Knowledge Base Retrieval for Kindergarten Healthcare and Meal Management: Coordinate Operations and Application Handover

Child Health and Meals: review retrieval for operational handover. Check source versions, input mappings and the values received by downstream nodes.

What this category's data looks like

Kindergarten health care and meal management materials include ingredient names and specifications, procurement documents, recipes, retention sample records, disinfection time-slot schedules, and absence cause classification tables. These materials are updated daily or per shift, and their fields include batch, time, specification, and cause category. After entering the knowledge base retrieval and recall function, they respectively become candidate chunk sources: ingredients and specifications support procurement verification, retention sample time and batch support batch traceability, disinfection time slots support time-slot completeness checks, and absence cause classification supports classification consistency judgment. In the retrieval node, the materials serve only as text to be recalled and do not directly generate statistical tables or business documents.

Configuration and implementation checks

This step is executed by the knowledge base retrieval service (retrieval node and search test), outputting candidate chunks and similarity scores (after mode, merge, rerank, filter, and limit trimming). The handover-ready deployment status and operation records are completed by subsequent nodes or manually.

ObjectCheckVerification action
Ingredient names and specificationsWhen the recipe and procurement category are inconsistent, can the candidate chunks hit both the recipe and the procurement entry at the same timeIn the search test, switch searchMode, compare semantic and full-text candidates, and record the number of candidates and scores
Retention sample time and batchWhen the document and retention sample batch do not match, do the candidates cover both batch fieldsTest once each with synonymous questions, check the batch and time in the candidate chunks, and record the number of candidates and scores
Disinfection time slotsWhen the disinfection record is missing a time slot, do the candidates include a complete time-slot schedule chunkCompare the ranking of the top few results before and after enabling usingReRank, check time-slot completeness, and record the number of candidates and scores
Absence cause classificationWhen the cause classification and absence record are inconsistent, do the candidates recall both the classification table and the record at the same timeAdjust similarity and limit, inspect the tokens of the cited text after filtering and limit trimming, and record the array length separately

Three common errors

  • Symptom: the document and retention sample batch do not match. Cause: the retrieval candidates did not hit the batch fields at the same time. When the source does not match, first check hits and associations; retrieval returns candidates and cited content, not statistical results.
  • Symptom: the recipe and procurement category are inconsistent. Cause: semantic recall and full-text recall results were not merged, or after reranking the top few rankings deviated from the material form, causing candidates to miss key categories.
  • Symptom: the disinfection record is missing a time slot. Cause: after similarity filtering and citation limit trimming, the retained chunks are incomplete, or the search mode does not match the material form, so the candidates do not cover the complete time-slot schedule.

How to verify the setup

This function's acceptance: test once each with synonymously rewritten questions, compare candidates and scores, then toggle reranking to inspect the ranking of the top few results; switch searchMode, adjust similarity and limit, and record the number of candidates and scores.

Business test: check item by item against the handover list once, and see which record is missing; check whether version, connection, and task records are consistent. The citation count is the array length, and the tokens of the cited text are recorded separately.

  • In the knowledge base search test, switch searchMode, test once each with synonymous questions, record the number of candidates and scores, and toggle usingReRank to compare the ranking of the top few results.
  • Adjust similarity and limit, inspect the tokens of the cited text after filtering and limit trimming, record the array length separately, and then check version, connection, and task records item by item against the handover list.