What this category's data looks like
Materials for pre-departure training and overseas safety reminders come from courseware, safety reminder checklists, and accommodation contact lists; the update cadence follows semester batches and destination policy changes. The courseware contains training topics, class hours, and instructor signatures; the safety reminder checklist contains risk items, effective dates, and receipt confirmations; the accommodation contact list contains accommodation names, addresses, emergency contacts, quantities, and dates. When these three materials enter this feature, the courseware and the checklist serve as text indexing sources, while the accommodation contact list is split by row; the same record can correspond to multiple vector identifiers, making it easy to recall by name, date, and quantity separately. Field characteristics center on names, quantities, dates, and signature receipts; when versions change, the index fields must be updated in sync.
Configuration and implementation checks
This step is executed by the indexing service and the vector database, outputting vectors and index status (whether recall is possible, whether data in the same group is merged). The "version ledger and change records" are then completed by subsequent AI nodes, code nodes, or manual work (retrieval and citation trimming only take supporting materials).
| Object | Check | Verification action |
|---|---|---|
| Pre-departure training courseware | Verify whether the name, quantity, date, and signature fields are consistent with the current version, and whether the old version is traceable | Open the courseware data row and check whether the index fields contain multiple vector identifiers and whether the status is indexed |
| Safety reminder checklist | Verify whether risk items, effective dates, and receipt confirmations are in sync, and whether the old version is traceable | In the knowledge base settings, check that the indexing model matches the model actually used by the checklist, and whether retrieval merges the same group |
| Accommodation contact list | Verify whether accommodation names, addresses, emergency contacts, quantities, and dates match the courseware and checklist | Check the data scope covered by the rebuild task, and confirm that after changing the model, only data that has already been indexed is triggered |
| Data row index fields | Verify whether multiple vector identifiers for one data row are all present, and whether data in the same group is merged by taking the highest score | Sample one data row to check the index fields, then run a retrieval to see whether duplicate entries are merged |
Three common errors
- Symptom: the name, quantity, and date of the same accommodation record do not match between the courseware and the contact list. Inspect the following: the two materials were imported separately.
- Symptom: the safety reminder checklist has been changed, but the old version is still recalled; the version number and citation are out of sync, and data rows whose indexing is not complete do not participate in recall, so indexing must be completed before rebuilding.
- Symptom: the accommodation contact list has content but cannot be retrieved, or duplicate entries appear in retrieval results. Inspect the following: the index fields are missing multiple vector identifiers and data in the same group is not merged; or after changing the model, recall quality drops and the rebuild task does not cover that batch of data.
How to verify the setup
This feature's acceptance check: sample one data row to inspect the index fields, confirm that multiple vector identifiers are all present and the status is indexed; then run a retrieval once to see whether data in the same group is merged by taking the highest score. If a citation limit is involved, check the tokens of the cited text; the number of array entries is recorded separately, the number of citations is the array length, and quoteMaxToken is the citation content budget; the two are not compared for equality.
Business test: change one piece of content and see whether the version number and citation are updated in sync; check limit, and the rebuild task records to confirm that after changing the model, only data that has already been indexed is triggered.
- In the knowledge base settings, check that the indexing model matches the vector model actually used by the knowledge base, and confirm that the data row index fields are complete.
- Check the data scope covered by the rebuild task records, and confirm that after changing the model, the old version vectors have been processed, the current version is unique, the old version is traceable, and citations are in sync.