What this category's data looks like
In the student disciplinary and appeal handling scenario, the materials entering the Q&A intake stage are mainly of three types: disciplinary fact materials (including time, place, description of conduct, involved persons, and names and quantities of items), the disciplinary decision (including the type of disciplinary action, the clauses relied upon, the effective date, signatures/seals, and delivery receipt), and the appeal statement (including the appeal request, reasons, evidence list, submission date, and signature). These materials are usually updated by semester or event milestone; for the same event, fact materials may come first, then the decision, and then the appeal statement, with versions rolling as review or supplementary investigation proceeds. When the materials enter the Q&A intake function, the fact materials and the decision are used to identify missing items and verify fields, while the appeal statement is used to confirm the claims and what the evidence points to; together, the three support the generation of the intake record and the items to be completed. Field characteristics center on whether names, quantities, dates, signatures, and receipts are consistent; if the same record does not match across two materials, this must be exposed in the follow-up questions and the user guided to correct it.
Configuration and implementation checks
This part is executed by the memory turn count and prompt configuration in the app settings, and outputs the reference range of historical messages and the answer constraints. The intake record and the items to be completed are then produced by the model. The memory turn count controls how many previous turns of dialogue can be referenced at most when answering; if the value is too small, the preceding text cannot be used, and if the value is too large, irrelevant content is brought in and consumption increases.
| Object | Check | Verification action |
|---|---|---|
| Disciplinary fact materials | Check against the disciplinary decision whether the names, quantities, and dates of the same record are consistent | Check the memory turn count in the app settings, run a trial with two consecutive questions, and check whether the second turn picks up the references from the first turn; inspect the historical range corresponding to messages and chatId |
| Disciplinary decision | Check against the disciplinary fact materials the type of disciplinary action, the clauses relied upon, the effective date, signatures, and receipt | Compare the usage records of two runs to confirm the change in consumption after adjusting the turn count; check whether variables bring in irrelevant preceding text |
| Appeal statement | Check against the disciplinary decision the appeal request, reasons, evidence list, submission date, and signature | Run a trial with two consecutive questions, with the first question missing key information, and observe whether it asks a follow-up before answering; verify whether the historical message range in messages matches the range of preceding text required by the question |
| Intake record and items to be completed | Missing items are identified and followed up on, and the answer has a material basis | Check the current memory turn count value in the app settings, and confirm that the second turn of consecutive follow-up questions correctly picks up the references from the first turn |
Three common errors
- Symptom: The names, quantities, and dates of the same record do not match between the disciplinary fact materials and the disciplinary decision. Cause: The two materials are updated separately, and no cross-check has been done.
- Symptom: The version of the appeal statement is not synchronized, and it still cites clauses from an old version of the decision. Cause: The materials are updated at inconsistent paces, and there is no version comparison.
- Symptom: The disciplinary decision lacks a signature or receipt, but the intake record is still generated as if the materials were complete. Cause: Missing-item identification does not cover the signature and receipt fields.
How to verify the setup
Acceptance for this feature: with two consecutive follow-up questions, see whether the second question picks up the references from the first question; check the historical range corresponding to messages and chatId, and confirm that the memory turn count matches the range of preceding text required by the question.
Business test: ask a question that lacks key information, and see whether it asks a follow-up before answering.
- Check the current memory turn count value in the app settings, run a trial with two consecutive questions, and record how the second turn picks up the references.
- Compare the usage records of two runs to confirm the change in consumption after adjusting the turn count, and check whether
variablesbring in irrelevant preceding text.