What this category's data looks like
This page addresses the enterprise compliance and safety training category. The source materials are compliance training courseware, safety operating procedures, exam records, and receipts. Courseware is updated by policy version, operating procedures are synchronized with equipment and process changes, exam records are archived by session and personnel, and receipts are collected by training batch. Field characteristics center on course name, clause number, equipment name, operating steps, personnel name, quantity, date, and signature field. When materials enter the knowledge base retrieval and recall feature, courseware and procedures serve as the primary candidate sources for semantic and full-text recall, while exam records and receipts serve as the basis for verifying dates, signatures, and quantities, used to locate the original material text during missing-item identification and follow-up questions.
Configuration and implementation checks
This step is performed by the knowledge base retrieval service (retrieval node and search test), outputting candidate chunks and similarity scores (after pattern, merge, re-rank, filter, and limit trimming). Intake records and completed follow-up questions are handled by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Compliance training courseware | Whether the course name, clause number, date, and signature field in the courseware fields are consistent with the exam records | Switch searchMode in the search test, check the course name and clause number in the candidate chunks, and record the number of candidates and the similarity score |
| Safety operating procedures | Whether the equipment name, operating steps, quantity, and version date in the procedures match the receipts or exam records | Compare the top few rankings before and after enabling usingReRank, check the procedure version date and signature field, and record the number of candidates |
| Exam records and receipts | Whether the name, quantity, and date of the same record are consistent across the two materials, and whether signatures or receipts are complete | Test once each with synonymous questions, check whether the candidate chunks contain both exam records and receipts, and separately record the array count and reference text tokens |
| Knowledge base retrieval service | Whether candidate chunks and similarity scores are returned after pattern, merge, re-rank, filter, and limit trimming | Adjust limit to observe changes in the number of candidates, check the similarity score and reference text tokens, and do not compare array length and reference budget for equality |
Three common errors
- Symptom: The course name, quantity, or date of the same training record does not match between the courseware and the exam records. Cause: The two materials are maintained separately, without field alignment by the same session and personnel.
- Symptom: The operating procedure version is not synchronized, and old steps are still recalled. Cause: After a version update, the material was not re-imported or not labeled with an effective date, so retrieval still hits old chunks.
- Symptom: Receipts lack signatures or exam records lack receipts, and follow-up questions cannot complete them. Cause: Material archiving is incomplete, and after missing-item identification there is no corresponding original text to cite.
How to verify the setup
Feature acceptance: Test once each with synonymously rewritten questions, compare candidate chunks and similarity scores, then toggle re-ranking to inspect the top few rankings, and record the number of candidates and reference text tokens.
Business test: Ask a question missing key information and see whether it asks a follow-up question before answering.
- In the search test, switch
searchModeand separately record the number of candidates andsimilarityscores under semantic, full-text, and hybrid modes. - Test once each with
usingReRankenabled and disabled, compare the ranking of the top few candidates, and adjustlimitto observe the reference text tokens and array count after trimming.