Nevar pievienot vairāk kā 25 tēmas Tēmai ir jāsākas ar burtu vai ciparu, tā var saturēt domu zīmes ('-') un var būt līdz 35 simboliem gara.

disable_segments_from_index_task.py 2.9KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384
  1. import logging
  2. import time
  3. import click
  4. from celery import shared_task
  5. from sqlalchemy import select
  6. from core.rag.index_processor.index_processor_factory import IndexProcessorFactory
  7. from extensions.ext_database import db
  8. from extensions.ext_redis import redis_client
  9. from models.dataset import Dataset, DocumentSegment
  10. from models.dataset import Document as DatasetDocument
  11. logger = logging.getLogger(__name__)
  12. @shared_task(queue="dataset")
  13. def disable_segments_from_index_task(segment_ids: list, dataset_id: str, document_id: str):
  14. """
  15. Async disable segments from index
  16. :param segment_ids: list of segment ids
  17. :param dataset_id: dataset id
  18. :param document_id: document id
  19. Usage: disable_segments_from_index_task.delay(segment_ids, dataset_id, document_id)
  20. """
  21. start_at = time.perf_counter()
  22. dataset = db.session.query(Dataset).where(Dataset.id == dataset_id).first()
  23. if not dataset:
  24. logger.info(click.style(f"Dataset {dataset_id} not found, pass.", fg="cyan"))
  25. db.session.close()
  26. return
  27. dataset_document = db.session.query(DatasetDocument).where(DatasetDocument.id == document_id).first()
  28. if not dataset_document:
  29. logger.info(click.style(f"Document {document_id} not found, pass.", fg="cyan"))
  30. db.session.close()
  31. return
  32. if not dataset_document.enabled or dataset_document.archived or dataset_document.indexing_status != "completed":
  33. logger.info(click.style(f"Document {document_id} status is invalid, pass.", fg="cyan"))
  34. db.session.close()
  35. return
  36. # sync index processor
  37. index_processor = IndexProcessorFactory(dataset_document.doc_form).init_index_processor()
  38. segments = db.session.scalars(
  39. select(DocumentSegment).where(
  40. DocumentSegment.id.in_(segment_ids),
  41. DocumentSegment.dataset_id == dataset_id,
  42. DocumentSegment.document_id == document_id,
  43. )
  44. ).all()
  45. if not segments:
  46. db.session.close()
  47. return
  48. try:
  49. index_node_ids = [segment.index_node_id for segment in segments]
  50. index_processor.clean(dataset, index_node_ids, with_keywords=True, delete_child_chunks=False)
  51. end_at = time.perf_counter()
  52. logger.info(click.style(f"Segments removed from index latency: {end_at - start_at}", fg="green"))
  53. except Exception:
  54. # update segment error msg
  55. db.session.query(DocumentSegment).where(
  56. DocumentSegment.id.in_(segment_ids),
  57. DocumentSegment.dataset_id == dataset_id,
  58. DocumentSegment.document_id == document_id,
  59. ).update(
  60. {
  61. "disabled_at": None,
  62. "disabled_by": None,
  63. "enabled": True,
  64. }
  65. )
  66. db.session.commit()
  67. finally:
  68. for segment in segments:
  69. indexing_cache_key = f"segment_{segment.id}_indexing"
  70. redis_client.delete(indexing_cache_key)
  71. db.session.close()