What this category's data looks like
Corporate compliance and security training materials typically include compliance training courseware, security operating procedures, exam records, and signed receipts. Their sources are scattered across policy documents, training notices, sign-in sheets, and assessment forms, and their update cadence follows regulatory revisions, incident reviews, and annual training plans. When materials enter this feature, their roles differ: courseware and procedures serve as knowledge body text for retrieval and citation, exam records and signed receipts serve as credential fields for traceability, and the version ledger and change records link the current status of each material. Field characteristics center on four categories: names, quantities, dates, and signatures. The same record often appears in two materials with inconsistent names, mismatched quantities, misaligned dates, or missing signatures or receipts, so fields and dates must be checked separately by material type, and old versions must be retained for traceability.
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. After parsing is complete, parsing and chunking are immediately persisted as visible data rows and bound to a unique data identifier. The index transitions through three states: parsed, indexing, and completed. Data pending indexing does not participate in recall, and changing the model only rebuilds data whose indexing is completed.
| Object | Check | Verification action |
|---|---|---|
| Compliance training courseware | Field check dimensions: whether the courseware name, clause number, publication date, signature field, and receipt are complete and consistent with the ledger | Open the data row to check whether the status is one of parsed / indexing / indexed, and separately check the identifiers of the file record and the data row to confirm the processing mode corresponding to trainingType |
| Security operating procedures | Field check dimensions: whether the procedure name, quantity, effective date, revision records, and signature match the exam records | Check whether the data rows covered by the rebuild task include only data whose indexing is completed, and confirm that the chunking results corresponding to chunkSize and chunkSplitter are consistent with the original text |
| Exam records and signed receipts | Field check dimensions: whether the exam name, number of people, exam date, signature, and receipt are consistent with the records in the courseware and procedures | Separately check the file record and the data row, and confirm that when the name, quantity, or date of the same record in two materials does not match, the current version prevails |
Three common errors
- Symptom: newly uploaded materials cannot be found. Cause: successful parsing does not mean retrieval will definitely hit; the index status also matters, and data pending indexing does not participate in recall.
- Symptom: empty data rows remain after training fails. Cause: parsing and chunking are immediately persisted as visible data rows, and failed training data is not cleaned up or retried.
- Symptom: the name, quantity, or date of the same record in two materials does not match. Cause: versions are not synchronized, old versions are not marked, and signatures or receipts are missing.
How to verify the setup
Feature acceptance: upload one material, observe the data row status transitions and the coverage scope of the rebuild task, confirm that the status transitions correctly among parsed / indexing / indexed, and that the rebuild task covers only data whose indexing is completed.
Business test: change one piece of content, observe whether the version number and references are updated synchronously, and check whether the file record and data row identifiers are each independent.
- Open the data row to check whether the status is consistent with the interface display, and confirm that the identities of the file record and the data row can be distinguished from each other.
- View the data rows covered by the rebuild task, confirm that it targets only data whose indexing is completed, and that the initial training covers data that is pending indexing and indexing.