What this category's data looks like
Graduate training process management involves materials such as training plans, proposal reports, and mid-term assessment forms. The sources are departmental and supervisor archives, and the update cadence advances with semesters and assessment milestones. The training plan contains course, credit, and research direction fields; the proposal report contains title, date, and supervisor signature; the mid-term assessment form contains assessment conclusion, date, signature, and receipt. When materials enter this feature, each is written as an independent data row. The vector model and index are responsible for aligning the names, quantities, and dates of the same record across different materials into basis entries that can be traced back to the original text, for citation when replying to parents or writing into a plan.
Configuration and implementation checks
This part is executed by the index service and vector database, outputting vectors and index status (whether recall is possible, whether data in the same group is merged). "Basis entries 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).
| Object | Check | Verification action |
|---|---|---|
| Training plan | Fields, dates, and signatures are complete, and the names and quantities match those in the proposal report | Check the index model in the knowledge base settings; open a data row and check whether the index field records all vector identifiers |
| Proposal report | The version is current, the date matches the mid-term assessment form, and the supervisor signature and receipt are not missing | View the data scope covered by the rebuild task; run a retrieval to see whether data in the same group is merged and whether the status is indexed |
| Mid-term assessment form | The assessment conclusion, date, and signature correspond item by item with the training plan and proposal report | After changing the model, trigger a rebuild only for data that has already completed indexing, and check the limit, and citation count records |
| Data row | Multiple vector identifiers for one piece of data are all present, and data in the same group is merged by taking the highest score | Open a data row and check the index field to confirm the vector identifiers are complete and can be traced back to the original text |
Three common errors
- Symptom: duplicate entries appear in retrieval results. Inspect the following: the names, quantities, and dates of the same record in two materials do not match, and the vector identifiers are not merged into the same group of data.
- Symptom: recall quality drops after changing the model. Inspect the following: the version is not synchronized, the rebuild task only overwrites, data rows that have not completed indexing do not participate in recall, and indexing must be completed before rebuilding.
- Symptom: a data row has content but cannot be retrieved. Inspect the following: a missing signature or receipt causes that entry not to complete indexing, the status does not reach indexed, and the vector is not written to the vector database.
How to verify the setup
Acceptance for this feature: take one data row and look at the index field to confirm that multiple vector identifiers are all present; then run a retrieval once to see whether data in the same group is merged, whether the highest score is taken, whether the status is indexed, and whether the rebuild task records cover it.
Business test: ask a basis question that spans materials, see whether the answer gives a traceable source, and check the citation count and the tokens budget of the cited text, recording the two separately.
- In the knowledge base settings, check that the index model is consistent with the vector model actually used by the knowledge base, and open a data row to check the index field.
- View the data scope covered by the rebuild task. After changing the model, trigger it only for data that has already completed indexing, and check the
limit, and citation count records.