(Optional) Running a custom job sweep for de-indexing existing documents from the vector database

You can use a de-indexing sweep to remove documents from the search index without deleting them or rebuilding the entire index.

About this task

A de-indexing sweep is a batch operation that performs the following actions:
  • De-indexes outdated documents or documents that no longer contain relevant content. These documents remain stored in the repository and are accessible for other business processes, workflows, or compliance purposes.
  • Updates the GenaiDateIndexed property to indicate when a document is no longer indexed. This improves metadata accuracy and can ensure the Content Assistant uses only relevant, indexed content when answering questions.

Procedure

  1. In the Administration Console for Content Platform Engine (ACCE), go to the object store where you want to remove documents from the vector index.
  2. Create a de-indexing sweep action for the object store.
    The sweep action specifies a Java-based action handler that defines the logic for how documents are de-indexed.
    1. In the object store navigation pane, right-click the Sweep Management > Sweep Actions folder and click New Sweep action.
    2. Enter a name for the de-indexing sweep action, and click Next.
      You can name the de-indexing sweep action as Gen AI vector De-indexing Request Job Sweep Action.
    3. Specify the Java class handler.
      The following table contains the Java class handler that you can specify based on your Content Platform Engine (CPE) version:
      Table 1. Java class handler
      CPE version Java class handler
      CPE 5.7 and later com.filenet.engine.genai.sweep.VectorDeIndexingJobSweepHandler
      CPE 5.6 and earlier com.filenet.genai.sweep.VectorDeIndexingJobSweepHandler
    4. Finish the wizard.
  3. Create a de-indexing sweep job for the object store.
    The sweep job uses the sweep action that you created in step 2 to de-index documents based on specific filter criteria.
    1. In the object store navigation pane, select the Sweep Management > Job Sweeps > Custom Jobs folder and click New.
    2. Enter a name for the de-indexing sweep job, and click Next.
      You can name the de-indexing sweep job as Genai Deindexing Job.
    3. Select your Target class.
      You can select the Document class or a subclass of the Document class.
    4. Specify the Filter expression for the de-indexing sweep job.
      You must use the filter expression "GenaiDateIndexed is not null" to identify documents that are currently indexed within the object store. You can also include additional conditions.
    5. Select the Sweep action for the de-indexing sweep job.
      You must select the de-indexing sweep action that you created in step 2.
    6. Finish the wizard.