What this category's data looks like
In the compulsory education dropout prevention and control scenario, the materials entering this function are mainly three types: dropout risk ledgers, return-to-school persuasion records, and home visit records. Dropout risk ledgers are updated by semester, with fields including student name, class, risk level, determination date, and responsible person's signature; return-to-school persuasion records are generated per instance, with fields including persuasion time, participants, persuasion method, and parent receipt; home visit records are generated per instance, with fields including home visit date, home visit teacher, interview key points, and parent signature. The three types of materials corroborate each other in business terms: the ledger provides the risk list and levels, the return-to-school persuasion records explain the handling process, and the home visit records reflect the family's cooperation. When materials enter this function, each assumes a different role: the ledger serves as the primary basis for the list and levels, while return-to-school persuasion records and home visit records serve as process-based corroboration, used to answer why a certain student was determined to be at risk and whether return-to-school persuasion and home visits have been implemented. In terms of update rhythm, the ledger rolls forward with semesters and monthly screenings, while return-to-school persuasion and home visit records are supplemented immediately as events occur, so the names, quantities, and dates of the same record in the two materials are prone to mismatch, and situations such as unsynchronized versions, missing signatures, or missing receipts are also common.
Configuration and implementation checks
This step 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). 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 index model is consistent with the vector model actually used by the knowledge base, and the index field of a data row records the multiple vector identifiers corresponding to one piece of data.
| Object | Check | Verification action |
|---|---|---|
| Dropout risk ledger | The Check column is written to the category verification dimensions: whether student name, class, risk level, determination date, and responsible person's signature are complete, and whether the names, quantities, and dates of the same record in the two materials match | Open one ledger data row and check whether the index field contains multiple vector identifiers; then search for that student's name and see whether data in the same group is merged into one entry |
| Return-to-school persuasion record | The Check column is written to the category verification dimensions: whether persuasion time, participants, persuasion method, and parent receipt are complete, whether the version is current, and whether the receipt is missing a signature | Sample one return-to-school persuasion record and check whether the data row status is indexed; after changing the index model, view the data range covered by the rebuild task record |
| Home visit record | The Check column is written to the category verification dimensions: whether home visit date, home visit teacher, interview key points, and parent signature are complete, and whether the dates are consistent with the ledger and return-to-school persuasion records | Search for one student's home visit record and check whether the recall result provides a source that can be traced back to the original text; check whether the tokens of the cited text under limit are trimmed |
| Knowledge base settings | The Check column is written to the category verification dimensions: whether the index model is consistent with the vector model actually used by the knowledge base, and whether the rebuild task is triggered only for data that has already completed indexing | Check the index model in the knowledge base settings; view the rebuild task record and confirm whether the data range covered after changing the model is consistent with the index field of the data row |
Three common errors
- Symptom: duplicate entries appear in retrieval results. Inspect the following: the same group of data corresponds to multiple vector identifiers, and the merge step does not merge by the highest score, causing the same record to be recalled multiple times.
- Symptom: the names, quantities, and dates of the same record in the two materials do not match. Inspect the following: the ledger and return-to-school persuasion and home visit records are supplemented separately without version synchronization, and records missing signatures or receipts are still treated as valid basis.
- Symptom: after changing the model, recall quality declines, and data rows have content but cannot be retrieved. Inspect the following: the rebuild task record only covers data that has already completed indexing; data rows that have not completed indexing do not participate in recall, so indexing must be completed before rebuilding.
How to verify the setup
Acceptance of this function: sample one data row and look at the index field to confirm that multiple vector identifiers for one piece of data are all present; then search once and see whether data in the same group is merged and whether the score takes the highest value; for status values, only recognize indexed and the rebuild task record.
Business test: ask a basis question across materials, for example whether a certain student's risk level this semester is consistent with the return-to-school persuasion and home visit records, and see whether the answer provides a source that can be traced back to the original text, and check the tokens of the cited text and the number of array entries and record them separately.
- In the knowledge base settings, check that the index model is consistent with the vector model actually used by the knowledge base, then open one data row and check the index field to confirm that multiple vector identifiers are all present, and view the data range covered by the rebuild task record.
- Sample one each from the dropout risk ledger, return-to-school persuasion records, and home visit records, check the fields, dates, and signatures, and after retrieval confirm that data in the same group is merged into one entry and can be traced back to the original text; for the citation limit, only check the tokens of the cited text, record the number of array entries separately, and set
limitaccording to the actual budget.