What this category's data looks like
This page addresses the business basis query for vocational training subsidy applications. Materials include training class filing records, sign-in records, completion assessment forms, and subsidy application materials retained by the park. Class filing records provide training program, institution, and date fields; sign-in records provide trainee name, sign-in date, and signature fields; completion assessment forms provide assessment results, date, and signature fields; application materials provide subsidy recipient, amount, and receipt. Materials enter the knowledge base by batch, with an update cadence aligned with the subsidy application cycle, and the current materials prevail as the version. When entering the vector model and indexing function, class filing records, sign-in records, and completion assessment forms each generate raw data rows, and the index field records the corresponding vector identifiers to support subsequent retrieval and citation trimming of supporting materials.
Configuration and implementation checks
This step is executed by the indexing service and vector store, outputting vectors and index status (whether recall is possible, whether data in the same group is merged). The 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 is consistent with the vector model actually used by the knowledge base, and the index field of a data row records multiple vector identifiers for one piece of data.
| Object | Check | Verification action |
|---|---|---|
| Training class filing record | Whether the fields, dates, and signatures of the class filing record match the completion assessment form | Check the training program, institution, date, and signature in the class filing record, then compare item by item with the completion assessment form |
| Sign-in record | Whether the trainee name, sign-in date, and signature in the sign-in record are consistent with the class filing record | Open the sign-in record data row, check the name, date, and signature field, then compare with the class filing record |
| Completion assessment form | Whether the assessment results, dates, and signatures of the completion assessment form are consistent with the sign-in record | Check the assessment results, date, and signature of the completion assessment form, then compare with the sign-in record |
| Index field of the data row | Whether multiple vector identifiers for one piece of data are all present and the status is indexed | Open a data row to check the index field, and view the data range covered by the rebuild task record |
Three common errors
- Phenomenon: The same trainee name is written differently in the sign-in record and the completion assessment form. Inspect the following: Names were not unified during material entry, and the lack of a receipt or signature makes it impossible to trace back to the original text.
- Phenomenon: The dates for the same training program do not match between the class filing record and the sign-in record. Inspect the following: Versions are not synchronized, and old materials remain in the knowledge base and participate in retrieval.
- Phenomenon: Duplicate entries appear in retrieval results. Inspect the following: Data in the same group is not merged, or after changing models, rebuild is triggered only for data that has already been indexed, reducing recall quality.
How to verify the setup
This function's acceptance check: sample a data row and inspect the index field to confirm that multiple vector identifiers are all present and the status is indexed, then run a retrieval once to see whether the same group is merged; check the data range covered by the rebuild task record, and inspect the values of limit, and indexes. The number of citations is the array length, and quoteMaxToken is the citation content budget; the two are not compared for equality. If a citation upper limit is involved, check the tokens of the cited text, and record the array count separately.
Business test: ask a basis question that spans materials, and see whether the answer gives a traceable source.
- In the knowledge base settings, check the indexing model and confirm it is consistent with the vector model actually used by the knowledge base.
- Open a data row to check the index field, confirm that multiple vector identifiers are all present, and view the data range covered by the rebuild task record.