81a aad2e16 8bec8f1 aad2e16 8bec8f1 aad2e16 8bec8f1 79f51f8 8bec8f1 4d1a43b aad2e16 c574b13 8bec8f1 86cf81a aad2e16 8bec8f1 86cf81a 8ebff9f 86cf81a 8ebff9f 86cf81a 8bec8f1 86cf81a 8bec8f1 86cf81a aad2e16 71830dc aad2e16 71830dc aad2e16 86cf81a aad2e16 86cf81a 71830dc 86cf81a 71830dc aad2e16 c6fc32c 71830dc c6fc32c 8ebff9f 8bec8f1 8ebff9f 70cc7b0 7ec4367 70cc7b0 c574b13 c6fc32c 70cc7b0 c6fc32c 70cc7b0 c7d38be 8e1caa1 71830dc c6fc32c ebf7d39 71830dc ebf7d39 8bec8f1 ebf7d39 624ba32 8bec8f1 c6fc32c ebf7d39 624ba32 c6fc32c 8ebff9f 8bec8f1 c6fc32c 624ba32 ebf7d39 c6fc32c 70cc7b0 c6fc32c 70cc7b0 c6fc32c 86cf81a c6fc32c 86cf81a c6fc32c 86cf81a c6fc32c 8e1caa1 33e11fc 8e1caa1 33e11fc 8e1caa1 33e11fc 8e1caa1 c6fc32c 86cf81a c6fc32c 8bec8f1 c574b13 c6fc32c 70cc7b0 8e1caa1 71830dc c6fc32c c7d38be c6fc32c 8ebff9f c6fc32c c7d38be c6fc32c 624ba32 ebf7d39 c6fc32c b3a5f93 c7d38be b3a5f93 c6fc32c c7d38be c6fc32c 70cc7b0 c6fc32c 70cc7b0 c7d38be 8e1caa1 c6fc32c c574b13 624ba32 c574b13 c6fc32c b3a5f93 c7d38be b3a5f93 c6fc32c 8e1caa1 b3a5f93 c7d38be c6fc32c c574b13 624ba32 c574b13 c6fc32c 33e11fc c6fc32c b3a5f93 c7d38be b3a5f93 c6fc32c c7d38be c6fc32c c574b13 624ba32 c574b13 c6fc32c b3a5f93 c7d38be b3a5f93 c6fc32c c7d38be c6fc32c b3a5f93 c7d38be b3a5f93 c6fc32c |
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 |
# components/review_dashboard_page.py
import gradio as gr
import datetime
import sentry_sdk
from sqlalchemy import orm
from components.header import Header
from utils.logger import Logger
from utils.cloud_server_audio_loader import CloudServerAudioLoader
from config import conf
from utils.database import get_db
from data.models import Annotation, TTSData, Annotator, Validation
from data.repository.annotator_workload_repo import AnnotatorWorkloadRepo
log = Logger()
LOADER = CloudServerAudioLoader(conf.FTP_URL)
class ReviewDashboardPage:
def __init__(self) -> None:
with gr.Column(visible=False) as self.container:
self.header = Header()
self.load_trigger = gr.Number(value=0, visible=False) # Add this hidden trigger
# Review info banner
with gr.Row():
self.review_info = gr.Markdown("", elem_classes="review-banner")
with gr.Row():
# Left Column - Review Content
with gr.Column(scale=3):
with gr.Row():
self.tts_id = gr.Textbox(label="ID", interactive=False, scale=1)
self.filename = gr.Textbox(label="Filename", interactive=False, scale=3)
self.sentence = gr.Textbox(
label="Original Sentence", interactive=False, max_lines=5, rtl=True
)
self.ann_sentence = gr.Textbox(
label="Annotated Sentence (by Original Annotator)",
interactive=False, max_lines=5, rtl=True
)
with gr.Row():
# self.annotator_name = gr.Textbox(label="Original Annotator", interactive=False, scale=1) # Removed for anonymization
self.annotated_at = gr.Textbox(label="Annotated At", interactive=False, scale=2)
# Review Actions
with gr.Row():
self.btn_approve = gr.Button("β
Approve", variant="primary", min_width=120)
self.btn_reject = gr.Button("β Reject", variant="stop", min_width=120)
self.btn_skip = gr.Button("βοΈ Skip (No Decision)", min_width=150)
# Navigation
with gr.Row():
self.btn_prev = gr.Button("β¬
οΈ Previous", min_width=120)
self.btn_next = gr.Button("Next β‘οΈ", min_width=120)
# Jump controls
with gr.Row():
self.jump_data_id_input = gr.Number(
label="Jump to ID",
value=None,
precision=0,
interactive=True,
min_width=120
)
self.btn_jump = gr.Button("Go to ID", min_width=70)
# Right Column - Audio
with gr.Column(scale=2):
self.btn_load_voice = gr.Button("Load Audio & Play", min_width=150)
self.audio = gr.Audio(
label="π Audio", interactive=False, autoplay=True
)
# Review status display
with gr.Group():
gr.Markdown("### Review Status")
self.current_validation_status = gr.Textbox(
label="Current Status", interactive=False
)
self.rejection_reason_input = gr.Textbox(
label="Rejection Reason",
placeholder="Enter reason and press Enter or click away...",
interactive=True,
visible=False,
max_lines=3,
elem_id="rejection_reason_input" # Added elem_id for clarity
)
# State variables
self.items_state = gr.State([])
self.idx_state = gr.State(0)
self.original_audio_state = gr.State(None)
self.rejection_mode_active = gr.State(False) # Track if waiting for rejection reason
# List of interactive UI elements for enabling/disabling
self.interactive_ui_elements = [
self.btn_prev, self.btn_next, self.btn_approve, self.btn_reject,
self.btn_skip, self.btn_jump, self.jump_data_id_input, self.btn_load_voice
]
def register_callbacks(self, login_page, session_state: gr.State, root_blocks: gr.Blocks):
self.header.register_callbacks(login_page, self, session_state)
def update_ui_interactive_state(is_interactive: bool):
updates = []
for elem in self.interactive_ui_elements:
if elem == self.btn_load_voice and not is_interactive:
updates.append(gr.update(value="β³ Loading Audio...", interactive=False))
elif elem == self.btn_load_voice and is_interactive:
updates.append(gr.update(value="Load Audio & Play", interactive=True))
else:
updates.append(gr.update(interactive=is_interactive))
return updates
def download_voice_fn(filename_to_load):
if not filename_to_load:
return None, None, gr.update(value=None, autoplay=False)
# try:
log.info(f"Downloading voice for review: {filename_to_load}")
# Show progress to user
# gr.Info(f"Loading audio file: {filename_to_load}")
sr, wav = LOADER.load_audio(filename_to_load)
log.info(f"Successfully loaded audio: {filename_to_load} (SR: {sr}, Length: {len(wav)} samples)")
# gr.Info(f"β
Audio loaded successfully!")
return (sr, wav), (sr, wav.copy()), gr.update(value=(sr, wav), autoplay=True)
# except TimeoutError as e:
# log.error(f"Audio download timeout for {filename_to_load}: {e}")
# sentry_sdk.capture_exception(e)
# raise
# except ConnectionError as e:
# log.error(f"Audio download connection error for {filename_to_load}: {e}")
# sentry_sdk.capture_exception(e)
# gr.Error(f"π Connection error loading audio: {filename_to_load}. Please check your internet connection.")
# return None, None, gr.update(value=None, autoplay=False)
# except FileNotFoundError as e:
# log.error(f"Audio file not found for {filename_to_load}: {e}")
# sentry_sdk.capture_exception(e)
# gr.Error(f"π Audio file not found: {filename_to_load}")
# return None, None, gr.update(value=None, autoplay=False)
# except Exception as e:
# log.error(f"Audio download failed for {filename_to_load}: {e}")
# sentry_sdk.capture_exception(e)
# gr.Error(f"β Failed to load audio: {filename_to_load}. Error: {e}")
# return None, None, gr.update(value=None, autoplay=False)
def get_validation_status_for_item(db, annotation_id, user_id, annotation_obj):
"""Get validation status for a specific item - called on-demand"""
validation = db.query(Validation).filter_by(
annotation_id=annotation_id,
validator_id=user_id
).first()
# Check if annotation is deleted
is_deleted = not annotation_obj.annotated_sentence or annotation_obj.annotated_sentence.strip() == ""
validation_status = "Not Reviewed"
if validation:
if validation.validated:
validation_status = "Approved"
else:
validation_status = "Rejected"
if validation.description:
validation_status += f" ({validation.description})"
# For deleted annotations, show special status
if is_deleted and validation_status == "Not Reviewed":
validation_status = "Not Reviewed (Deleted)"
return validation_status, is_deleted
def get_review_progress_fn(session):
"""Calculate review progress for the current reviewer with beautiful tqdm-style display"""
user_id = session.get("user_id")
username = session.get("username")
if not user_id or not username:
return ""
# Check if user is a reviewer
if username not in conf.REVIEW_MAPPING.values():
return ""
# Find target annotator
target_annotator = None
for annotator_name, reviewer_name in conf.REVIEW_MAPPING.items():
if reviewer_name == username:
target_annotator = annotator_name
break
if not target_annotator:
return ""
with get_db() as db:
try:
# Get target annotator's ID
target_annotator_obj = db.query(Annotator).filter_by(name=target_annotator).first()
if not target_annotator_obj:
return f"β οΈ **Error:** Annotator '{target_annotator}' not found"
# Count total annotations for target annotator
total_count = db.query(Annotation).filter(
Annotation.annotator_id == target_annotator_obj.id
).count()
# Count reviewed annotations (have validation from this reviewer)
reviewed_count = db.query(Annotation).join(
Validation, Annotation.id == Validation.annotation_id
).filter(
Annotation.annotator_id == target_annotator_obj.id,
Validation.validator_id == user_id
).count()
if total_count > 0:
percentage = (reviewed_count / total_count) * 100
# Create tqdm-style progress bar
bar_width = 30 # Width of the progress bar in characters
filled = int((percentage / 100) * bar_width)
empty = bar_width - filled
# Different colors based on progress
if percentage < 25:
color = "π΄" # Red for low progress
bar_color = "progress-bar-low"
elif percentage < 50:
color = "π‘" # Yellow for medium-low progress
bar_color = "progress-bar-medium-low"
elif percentage < 75:
color = "π " # Orange for medium progress
bar_color = "progress-bar-medium"
elif percentage < 100:
color = "π’" # Green for high progress
bar_color = "progress-bar-high"
else:
color = "β
" # Check mark for complete
bar_color = "progress-bar-complete"
# Create the visual progress bar with Unicode blocks
progress_bar = "β" * filled + "β" * empty
# Estimate remaining items
remaining = total_count - reviewed_count
# Create the beautiful progress display
progress_html = f"""
<div class="progress-container">
<div class="progress-header">
<span class="progress-icon">{color}</span>
<strong>Review Progress</strong>
</div>
<div class="progress-bar-container">
<span class="progress-percentage">{percentage:.1f}%</span>
<div class="progress-bar {bar_color}">
<span class="progress-fill" style="width: {percentage:.1f}%"></span>
</div>
<span class="progress-stats">{reviewed_count}/{total_count}</span>
</div>
<div class="progress-details">
π <code>{progress_bar}</code>
<span class="remaining-items">({remaining} remaining)</span>
</div>
</div>
"""
return progress_html
else:
return f"π **No items found for {target_annotator}**"
except Exception as e:
log.error(f"Error calculating review progress for user {user_id}: {e}")
return f"β οΈ **Error calculating progress**"
def load_review_items_fn(session):
user_id = session.get("user_id")
username = session.get("username")
if not user_id or not username:
log.warning("load_review_items_fn: user not found in session")
return [], 0, "", "", "", "", "", "", "", "", gr.update(value=None, autoplay=False), gr.update(visible=False, value=""), False, gr.update(value="β Reject")
# Check if user is in Phase 2 (should be a reviewer)
if username not in conf.REVIEW_MAPPING.values():
log.warning(f"User {username} is not assigned as a reviewer")
return [], 0, "", "", "", "", "", "", "", "", gr.update(value=None, autoplay=False), gr.update(visible=False, value=""), False, gr.update(value="β Reject")
# Find which annotator this user should review
target_annotator = None
for annotator_name, reviewer_name in conf.REVIEW_MAPPING.items():
if reviewer_name == username:
target_annotator = annotator_name
break
if not target_annotator:
log.warning(f"No target annotator found for reviewer {username}")
return [], 0, "", "", "", "", "", "", "", "", gr.update(value=None, autoplay=False), gr.update(visible=False, value=""), False, gr.update(value="β Reject")
# Load annotations from target annotator with FAST INITIAL LOADING
with get_db() as db:
# Get target annotator's ID
target_annotator_obj = db.query(Annotator).filter_by(name=target_annotator).first()
if not target_annotator_obj:
log.error(f"Target annotator {target_annotator} not found in database")
return [], 0, f"Review Target Error: Annotator '{target_annotator}' not found.", "", "", "", "", "", "", "", gr.update(value=None, autoplay=False), gr.update(visible=False, value=""), False, gr.update(value="β Reject")
log.info(f"Found target annotator with ID: {target_annotator_obj.id}")
# FAST INITIAL QUERY: Load only essential data without complex validation processing
# Reduced batch size for instant loading in HuggingFace spaces
INITIAL_BATCH_SIZE = 5 # Load only 5 items initially for instant response
# Simple query to get basic annotation data quickly
initial_query = db.query(
Annotation,
TTSData.filename,
TTSData.sentence
).join(
TTSData, Annotation.tts_data_id == TTSData.id
).filter(
Annotation.annotator_id == target_annotator_obj.id
).order_by(Annotation.id).limit(INITIAL_BATCH_SIZE)
initial_results = initial_query.all()
# Get total count for progress info (this is fast)
total_count = db.query(Annotation).filter(
Annotation.annotator_id == target_annotator_obj.id
).count()
log.info(f"Fast initial load: {len(initial_results)} annotations out of {total_count} total for target annotator ID {target_annotator_obj.id}")
# Process items with minimal data - validation status will be loaded on-demand
items = []
for annotation, filename, sentence in initial_results:
# Check if annotation is deleted (minimal processing)
is_deleted = not annotation.annotated_sentence or annotation.annotated_sentence.strip() == ""
annotated_sentence_display = "[DELETED ANNOTATION]" if is_deleted else annotation.annotated_sentence
items.append({
"annotation_id": annotation.id,
"tts_id": annotation.tts_data_id,
"filename": filename,
"sentence": sentence,
"annotated_sentence": annotated_sentence_display,
"is_deleted": is_deleted,
"annotated_at": annotation.annotated_at.isoformat() if annotation.annotated_at else "",
"validation_status": "Loading...", # Will be loaded on-demand
"validation_loaded": False # Track if validation status has been loaded
})
# Find the first item that is not reviewed (prioritize non-deleted annotations)
initial_idx = 0
if items:
found_unreviewed = False
# First, try to find unreviewed non-deleted annotations
for i, item_data in enumerate(items):
if (item_data["validation_status"] == "Not Reviewed" and
not item_data.get("is_deleted", False)):
initial_idx = i
found_unreviewed = True
break
# If no unreviewed non-deleted items, look for any unreviewed items
if not found_unreviewed:
for i, item_data in enumerate(items):
if item_data["validation_status"].startswith("Not Reviewed"):
initial_idx = i
found_unreviewed = True
break
# If no unreviewed items at all, use the last item
if not found_unreviewed:
initial_idx = len(items) - 1 if items else 0
# Set initial display
if items:
initial_item = items[initial_idx]
review_info_text = f"π **Phase 2 Review Mode** - Reviewing assigned annotations. Loaded {len(items)} of {total_count} total items."
# Ensure correct order of return values for 12 outputs
# items, idx, review_info, tts_id, filename, sentence, ann_sentence, annotated_at, validation_status, annotator_placeholder, audio_update, rejection_reason_update
rejection_reason_val = ""
rejection_visible_val = False
if initial_item["validation_status"].startswith("Rejected"):
start_paren = initial_item["validation_status"].find("(")
end_paren = initial_item["validation_status"].find(")")
if start_paren != -1 and end_paren != -1:
rejection_reason_val = initial_item["validation_status"][start_paren+1:end_paren]
rejection_visible_val = True
return (
items,
initial_idx,
review_info_text,
str(initial_item["tts_id"]),
initial_item["filename"],
initial_item["sentence"],
initial_item["annotated_sentence"],
initial_item["annotated_at"],
initial_item["validation_status"],
"", # Placeholder for the original annotator name (maps to header.welcome)
gr.update(value=None, autoplay=False), # audio_update
gr.update(visible=rejection_visible_val, value=rejection_reason_val), # rejection_reason_input update
False, # Reset rejection mode
gr.update(value="β Reject") # Reset reject button
)
else:
# Ensure correct order and number of return values for empty items (14 outputs)
return [], 0, f"π **Phase 2 Review Mode** - No annotations found for review.", "", "", "", "", "", "", "", gr.update(value=None, autoplay=False), gr.update(visible=False, value=""), False, gr.update(value="β Reject")
# except Exception as e:
# log.error(f"Error loading review items: {e}")
# sentry_sdk.capture_exception(e)
# gr.Error(f"Failed to load review data: {e}")
# # Ensure correct order and number of return values for error case (14 outputs)
# return [], 0, "", "", "", "", "", "", "", "", gr.update(value=None, autoplay=False), gr.update(visible=False, value=""), False, gr.update(value="β Reject")
def show_current_review_item_fn(items, idx, session):
if not items or idx >= len(items) or idx < 0:
# tts_id, filename, sentence, ann_sentence, annotated_at, validation_status, annotator_name_placeholder, audio_update, rejection_reason_update, rejection_mode_reset, btn_reject_update
return "", "", "", "", "", "", "", gr.update(value=None, autoplay=False), gr.update(visible=False, value=""), False, gr.update(value="β Reject")
current_item = items[idx]
# Load validation status on-demand if not already loaded
if not current_item.get("validation_loaded", False):
user_id = session.get("user_id")
if user_id:
with get_db() as db:
try:
# Get the full annotation object for validation processing
annotation_obj = db.query(Annotation).filter_by(id=current_item["annotation_id"]).first()
if annotation_obj:
validation_status, is_deleted = get_validation_status_for_item(db, current_item["annotation_id"], user_id, annotation_obj)
current_item["validation_status"] = validation_status
current_item["is_deleted"] = is_deleted
current_item["validation_loaded"] = True
# Update displayed annotation if deleted
if is_deleted:
current_item["annotated_sentence"] = "[DELETED ANNOTATION]"
log.info(f"Loaded validation status for item {idx}: {validation_status}")
except Exception as e:
log.error(f"Error loading validation status for item {idx}: {e}")
current_item["validation_status"] = "Error loading status"
rejection_reason = ""
rejection_visible = False
# Check if this is a deleted annotation
is_deleted = current_item.get("is_deleted", False)
if current_item["validation_status"].startswith("Rejected"):
# Extract reason from status like "Rejected (reason)" or just use empty if no parenthesis
start_paren = current_item["validation_status"].find("(")
end_paren = current_item["validation_status"].find(")")
if start_paren != -1 and end_paren != -1:
rejection_reason = current_item["validation_status"][start_paren+1:end_paren]
rejection_visible = True
return (
str(current_item["tts_id"]),
current_item["filename"],
current_item["sentence"],
current_item["annotated_sentence"],
current_item["annotated_at"],
current_item["validation_status"],
"", # Placeholder for annotator_name
gr.update(value=None, autoplay=False),
gr.update(visible=rejection_visible, value=rejection_reason),
False, # Reset rejection mode
gr.update(value="β Reject") # Reset reject button text
)
def update_review_info_fn(items, total_count):
"""Update the review info banner with current loaded items count"""
if items:
return f"π **Phase 2 Review Mode** - Reviewing assigned annotations. Loaded {len(items)} of {total_count} total items."
else:
return f"π **Phase 2 Review Mode** - No annotations found for review."
def navigate_and_load_fn(items, current_idx, direction, session):
"""Combined navigation and loading function"""
if not items:
return items, 0, ""
# Navigate
if direction == "next":
new_idx = min(current_idx + 1, len(items) - 1)
# Only load more items when user reaches the LAST item of a batch
should_load_more = (new_idx == len(items) - 1 and len(items) % 5 == 0)
if should_load_more:
log.info(f"User reached end of loaded items ({new_idx}/{len(items)}), will load more items")
# Load more items
updated_items, total_count = load_more_items_fn(items, session, current_batch_size=10)
# Update review info with new count
review_info = update_review_info_fn(updated_items, total_count)
return updated_items, new_idx, review_info
else:
return items, new_idx, "" # No review info update needed
else: # prev
new_idx = max(current_idx - 1, 0)
return items, new_idx, "" # No review info update needed
def save_validation_fn(items, idx, session, approved: bool, rejection_reason: str = ""):
if not items or idx >= len(items):
gr.Error("Invalid item index")
return items, "Error: Invalid item index", gr.update(visible=False)
user_id = session.get("user_id")
if not user_id:
gr.Error("User not logged in")
return items, "Error: User not logged in", gr.update(visible=False)
current_item = items[idx]
annotation_id = current_item["annotation_id"]
log.info(f"Saving validation for annotation_id: {annotation_id}, validator_id: {user_id}, approved: {approved}, reason: {rejection_reason}")
with get_db() as db:
# try:
existing_validation = db.query(Validation).filter_by(
annotation_id=annotation_id,
validator_id=user_id
).first()
if existing_validation:
log.info(f"Updating existing validation for annotation_id: {annotation_id}")
existing_validation.validated = approved
existing_validation.description = rejection_reason if not approved else None
existing_validation.validated_at = datetime.datetime.utcnow()
else:
log.info(f"Creating new validation for annotation_id: {annotation_id}")
new_validation = Validation(
annotation_id=annotation_id,
validator_id=user_id,
validated=approved,
description=rejection_reason if not approved else None,
validated_at=datetime.datetime.utcnow(),
)
db.add(new_validation)
db.commit()
log.info(f"Validation saved successfully for annotation_id: {annotation_id}")
items[idx]["validation_status"] = "Approved" if approved else f"Rejected ({rejection_reason})" if rejection_reason else "Rejected"
# Show rejection reason input only if rejected, otherwise hide and clear
rejection_input_update = gr.update(visible=not approved, value="" if approved else rejection_reason)
return items, items[idx]["validation_status"], rejection_input_update
# except Exception as e:
# db.rollback()
# log.error(f"Error saving validation: {e}")
# sentry_sdk.capture_exception(e)
# gr.Error(f"Failed to save validation: {e}")
# return items, current_item["validation_status"], gr.update(visible=False) # Return original status and hide input on error
def handle_rejection_fn(items, idx, session, rejection_reason, rejection_mode_active):
"""Handle rejection button click - two-step process"""
if not items or idx >= len(items):
return items, "Error: Invalid item", gr.update(visible=False), False, gr.update(value="β Reject")
current_item = items[idx]
current_status = current_item["validation_status"]
if not rejection_mode_active:
# First click - show rejection reason input and change button text
return (
items, # items unchanged
current_status, # Keep current validation status
gr.update(visible=True, value=""), # Show rejection reason input, clear any existing value
True, # Set rejection mode active
gr.update(value="β οΈ Confirm Reject") # Change button text
)
else:
# Second click - validate reason and save if provided
if not rejection_reason or rejection_reason.strip() == "":
gr.Warning("Rejection reason cannot be empty. Please provide a reason before confirming rejection.")
return (
items, # items unchanged
current_status, # Keep current validation status
gr.update(visible=True, value=rejection_reason), # Keep input visible
True, # Keep rejection mode active
gr.update(value="β οΈ Confirm Reject") # Keep button text
)
else:
# Save the rejection with reason
updated_items, validation_status, rejection_input_update = save_validation_fn(
items, idx, session, approved=False, rejection_reason=rejection_reason.strip()
)
return (
updated_items,
validation_status,
gr.update(visible=False, value=""), # Hide rejection input after successful save
False, # Reset rejection mode
gr.update(value="β Reject") # Reset button text
)
def jump_by_data_id_fn(items, target_data_id, current_idx):
if not target_data_id:
return current_idx
# try:
target_id = int(target_data_id)
for i, item in enumerate(items):
if item["tts_id"] == target_id:
return i
gr.Warning(f"Data ID {target_id} not found in review items")
# except ValueError:
# sentry_sdk.capture_exception()
# gr.Warning(f"Invalid Data ID format: {target_data_id}")
return current_idx
def load_more_items_fn(items, session, current_batch_size=10):
"""Load more items when user needs them (pagination support)"""
user_id = session.get("user_id")
username = session.get("username")
if not user_id or not username:
return items, 0 # Return existing items if no user session
# Find target annotator
target_annotator = None
for annotator_name, reviewer_name in conf.REVIEW_MAPPING.items():
if reviewer_name == username:
target_annotator = annotator_name
break
if not target_annotator:
return items, 0
with get_db() as db:
target_annotator_obj = db.query(Annotator).filter_by(name=target_annotator).first()
if not target_annotator_obj:
return items, 0
# Get total count for updated review info
total_count = db.query(Annotation).filter(
Annotation.annotator_id == target_annotator_obj.id
).count()
# Load next batch starting from where we left off
offset = len(items)
# FAST LOADING: Use same strategy as initial load - simple query without complex JOINs
query = db.query(
Annotation,
TTSData.filename,
TTSData.sentence
).join(
TTSData, Annotation.tts_data_id == TTSData.id
).filter(
Annotation.annotator_id == target_annotator_obj.id
).order_by(Annotation.id).offset(offset).limit(current_batch_size)
results = query.all()
# Process new items with minimal data - validation status loaded on-demand
new_items = []
for annotation, filename, sentence in results:
# Check if annotation is deleted (minimal processing)
is_deleted = not annotation.annotated_sentence or annotation.annotated_sentence.strip() == ""
annotated_sentence_display = "[DELETED ANNOTATION]" if is_deleted else annotation.annotated_sentence
new_items.append({
"annotation_id": annotation.id,
"tts_id": annotation.tts_data_id,
"filename": filename,
"sentence": sentence,
"annotated_sentence": annotated_sentence_display,
"is_deleted": is_deleted,
"annotated_at": annotation.annotated_at.isoformat() if annotation.annotated_at else "",
"validation_status": "Loading...", # Will be loaded on-demand
"validation_loaded": False # Track if validation status has been loaded
})
# Combine with existing items
all_items = items + new_items
log.info(f"Loaded {len(new_items)} more items, total now: {len(all_items)}")
return all_items, total_count
# Output definitions
review_display_outputs = [
self.tts_id, self.filename, self.sentence, self.ann_sentence,
self.annotated_at,
self.current_validation_status,
self.header.welcome, # Placeholder for anonymized annotator name
self.audio,
self.rejection_reason_input, # Added rejection reason input to display outputs
self.rejection_mode_active, # Added rejection mode state
self.btn_reject # Added reject button to display outputs
]
# Trigger data loading when load_trigger changes (after successful login for a reviewer)
self.load_trigger.change(
fn=lambda: update_ui_interactive_state(False),
outputs=self.interactive_ui_elements
).then(
fn=load_review_items_fn,
inputs=[session_state],
outputs=[self.items_state, self.idx_state, self.review_info] + review_display_outputs
).then(
fn=get_review_progress_fn,
inputs=[session_state],
outputs=[self.header.progress_display]
).then(
fn=lambda: (None, gr.update(value=None)), # Clear audio state
outputs=[self.original_audio_state, self.audio]
).then(
fn=lambda: update_ui_interactive_state(True),
outputs=self.interactive_ui_elements
)
# Audio loading is now manual only via the Load Audio button
# Removed automatic filename.change callback to prevent slow loading during initialization
# Navigation buttons
for btn, direction in [(self.btn_prev, "prev"), (self.btn_next, "next")]:
btn.click(
fn=lambda: update_ui_interactive_state(False),
outputs=self.interactive_ui_elements
).then(
fn=lambda items, idx, session, dir=direction: navigate_and_load_fn(items, idx, dir, session),
inputs=[self.items_state, self.idx_state, session_state],
outputs=[self.items_state, self.idx_state, self.review_info]
).then(
fn=show_current_review_item_fn,
inputs=[self.items_state, self.idx_state, session_state],
outputs=review_display_outputs
).then(
# Auto-load audio with autoplay for smooth navigation
fn=download_voice_fn,
inputs=[self.filename],
outputs=[self.audio, self.original_audio_state, self.audio]
).then(
lambda: gr.update(value=None),
outputs=self.jump_data_id_input
).then(
fn=lambda: update_ui_interactive_state(True),
outputs=self.interactive_ui_elements
)
# Approve/Reject buttons
self.btn_approve.click(
fn=lambda items, idx, session: save_validation_fn(items, idx, session, approved=True, rejection_reason=""), # Pass empty rejection_reason
inputs=[self.items_state, self.idx_state, session_state],
outputs=[self.items_state, self.current_validation_status, self.rejection_reason_input]
).then(
fn=get_review_progress_fn, # Update progress after approval
inputs=[session_state],
outputs=[self.header.progress_display]
).then(
fn=lambda: False, # Reset rejection mode
outputs=[self.rejection_mode_active]
).then(
fn=lambda: gr.update(value="β Reject"), # Reset reject button
outputs=[self.btn_reject]
).then(
fn=lambda items, idx, session: navigate_and_load_fn(items, idx, "next", session),
inputs=[self.items_state, self.idx_state, session_state],
outputs=[self.items_state, self.idx_state, self.review_info]
).then(
fn=show_current_review_item_fn,
inputs=[self.items_state, self.idx_state, session_state],
outputs=review_display_outputs
).then(
# Auto-load audio with autoplay after moving to next item
fn=download_voice_fn,
inputs=[self.filename],
outputs=[self.audio, self.original_audio_state, self.audio]
)
self.btn_reject.click(
fn=handle_rejection_fn,
inputs=[self.items_state, self.idx_state, session_state, self.rejection_reason_input, self.rejection_mode_active],
outputs=[self.items_state, self.current_validation_status, self.rejection_reason_input, self.rejection_mode_active, self.btn_reject]
).then(
fn=lambda items, idx, session, rejection_mode: get_review_progress_fn(session) if not rejection_mode else "", # Update progress only after successful rejection
inputs=[self.items_state, self.idx_state, session_state, self.rejection_mode_active],
outputs=[self.header.progress_display]
).then(
fn=lambda items, idx, session, rejection_mode: navigate_and_load_fn(items, idx, "next", session) if not rejection_mode else (items, idx, ""),
inputs=[self.items_state, self.idx_state, session_state, self.rejection_mode_active],
outputs=[self.items_state, self.idx_state, self.review_info]
).then(
fn=lambda items, idx, session, rejection_mode: show_current_review_item_fn(items, idx, session) if not rejection_mode else (
str(items[idx]["tts_id"]) if items and idx < len(items) else "",
items[idx]["filename"] if items and idx < len(items) else "",
items[idx]["sentence"] if items and idx < len(items) else "",
items[idx]["annotated_sentence"] if items and idx < len(items) else "",
items[idx]["annotated_at"] if items and idx < len(items) else "",
items[idx]["validation_status"] if items and idx < len(items) else "",
"", # annotator placeholder
gr.update(value=None, autoplay=False), # audio
gr.update(), # rejection_reason_input - don't change
rejection_mode, # keep rejection mode as is
gr.update() # btn_reject - don't change
),
inputs=[self.items_state, self.idx_state, session_state, self.rejection_mode_active],
outputs=review_display_outputs
).then(
# Auto-load audio with autoplay only if we moved to next item (not in rejection mode)
fn=lambda filename, rejection_mode: download_voice_fn(filename) if not rejection_mode else (None, None, gr.update(value=None, autoplay=False)),
inputs=[self.filename, self.rejection_mode_active],
outputs=[self.audio, self.original_audio_state, self.audio]
)
# Skip button (just navigate to next)
self.btn_skip.click(
fn=lambda items, idx, session: navigate_and_load_fn(items, idx, "next", session),
inputs=[self.items_state, self.idx_state, session_state],
outputs=[self.items_state, self.idx_state, self.review_info]
).then(
fn=show_current_review_item_fn,
inputs=[self.items_state, self.idx_state, session_state],
outputs=review_display_outputs
).then(
# Auto-load audio with autoplay after skipping
fn=download_voice_fn,
inputs=[self.filename],
outputs=[self.audio, self.original_audio_state, self.audio]
)
# Jump button
self.btn_jump.click(
fn=jump_by_data_id_fn,
inputs=[self.items_state, self.jump_data_id_input, self.idx_state],
outputs=self.idx_state
).then(
fn=show_current_review_item_fn,
inputs=[self.items_state, self.idx_state, session_state],
outputs=review_display_outputs
).then(
# Auto-load audio with autoplay after jumping
fn=download_voice_fn,
inputs=[self.filename],
outputs=[self.audio, self.original_audio_state, self.audio]
).then(
lambda: gr.update(value=None),
outputs=self.jump_data_id_input
)
# Load audio button
self.btn_load_voice.click(
fn=lambda: update_ui_interactive_state(False),
outputs=self.interactive_ui_elements
).then(
fn=download_voice_fn,
inputs=[self.filename],
outputs=[self.audio, self.original_audio_state, self.audio]
).then(
fn=lambda: update_ui_interactive_state(True),
outputs=self.interactive_ui_elements
)
return self.container
|