What this category's data looks like
This page addresses the kindergarten health care and meal management scenario. The materials come from daily health care and meal documents, and the update cadence advances daily or per meal. Ingredient names and specifications, sample retention times and batches, disinfection periods, and absence cause classification are the four core fields. When ingredient names and specifications enter authorized access and external services, they determine the recipe entries visible to parents; when sample retention times and batches enter multi-turn dialogue, they are used to check whether documents and sample retention batches match; when disinfection periods enter external services, they are used to respond to period coverage questions from higher-level inspections; when absence cause classification enters prompt constraints, it is used to limit answers to only cite already classified cause categories. After materials enter this feature, each assumes the role of being retrieved, questioned, and constrained, and does not directly generate access entry points.
Configuration and implementation checks
This part is executed by the memory turns and prompt configuration in application settings, outputting the referenced historical message range and answer constraints. The externally available entry points and authorization scope are determined by the publishing service and file permissions. The model only processes filtered materials, while the publishing service and file permissions are responsible for mapping entry points and visibility scope to specific identities. Memory turns control how many previous dialogue turns can be referenced at most when answering. If the value is too small, prior context cannot be used; if too large, irrelevant content is brought in and consumption increases.
| Object | Check | Verification action |
|---|---|---|
| Memory turns | Matches the prior context range required by the question | Check the current value in application settings, and run a trial with two consecutive questions to observe whether the second question carries forward the reference from the first question |
| Sample retention times and batches | Documents and sample retention batches match | Extract one sample retention document, check whether the batch number and sample retention time are consistent with the record, then put the same batch into multi-turn dialogue for follow-up questioning |
| Disinfection periods | Disinfection records have no missing periods | Check records period by period according to disinfection periods, confirm that each period has a corresponding entry, then access once with a restricted identity to see the visible scope |
| Absence cause classification | Classification categories are consistent with external answers | Check the absence cause classification field, confirm that answer constraints only cite already classified categories, and inspect the values of messages and variables |
Three common errors
- Phenomenon: Documents and sample retention batches do not match. Cause: Sample retention times and batches were not updated in sync with the documents, and multi-turn dialogue answered by citing an old batch.
- Phenomenon: Recipes and procurement categories are inconsistent. Cause: Ingredient names and specifications were not filtered according to the same categories in authorized access and external services.
- Phenomenon: Disinfection records are missing periods. Cause: The disinfection period field was not completed before entering prompt constraints, so answers could not cover the missing periods.
How to verify the setup
Acceptance for this feature: ask two consecutive follow-up questions, see whether the second question carries forward the reference from the first question, and check whether the historical message range referenced in messages conforms to the memory turns value.
Business test: access once with a restricted identity, see whether the visible scope conforms to the settings, and confirm that unauthorized content is not visible.
- Check the memory turns value in application settings, run a trial with two consecutive questions, and record the historical message range corresponding to
chatId. - Compare the usage records of two runs, confirm that consumption changes accordingly after adjusting the turns, and separately record the number of entries in the citation array.