What this category's data looks like
This page addresses kindergarten health care and meal management; the materials come from the kindergarten's health room, cafeteria, and class daily records. Division of roles: ingredient acceptance sheets and purchase orders provide names, specifications, and batches; retention-sample records provide retention time and batch; disinfection records provide disinfection time periods and the person who performed them; menus provide meal times and categories; absence registration provides cause classification. The update rhythm is daily recording with the class, weekly aggregation, and monthly archiving, with versions bounded by the current semester. When the materials enter this function, each serves as an independent data row, an index is generated by the vector model, the original data is stored in the business database, and multiple vector identifiers are recorded in the index field; during retrieval, vectors are recalled first and then merged back with the original data. Field characteristics include date, class, ingredient name and specification, batch number, time period, and cause category. These fields determine whether cross-material basis queries can match the original text.
Configuration and implementation checks
This step is executed by the index service and the vector database, outputting vectors and index status (whether recall is possible and 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 take only supporting materials).
| Object | Check | Verification action |
|---|---|---|
| Ingredient acceptance sheets and purchase orders | Whether the ingredient name and specification match the menu and the retention-sample batch | Open one acceptance sheet data row, check whether the index field contains multiple vector identifiers, then search the ingredient name to see whether the same group is merged |
| Retention-sample records | Whether the retention time and batch are consistent with that day's menu and purchased categories | Search the retention-sample batch, check whether the recall results trace back to the same original text, and view the data range covered by the rebuild task record |
| Disinfection records | Whether the disinfection time periods are complete and whether any time period is missing | Search disinfection time periods by class and date, check the data row index field and the recalled content, and confirm that data in the same group is merged by taking the highest score |
| Absence registration | Whether the absence cause classification is consistent with the health room records | Search the cause category, check whether the index model and the vector model actually used by the knowledge base are consistent, and view the data row index field |
Three common errors
- Symptom: the document and the retention-sample batch do not match. Inspect the following: the acceptance sheet batch number and the retention-sample record batch number were not generated according to the same rule, when merging back with the original data they cannot be grouped into the same group.
- Symptom: the menu and the purchased categories are inconsistent. Inspect the following: the menu category and the purchased item name have different granularity.
- Symptom: disinfection records are missing time periods. Inspect the following: disinfection time periods were not entered segment by segment, the data row content is incomplete.
How to verify the setup
Acceptance of this function: take one data row and look at the index field, confirm that multiple vector identifiers are all present and the status is indexed; then search once to see whether the same group is merged, and view the coverage of the rebuild task record.
Business test: ask a cross-material basis question, for example whether a certain class's retention-sample batch on a certain day is consistent with that day's menu and disinfection time periods, and see whether the answer gives a traceable source.
- In the knowledge base settings, check that the index model and the vector model actually used by the knowledge base are consistent, and open one data row to check the index field, confirming that the values of
defaultTokenandmaxTokendo not affect the recording of the index field. - View the data range covered by the rebuild task record, confirm that after changing the model it is triggered only for data that has already completed indexing, and record the number of citations and the tokens of the citation text.