batch_clean_document_task.py 3.0 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475
  1. import logging
  2. import time
  3. import click
  4. from celery import shared_task # type: ignore
  5. from core.rag.index_processor.index_processor_factory import IndexProcessorFactory
  6. from core.tools.utils.web_reader_tool import get_image_upload_file_ids
  7. from extensions.ext_database import db
  8. from extensions.ext_storage import storage
  9. from models.dataset import Dataset, DocumentSegment
  10. from models.model import UploadFile
  11. @shared_task(queue="dataset")
  12. def batch_clean_document_task(document_ids: list[str], dataset_id: str, doc_form: str, file_ids: list[str]):
  13. """
  14. Clean document when document deleted.
  15. :param document_ids: document ids
  16. :param dataset_id: dataset id
  17. :param doc_form: doc_form
  18. :param file_ids: file ids
  19. Usage: clean_document_task.delay(document_id, dataset_id)
  20. """
  21. logging.info(click.style("Start batch clean documents when documents deleted", fg="green"))
  22. start_at = time.perf_counter()
  23. try:
  24. dataset = db.session.query(Dataset).filter(Dataset.id == dataset_id).first()
  25. if not dataset:
  26. raise Exception("Document has no dataset")
  27. segments = db.session.query(DocumentSegment).filter(DocumentSegment.document_id.in_(document_ids)).all()
  28. # check segment is exist
  29. if segments:
  30. index_node_ids = [segment.index_node_id for segment in segments]
  31. index_processor = IndexProcessorFactory(doc_form).init_index_processor()
  32. index_processor.clean(dataset, index_node_ids, with_keywords=True, delete_child_chunks=True)
  33. for segment in segments:
  34. image_upload_file_ids = get_image_upload_file_ids(segment.content)
  35. for upload_file_id in image_upload_file_ids:
  36. image_file = db.session.query(UploadFile).filter(UploadFile.id == upload_file_id).first()
  37. try:
  38. storage.delete(image_file.key)
  39. except Exception:
  40. logging.exception(
  41. "Delete image_files failed when storage deleted, \
  42. image_upload_file_is: {}".format(upload_file_id)
  43. )
  44. db.session.delete(image_file)
  45. db.session.delete(segment)
  46. db.session.commit()
  47. if file_ids:
  48. files = db.session.query(UploadFile).filter(UploadFile.id.in_(file_ids)).all()
  49. for file in files:
  50. try:
  51. storage.delete(file.key)
  52. except Exception:
  53. logging.exception("Delete file failed when document deleted, file_id: {}".format(file.id))
  54. db.session.delete(file)
  55. db.session.commit()
  56. end_at = time.perf_counter()
  57. logging.info(
  58. click.style(
  59. "Cleaned documents when documents deleted latency: {}".format(end_at - start_at),
  60. fg="green",
  61. )
  62. )
  63. except Exception:
  64. logging.exception("Cleaned documents when documents deleted failed")