What this category's data looks like
This page focuses on data versioning and change maintenance in enterprise compliance and security training scenarios, covering three types of materials: compliance training courseware, safety operating procedures, and exam records. Courseware and procedures require verification of names, quantities, dates, and signatures; exam records require verification of receipts and signature fields. When materials enter the vector model and indexing feature, courseware and procedures serve as text source data to generate vector identifiers, while exam records participate in indexing as table fields. The update cadence is triggered by training cycles and procedure revisions; old versions must be traceable, and the current version must be unique. In terms of material roles, courseware and procedures are the indexing subjects, while exam records are the verification basis for reference synchronization.
Configuration and implementation checks
The indexing service and vector database only output vectors and indexing status (whether recall is possible, whether data in the same group is merged). The version ledger and change records must be handled subsequently by AI nodes, code nodes, or manual work (retrieval and reference trimming only take supporting materials). Source data is stored in business database tables, and multiple vector identifiers are recorded via the index field. Retrieval first recalls vectors and then merges them back to the source data by identifier, taking the highest score for the same group of data.
| Object | Check | Verification action |
|---|---|---|
| Compliance training courseware | Whether names, quantities, dates, and signatures are consistent with the current version | Open a data row to check the index field, confirm that multiple vector identifiers are all present, perform a retrieval to see whether data in the same group is merged, and check whether the status is indexed or a rebuild task record exists |
| Safety operating procedures | Whether the version number and references are updated synchronously, and whether the old version is traceable | After changing one piece of content, check the data range covered by the rebuild task record, verify the reference text tokens corresponding to limit, and record the array length separately |
| Exam records | Whether signatures or receipts are complete, and whether dates and quantities match | In the knowledge base settings, verify that the indexing model is consistent with the vector model actually used by the knowledge base, and sample a data row to inspect the index field |
| Version ledger and change records | The current version is unique, the old version is traceable, and references are synchronized | Change one piece of content, see whether the version number and references are updated synchronously, and check whether the rebuild task record covers data that has already been indexed |
Three common errors
- Symptom: Duplicate entries appear in retrieval results. Inspect the following: Multiple vector identifiers for the same group of data are not correctly merged, or versions are not synchronized, causing both the old and current versions to be recalled.
- Symptom: Recall quality drops after changing models. Inspect the following: After changing models, only data that has already been indexed is processed; data rows that have not completed indexing do not participate in recall, so indexing must be completed before rebuilding.
- Symptom: A data row has content but cannot be retrieved. Inspect the following: The index field is missing or no vector identifier has been generated, or exam records lack signatures and receipts, causing reference synchronization to break.
How to verify the setup
Acceptance for this feature: sample a data row to inspect the index field, confirm that multiple vector identifiers are all present, then perform a retrieval to see whether the same group is merged, and check whether the status is indexed or a rebuild task record exists.
Business test: change one piece of content, see whether the version number and references are updated synchronously, whether the old version is traceable, and whether the current version is unique.
- Open a data row to check the index field, confirm that multiple vector identifiers are all present, and perform a retrieval to see whether data in the same group is merged.
- In the knowledge base settings, verify that the indexing model is consistent with the vector model actually used by the knowledge base, and check the data range covered by the rebuild task record.