Context and Token Budgets for Kindergarten Healthcare and Meal Management: Draft Materials and Hand Over Responsibilities

Child Health and Meals: review context & tokens for material handover. Check source versions, input mappings and the values received by downstream nodes.

What this category's data looks like

This page addresses the kindergarten health care and meal management scenario. Materials come from the daily records of the health room and the cafeteria: ingredient names and specifications, retention sample times and batches, disinfection periods, categorization of absence causes, and lists of recipes and procurement categories. These materials are updated daily or weekly, require complete fields, consistent responsible persons and dates, and flow along with handover records. When entering this function, ingredient and retention sample materials take on the role of batch verification, disinfection records take on the role of period verification, recipe and procurement lists take on the role of category consistency, and absence categorization takes on the role of cause statistics. Materials are first drafted into a signable draft according to the template, then handed over to the next stage for sign-off along with the handover record; missing items are blocked at the two nodes of drafting and handover.

Configuration and implementation checks

This part is executed by application settings (response limit, etc.), which output the length of a single answer and the range of context that can be accommodated. "Signable material drafts and handover records" are then completed by the model. The response limit uses a slider to control the length of a single answer; if set too low, the answer will be truncated, and if set too high, consumption will increase.

ObjectCheckVerification action
Ingredient names and specificationsCorrespond item by item with the recipe and procurement categories, and the specification wording is consistentCompare against the recipe and procurement list, and check ingredient names, specifications, and batches
Retention sample times and batchesThe retention sample batch on the document matches the retention sample recordSample one document and check whether the retention sample time, batch, and record are consistent
Disinfection periodsThe disinfection record covers every period that should be disinfected, with no gapsCheck the disinfection record against the period table and find the locations of missing periods
Categorization of absence causesThe categorization criteria for causes are unified, and the responsible person and date are consistentCheck the categorization field of the absence registration form and the signing date of the handover record

Three common errors

  • Phenomenon: The document and the retention sample batch do not match. Cause: The batch was not filled in according to the retention sample record during drafting, and item-by-item verification was not performed before handover.
  • Phenomenon: The recipe and procurement categories are inconsistent. Cause: After the recipe was updated, the procurement list was not synchronized, and the category names and specification wording were not unified.
  • Phenomenon: The disinfection record is missing periods. Cause: During handover, only the total number of times was checked, and each period was not confirmed segment by segment according to the period table, so gaps were missed.

How to verify the setup

This function's acceptance: ask a question that requires a long answer and see whether the end of the answer is complete; then check whether the settings of maxResponse and maxContext match the answer length required by the page. quoteMaxToken only checks the tokens of the quoted text, and the number of quotes is recorded separately.

Business test: draft one according to the template and complete one handover, and see whether missing items are blocked.

  • In the application settings, check the response limit switch and slider position, and ask a question that requires a long answer to see whether the end is complete
  • Compare the usage of two runs to confirm the change in single-run usage after adjusting the limit, and check the values of maxResponse, maxContext, and quoteMaxToken