Multi-Turn Conversation and Prompts for School Moral Education and Class Teacher Work: Trace Exceptions and Review Results

Pastoral Education: review conversation for exception review. Check source versions, input mappings and the values received by downstream nodes.

What this category's data looks like

This page addresses school moral education and homeroom teacher work scenarios; the material source is the daily class management process, and the update cadence advances weekly or by activity milestones. When the three types of materials—class meeting plans, student comprehensive quality records, and home visit records—enter the anomaly tracing and manual review function, their roles differ: class meeting plans provide activity arrangements and sign-in evidence, student comprehensive quality records provide evaluation items and a timeline, and home visit records provide communication receipts and signatures. Field characteristics center on names, quantities, dates, signatures, and receipts; common problems are that the same record's name, quantity, or date does not match across two materials, versions are not synchronized, or signatures or receipts are missing. Multi-turn dialogue and prompts are used to help sort out the historical message scope, the answer constraints are generated by the model based on this, and the anomaly list and review conclusions require manual confirmation.

Configuration and implementation checks

This step is executed by the memory turns and prompt configuration in the application settings, outputting the referenced historical message scope and answer constraints. The anomaly list and review conclusions are then completed by the model, while the external entry point and authorization are handled by the publishing service and file permissions.

ObjectCheckVerification action
Class meeting planWhether the name, date, and signature are consistent with the comprehensive quality recordCheck the activity name, implementation date, and homeroom teacher signature in the plan, compare them item by item with the comprehensive quality record, and run two consecutive follow-up questions to check whether messages and chatId carry forward the previous reference
Student comprehensive quality recordWhether the quantity, date, and version are synchronized with the home visit recordCheck the item count, evaluation date, and version number in the record, compare them with the home visit record receipt, and check whether variables are correctly brought in with the turns
Home visit recordWhether the receipt, signature, and date correspond to the class meeting planCheck the home visit receipt signature and communication date, compare them with the class meeting plan timeline, and confirm there are no missing signatures or receipts
Anomaly list and review conclusionsWhether there is support from the original record and a reviewer signatureExtract one anomaly record, trace it back to the source material, confirm the name, quantity, and date are consistent, and attach the reviewer signature

Three common errors

  • Phenomenon: The same record's name, quantity, or date does not match across two materials. Cause: The materials are updated separately, and no cross-check was performed.
  • Phenomenon: Versions are not synchronized, and an old version of the plan coexists with a new version of the record. Cause: The update cadence is inconsistent, and there is no unified version marker.
  • Phenomenon: Signatures or receipts are missing, and the review conclusion has no basis. Cause: Receipts were not collected, and the signature step was omitted.

How to verify the setup

This feature's acceptance: ask two consecutive follow-up questions, see whether the second question carries forward the reference from the first, and check whether messages and chatId are correctly brought in according to the memory turns.

Business test: create one anomaly record, see whether it can be traced back to the source material, and confirm that the name, quantity, date, and signature receipt are consistent.

  • In the application settings, check the current value of memory turns and confirm it matches the prior context range required by the question.
  • Run a trial with two consecutive question sentences, compare the usage records of the two runs, and check whether variables are correctly brought in with the turns.