What this category's data looks like
This page's category is primary and secondary school teacher training credit management. The source materials include training notices, credit recognition forms, and attendance rosters, and the update cadence advances with training batches and recognition cycles. Training notices provide the project name, start and end dates, and required credit hours; credit recognition forms carry teachers' individual credit declarations and review signatures; attendance rosters record the personnel who actually attended and the attendance dates. When these three types of materials enter the Q&A intake and missing-item follow-up function, they respectively take on the roles of rule basis, declaration evidence, and attendance fact: notices are used to determine whether credits fall within the recognizable scope, recognition forms are used to check declaration fields and signatures, and rosters are used to verify consistency of personnel, quantities, and dates. Field characteristics are concentrated in four categories: names, quantities, dates, and signatures. The same record often shows inconsistent name spellings across different materials, mismatched headcounts or credit-hour quantities, dates not synchronized across versions, and missing signatures or receipts, so the intake step must first identify missing items and then ask follow-up questions to complete them.
Configuration and implementation checks
This part is executed by the memory turn count and prompt configuration in the app settings, and outputs the historical message range referenced and the answer constraints. The "intake record and items to complete" is then completed by the model. The memory turn count controls how many previous turns of dialogue can be referenced at most when answering; the value only needs to match the preceding context range required by the question, and in continuous follow-up scenarios the second turn should be able to correctly carry forward the references from the first turn.
| Object | Check | Verification action |
|---|---|---|
| Memory turn count | Matches the preceding context range required by the question; in continuous follow-up, the second turn can carry forward the references from the first turn | Check the current value in the app settings, run a trial with two consecutive turns of questions, and observe whether the historical range brought into messages covers the preceding context required for the references |
| Training notice | The name, dates, and credit-hour quantity match the credit recognition form and attendance roster, and versions are synchronized | Compare item by item the project name, start and end dates, and required credit hours in the notice against the corresponding fields in the recognition form and roster, and check whether signatures or receipts are missing |
| Credit recognition form | Declaration fields and signatures are complete, and the name, quantity, and date of the same record are consistent between the two materials | Check the name, credit quantity, and recognition date of the same teacher in the recognition form and attendance roster, and check whether the signature column and receipt are complete |
| Attendance roster | The attending personnel, quantity, and dates are consistent with the notice and recognition form, with no unsynchronized versions | Compare the batch dates and headcount in the roster against the notice, compare with the declaration records in the recognition form, and confirm that the material version referenced in variables is the latest |
Three common errors
- Symptom: the second question uses a reference but the answer misses the point. Cause: the memory turn count value is too small, the preceding context is not brought in, and the model cannot carry forward the referenced object from the previous turn.
- Symptom: the name, quantity, or date of the same record does not match between the training notice and the attendance roster. Cause: the two material versions are not synchronized, and at intake no follow-up was first asked about which one prevails.
- Symptom: the credit recognition form lacks a signature or receipt, but intake answers directly. Cause: the missing item was not identified and followed up, and the model gives a conclusion based on incomplete materials.
How to verify the setup
This function's acceptance: with two consecutive follow-up turns, see whether the second question carries forward the references from the first question, and check whether the historical range brought into messages is consistent with the conversation context corresponding to chatId.
Business test: ask a question that lacks key information and see whether it first follows up before answering
- Check the memory turn count value in the app settings, run a trial with two consecutive turns of questions, and confirm that the second question can carry forward the references from the first question
- Compare the usage records of two runs, and confirm that after adjusting the turn count the consumption changes accordingly and no irrelevant content is brought in