What this category's data looks like
This category's data covers three types of materials: degree application forms, thesis review forms, and defense resolutions. Sources are training units and departments, and the update cadence rolls with defense batches and degree committee milestones. Fields include name, quantity, date, signature, and receipt. The degree application form carries personal and degree application information, the thesis review form carries review comments and reviewer signatures, and the defense resolution carries the defense conclusion and committee member signatures. When these three types of materials enter the vector model and indexing feature, each is split, vectorized, and written into the vector store as an independent data row. The original data and index fields are retained in the business database for later retrieval and citation trimming. The name, quantity, and date of the same record across the three materials must stay consistent, signatures and receipts must be complete, and versions must be synchronized.
Configuration and implementation checks
This step is executed by the indexing service and the vector store, outputting vectors and index status (whether recall is possible, whether data in the same group is merged). "Version ledger and change records" is then completed by subsequent AI nodes, code nodes, or humans (retrieval and citation trimming only take supporting materials).
| Object | Check | Verification action |
|---|---|---|
| Degree application form | Name, quantity, date, and signature must align item by item with the thesis review form and defense resolution, and the version must be the current unique one | Open a data row and check the index fields, confirm that multiple vector identifiers are complete and the status is indexed, then run a retrieval once to see whether the same group is merged |
| Thesis review form | Reviewer signatures and receipts must not be missing, and the name, quantity, and date of the same record must not mismatch across the three materials | In the knowledge base settings, check that the index model matches the vector model actually in use, and review the data range covered by the rebuild task records |
| Defense resolution | The defense conclusion, committee member signatures, and date must be consistent with the application form and review form, old versions must be traceable, and citations must be synchronized | After changing one piece of content, re-trigger indexing, run a retrieval to check whether the version number and citations are updated in sync, and record the array count separately |
| Data row index fields | All multiple vector identifiers of one piece of data are present, and when retrieving data in the same group, the highest score is taken and merged | Check the values of limit in splitting and citation budgeting, and confirm that the tokens of the cited text and the array count are recorded separately |
Three common errors
- Symptom: the name, quantity, and date of the same record do not match between the degree application form and the defense resolution. Inspect the following: the three materials are reported separately, without field-level alignment and version synchronization.
- Symptom: the thesis review form lacks reviewer signatures or receipts. Inspect the following: the reporting step did not verify signature and receipt completeness.
- Symptom: duplicate entries appear in retrieval results, or a data row has content but cannot be retrieved. Inspect the following: the index fields did not record all vector identifiers, and after switching models the rebuild task did not cover data that had already been indexed.
How to verify the setup
This feature's acceptance: sample one data row and look at the index fields, confirm that multiple vector identifiers are complete and the status is indexed, then run a retrieval once to see whether the same group is merged and whether the highest score is taken.
Business test: change one piece of content, see whether the version number and citations are updated in sync, record the array count separately, and do not compare the tokens of the cited text with the quoteMaxToken budget for equality.
- In the knowledge base settings, check that the index model matches the vector model actually used by the knowledge base, and review the data range covered by the rebuild task records.
- Check the values of
limit, andquoteMaxTokenin splitting and citation budgeting, and confirm that the citation count and the tokens of the cited text are recorded separately.