Vector Models and Indexes for Kindergarten Operations and Quality Self-Assessment: Find Supporting Evidence

Kindergarten Quality: review vector indexes for evidence lookup. Check source versions, input mappings and the values received by downstream nodes.

What this category's data looks like

This category's data includes self-assessment reports, issue lists, and rectification plans. Fields involve names, quantities, dates, signatures, and receipts. Self-assessment reports are updated by semester, issue lists roll with inspections, and rectification plans are adjusted by milestone. When these three types of materials enter this feature, the self-assessment report serves as the primary source of basis, the issue list provides item clues, and the rectification plan is used for version comparison. Material versions must be synchronized with current requirements, signatures and receipts must be complete, and the name, quantity, and date of the same record should be consistent across the three materials. During indexing, each material is split by paragraph, vectors and original data are stored separately, and retrieval first recalls vectors and then merges them with the original data.

Configuration and implementation checks

This step is executed by the indexing service and vector database, outputting vectors and index status (whether recall is possible and whether data in the same group is merged). "Basis items that can be traced back to the original text" are then completed by subsequent AI nodes, code nodes, or manual work (retrieval and citation trimming only take supporting materials).

ObjectCheckVerification action
Index modelConsistent with the vector model actually used by the knowledge baseCheck the index model in the knowledge base settings, confirm it is the same as the currently used model, and check the values of defaultToken and maxToken
Index fields of the data rowMultiple vector identifiers for one data item are all present, and the status is indexedOpen a data row and check the index fields, check the number of vector identifiers and the status, and confirm that defaultToken and maxToken are not exceeded
Rebuild task recordAfter changing the model, trigger only for data that has already been indexed, and the coverage scope is consistent with the task recordView the data scope covered by the rebuild task, check the status in the task record, and confirm that the maxToken and defaultToken settings have not changed

Three common errors

  • Symptom: The quantity of the same record does not match between the self-assessment report and the issue list. Inspect the following: The two materials were not synchronized when entered, or the versions were not synchronized, causing data discrepancies.
  • Symptom: The dates in the rectification plan are inconsistent with the self-assessment report. Inspect the following: The versions were not synchronized, or related materials were not updated after modification.
  • Symptom: Some materials lack signatures or receipts. Inspect the following: The process was not closed-loop, or the scanned copies omitted signature pages.

How to verify the setup

Feature acceptance: sample a data row and check the index fields to confirm that multiple vector identifiers are all present and the status is indexed; then retrieve once to see whether data in the same group is merged, with the highest score taken, and check the reference budget of limit.

Business test: ask a cross-material basis question and see whether the answer provides a traceable source.

  • Check whether the index model is consistent with the vector model actually used by the knowledge base
  • Check the index fields of the data row and confirm that the vector identifiers are complete and the status is indexed