What this category's data looks like
This page addresses the kindergarten health care and meal management scenario. Materials cover the kindergarten's own recipes, procurement documents, retention sample records, disinfection records, absence cause registration forms, etc., with sources from the health room, canteen, classes, and logistics, updated daily or weekly. Field characteristics include ingredient names and specifications, retention sample time and batch, disinfection time period, and absence cause classification. When entering the parsing and chunking function, recipes and procurement documents play the role of category comparison, retention sample records play the role of batch back-reference, disinfection records play the role of time period verification, and absence registration plays the role of classification back-reference. After parsing, each material forms basis entries that can be back-referenced to the original text, for citation when replying to parents or writing into plans. Material versions must be the current effective version, entries must correspond to the original text, and cross-material questions must provide a back-referenceable source.
Configuration and implementation checks
This step is executed by the parsing and training task (server-side task), outputting data rows and index status (parsed, indexing, indexed). The basis entries that can be back-referenced to the original text are then completed by retrieval and subsequent processing. After parsing is completed, parsing and chunking immediately persist to the database as visible data rows and bind a unique data identifier; the index transitions through three states; data pending indexing does not participate in recall; changing the model only rebuilds data whose indexing has been completed.
| Object | Check | Verification action |
|---|---|---|
| Ingredient names and specifications | Recipes and procurement categories are consistent, and documents match retention sample batches | Check procurement documents and retention sample records item by item against the recipes, and check whether entries can be back-referenced to the original text after chunkSize chunking |
| Retention sample time and batch | Retention sample time and batch can correspond between documents and retention sample records | Check file records and data rows separately, confirm the batch field is not misaligned, and check the processing mode corresponding to trainingType |
| Disinfection time period | Disinfection record time periods are continuous with no missing periods | Check disinfection record time periods by date, and confirm the time period field is complete after chunkSplitter splitting |
| Absence cause classification | Absence cause classification is consistent with the registration form fields | Check the classification field of the absence registration form, and confirm the data row status is parsed, indexing, or indexed and is consistent with the interface |
Three common errors
- Symptom: Materials just uploaded cannot be found. Cause: Successful parsing does not mean retrieval will definitely hit; the data row may still be in parsed or indexing, and data pending indexing is naturally isolated at the retrieval layer and is not recalled.
- Symptom: Documents and retention sample batches do not match. Cause: Retention sample records and procurement documents belong to different materials, the batch fields are not aligned, and the back-reference to the original text points to the wrong entry.
- Symptom: Disinfection records are missing time periods. Cause: Disinfection records are updated daily but the time period field is cut off during chunking, or the material version is not the current effective version, resulting in gaps in time period verification.
How to verify the setup
This function's acceptance: Upload one material, observe the data row status transitions and the coverage scope of the rebuild task, and confirm the data row goes from parsed to indexing to indexed, and that the rebuild only covers data whose indexing has been completed.
Business test: Ask a cross-material basis question, such as whether a certain ingredient in the recipes is consistent with the procurement documents and retention sample batch, see whether the answer provides a back-referenceable source, and check the number of cited entries and the tokens of the cited text.
- Open the data row to check the status, confirm it is in one of parsed, indexing, or indexed and is consistent with the interface display, and check the file records and data row identifiers separately.
- View the data rows covered by the rebuild task, confirm it only targets data whose indexing has been completed; if training fails, directly edit or delete that data row to avoid leaving an empty data row.