What this category's data looks like
In the enrollment consultation conversion scenario at training institutions, materials come from three categories: enrollment brochures, consultation records, and trial class schedules, and their update cadence changes on a rolling basis with class terms and instructor adjustments. Enrollment brochures contain course names, start dates, class hours, and signature fields; consultation records contain the consultant, intended course, follow-up date, and receipt; trial class schedules contain the trial time, instructor, and confirmation signature. When these materials enter the document parsing and chunking feature, enrollment brochures serve as the terminology source and translation-name baseline, consultation records serve as samples for multilingual comparison and back-referencing the source text, and trial class schedules serve as cross-validation for dates, quantities, and signature fields. Field characteristics center on names, quantities, dates, signatures, and receipts. The same record is prone to wording and numerical discrepancies across the three material categories, so it must first be parsed and persisted before entering index flow.
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 unified terminology table and comparison draft are then completed by retrieval and subsequent processing. After parsing is complete, parsed chunks are immediately persisted as visible data rows and bound to a unique data identifier. The index flows through three states, and data pending indexing does not participate in recall.
| Object | Check | Verification action |
|---|---|---|
| Enrollment brochure | Course names, start dates, class hours, and signature fields must correspond item by item with consultation records and trial class schedules; the same term must be written consistently throughout the entire document, and the translation name must be able to back-reference the source text. | Check the file records and data rows separately, confirm that the data row status is consistent with the interface display, and check whether the chunk boundaries corresponding to chunkSize and chunkSplitter cover the paragraph where the term is located. |
| Consultation record | The consultant, intended course, follow-up date, and receipt must be consistent with the names, quantities, and dates in the enrollment brochure and trial class schedule; the version must be synchronized and no signature or receipt may be missing. | Open the data row and check that the status is one of parsed / indexing / indexed, sample three terms to check whether their wording is consistent throughout the entire document, and record the processing mode corresponding to trainingType. |
| Trial class schedule | The trial time, instructor, and confirmation signature must align with the enrollment brochure and consultation record; the names, quantities, and dates of the same record across the three materials must not mismatch. | Check whether the data rows covered by the rebuild task contain only data with completed indexing; data pending indexing does not participate in the rebuild chain, and check trainingType and the index status transition records. |
Three common errors
- Symptom: newly uploaded materials cannot be found. Cause: successful parsing does not mean retrieval will definitely hit; data pending indexing is naturally isolated at the retrieval layer and is not recalled, so you must wait for the index status to transition to completed.
- Symptom: the names, quantities, and dates of the same record do not match across two materials. Cause: versions are not synchronized; after the enrollment brochure and consultation record are updated separately, the terminology table is not unified, and the translation name cannot back-reference the source text.
- Symptom: after training fails, empty data rows remain, and signatures or receipts are missing. Cause: training interruption leaves orphan data; failed data is not edited or deleted in time, and receipt fields are not completed in sync with the version.
How to verify the setup
Acceptance for this feature: upload one material, observe the data row status transitions and the coverage scope of the rebuild task, confirm that after parsing is complete it is immediately persisted as visible data rows and bound to a unique data identifier, that the index flows through the three states parsed / indexing / indexed, and that changing the model rebuilds only data with completed indexing.
Business test: sample three terms to check whether their wording is consistent throughout the entire document; the same term must be consistent throughout and the translation name must be able to back-reference the source text. Before external release, check the fields, dates, and signatures of the enrollment brochure, consultation record, and trial class schedule.
- Open the data row and check the status, confirm that it is one of parsed / indexing / indexed and consistent with the interface display, check the file records and data rows separately, and confirm that the identifiers of the file and the data row are distinguishable.
- Look at the data rows covered by the rebuild task, confirm that only data with completed indexing is covered, that data pending indexing does not participate in the rebuild chain, and check the parameters corresponding to
chunkSize,chunkSplitter, andtrainingTypeand the chunking results.