What this category's data looks like
For the pre-departure training and overseas safety reminders category, materials include pre-departure training courseware, safety reminder checklists, and accommodation contact forms. Courseware is updated by semester, with fields including course name, class hours, date, and instructor signature; safety reminder checklists are updated by destination country, with fields including risk item, response steps, version number, and receipt signature; accommodation contact forms are updated by batch, with fields including accommodation name, address, contact person, check-in date, and confirmation signature. When materials enter this feature, courseware and checklists generate text indexes, accommodation contact forms generate indexes by row, and image attachments can generate image description indexes if the model supports it. The same record may appear in two materials, and the name, quantity, and date must be consistent, the version must be synchronized, and signatures or receipts must not be missing. The indexing service and vector database are responsible for recall and merging, and subsequent nodes trim citations accordingly, providing basis entries that can be traced back to the original text when replying to parents or writing into a plan.
Configuration and implementation checks
This step is executed by the indexing 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). The indexing model must be consistent with the vector model actually used by the knowledge base. The index field of a data row records multiple vector identifiers. After changing the model, rebuild is triggered only for data that has already been indexed.
| Object | Check | Verification action |
|---|---|---|
| Pre-departure training courseware | Fields include course name, class hours, date, and instructor signature, and the name, quantity, and date for the same record are consistent with the safety reminder checklist and accommodation contact form | Open a courseware data row, check whether the index field contains multiple vector identifiers, whether the status is indexed, and run a retrieval to see whether the same group is merged |
| Safety reminder checklist | Fields include risk item, response steps, version number, and receipt signature, and the version is synchronized with the current materials | In the knowledge base settings, check that the indexing model is consistent with the vector model actually used by the knowledge base, and view the data range covered by the rebuild task record |
| Accommodation contact form | Fields include accommodation name, address, contact person, check-in date, and confirmation signature, and the name and date for the same record are consistent with the courseware and checklist | Retrieve an accommodation record to see whether the recall results are merged back to the original data by vector identifier, whether the highest score is taken for the same group, and whether the array count is recorded separately |
| Data row index field | Multiple vector identifiers for one piece of data are all present and can be traced back to the original text | Check the budget of limit during citation trimming; check the tokens of the cited text separately, and record the array count separately |
Three common errors
- Symptom: The name, quantity, and date of the same record do not match between the courseware and the checklist. Inspect the following: The two materials are updated separately, without field alignment and version synchronization.
- Symptom: The version of the safety reminder checklist is not synchronized, and the receipt signature is missing. Inspect the following: compare the relevant source fields, document versions and downstream processing records.
- Symptom: Duplicate entries appear in retrieval results. Inspect the following: Data in the same group corresponds to multiple vector identifiers, the highest score is not taken during merging, and the array count is not recorded separately.
How to verify the setup
This feature's acceptance: sample a data row and check the index field to confirm that multiple vector identifiers are all present and the status is indexed, then run a retrieval to see whether the same group is merged and whether the highest score is taken, and view the data range covered by the rebuild task record.
Business test: ask a basis question across materials and see whether the answer provides a traceable source, and whether the source matches the name, date, and signature in the original text.
- In the knowledge base settings, check that the indexing model is consistent with the vector model actually used by the knowledge base; after changing the model, check that the rebuild task record covers only data that has already been indexed.
- Open a data row and check the index field to confirm that multiple vector identifiers are all present; after retrieval, check whether the same group is merged and whether the highest score is taken, and separately record the array count and the tokens of the cited text.