What this category's data looks like
Kindergarten home-school co-education communication materials mainly consist of class notices, child observation records, parent request forms, and commitment item tables. They come from teachers' daily entries and the kindergarten's unified templates, and their update cadence changes with activity milestones and children's performance in kindergarten. When the materials enter the document parsing and chunking function, each plays a different role: notice-type materials handle delivery and receipt verification, observation records serve as the factual source for descriptions of children's performance, request forms serve as the basis for aggregating parent request classifications, and commitment item tables serve as the tracking record for response times and responsible persons. Field characteristics are concentrated in four dimensions: commitment items and response times, descriptions of children's performance, parent request classifications, and notice delivery. The version ledger and change records must be registered item by item according to these dimensions.
Configuration and implementation checks
This step is executed by the parsing and training task (server-side task), which outputs 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 |
|---|---|---|
| Class notice materials | Whether notice delivery and commitment item response times are registered item by item in the version ledger | Check the file records and data row identities, confirm the status is parsed or indexing or indexed, and check whether the chunk boundaries corresponding to chunkSize and chunkSplitter are split between notice items and response times |
| Child observation records | Whether the descriptions of children's performance are consistent with the original observation records, and whether there are deviations in wording | Open the data rows to check that the status matches what is displayed in the interface, check whether the processing mode corresponding to trainingType is suitable for the observation record text, and compare the descriptions with the original text item by item |
| Parent request forms | Whether parent request classifications are assigned to the corresponding categories, and whether any feedback is omitted from classification | Check the file records and data rows separately, confirm that the request forms have been stored as visible data rows, and check the correspondence between the classification fields and the request content |
| Commitment item tables | Whether commitment items and response times appear in pairs, and whether unresponded items can be traced | Check whether the data rows covered by the rebuild task contain only data with completed indexing, and confirm that old-version commitments remain traceable |
Three common errors
- Phenomenon: commitment items have been registered but response times are missing, and parent requests have not received a response for a long time. Cause: commitments and responses belong to different materials, the version ledger does not bind response times item by item, and response records are not synchronized after old versions are replaced.
- Phenomenon: descriptions of children's performance do not match the original observation records, and teachers' paraphrasing deviates. Cause: description fields are split apart during chunking, the quoted text and the original observation records do not remain on the same version, and retrieval returns old-version fragments.
- Phenomenon: parent requests are not classified, and feedback is scattered across multiple materials and difficult to track. Cause: request forms are not stored according to classification dimensions, data row status is still parsed or indexing, the retrieval layer does not recall them, and subsequent processing cannot aggregate them.
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 that the statuses are parsed, indexing, and indexed in sequence, and that the rebuild covers only data rows with completed indexing.
Business test: change one piece of content, observe whether the version number and references are updated synchronously, check that the tokens of the quoted text and the number of array entries are recorded separately, and note that quoteMaxToken only constrains the quoted content budget.
- Open the data rows to check that the status matches what is displayed in the interface, and check the identity identifiers of the file records and data rows separately
- View the data rows covered by the rebuild task, confirm that it targets only data with completed indexing, and that data pending indexing does not participate in the rebuild chain