What this category's data looks like
For kindergarten health care and meal management scenarios, materials cover quantified recipes, food ingredient purchase orders, warehousing and sample retention records, disinfection time schedules, absence cause registration, and health room shift handover records. Sources are mostly daily submissions by health care workers, canteen administrators, and class teachers, with update rhythms overlapping daily, weekly, and monthly: recipes roll weekly, purchases and sample retention change by batch, and disinfection and absences accumulate by shift and date. Field characteristics center on ingredient names and specifications, sample retention time and batch, disinfection time periods, and absence cause classification. After materials enter document parsing and chunking, each plays a different role: recipes and purchase orders are used to align categories and quantities, sample retention and disinfection records are used to lock time and batch, and absence registration is used to unify the criteria for cause classification.
Configuration and implementation checks
This step is executed by parsing and training tasks (server-side tasks), outputting data rows and index status (parsed, indexing, indexed). The "version ledger and change records" are then completed by retrieval and subsequent processing.
| Object | Check | Verification action |
|---|---|---|
| Food ingredient purchase orders and quantified recipes | Whether ingredient names and specifications, and recipe and purchase categories are aligned item by item | Extract both materials for the same week, check whether names, specifications, and quantities are consistent, and record the data rows where differences occur |
| Sample retention records and warehousing documents | Whether sample retention time and batch can correspond one-to-one with warehousing batches | Extract sample retention records and warehousing documents by batch number, check whether time and batch match, and observe whether data row status updates with parsing |
| Disinfection time schedules and class handover records | Whether disinfection time periods cover all key nodes throughout the day and whether any time periods are missing | Extract disinfection time schedules by date, check segment by segment whether any time periods are missing, and check whether the time period descriptions in handover records are synchronized |
| Absence cause registration and health room summary | Whether the criteria for absence cause classification are unified and whether old versions can be traced | Extract registration and summary for the same week, check whether classification names are consistent, and check whether old versions are still traceable |
Three common errors
- Phenomenon: Documents and sample retention batches do not match, and retrieval hits old batches. Cause: After parsing is persisted, batch fields were not checked, and the index status was still indexing when misused.
- Phenomenon: Recipes and purchase categories are inconsistent, and referenced content contradicts each other. Cause: After the two materials were chunked, category alignment was not performed, and the version ledger did not record change points.
- Phenomenon: Disinfection records are missing time periods, and retrieval results are incomplete. Cause: Chunking split by fixed granularity, time period boundaries were cut off, and differences between parsed and indexed were not checked.
How to verify the setup
Acceptance for this feature: Upload one material, see whether data row status transitions through parsed, indexing, and indexed, and check whether the rebuild task only covers data rows with completed indexing.
Business test: Change one piece of content, see whether the version number and references are updated synchronously, and check whether the old version is still traceable.
- After checking
chunkSizeandchunkSplittersettings, check whether chunk boundaries cut off ingredient specifications, sample retention batches, and disinfection time periods, and record data row status. - Check the training mode corresponding to
trainingType, confirm that the rebuild scope only covers data rows with completed indexing, and check whether the tokens of referenced text and the number of references are recorded separately.