What this category's data looks like
This category's materials for enterprise sales product training include product manuals, sales scripts, and channel training schedules. Product manuals record product names, feature descriptions, and version dates; sales scripts record sales script entries, applicable scenarios, and signed confirmations; channel training schedules record course names, start dates, and attendance receipts. The three types of materials play different roles in knowledge base retrieval and recall: product manuals provide product fact fragments, sales scripts provide repeatable expressions, and channel training schedules provide time and personnel arrangements. The material update cadence is: sync after version release, backfill after channel adjustments, and archive receipts after training ends. Field characteristics center on names, quantities, dates, and signatures; the same record often has mismatched names, quantities, or dates across two materials, unsynchronized versions, or missing signatures or receipts. When entering retrieval, the three types of materials serve as candidate fragment sources respectively, recalled and merged by the retrieval node according to the mode.
Configuration and implementation checks
This step is executed by the knowledge base retrieval service (retrieval node and search test), outputting candidate fragments and similarity scores (after mode, merge, rerank, filter, and limit trimming). The externally available entry points and authorization scope are completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Product manual | The product names, feature descriptions, and version dates in the manual must be consistent with the sales scripts and channel training schedules; if the names and dates of the same record do not match across two materials, it is considered unsynchronized | In the knowledge base search test, test once each with synonymous rephrased questions, record the number of candidates and similarity scores, and check whether the hit fragments come from the same version |
| Sales script | The sales script entries, applicable scenarios, and signed confirmations must correspond to the product manuals and channel training schedules; records missing signatures or receipts do not enter the externally visible scope | Switch searchMode to semantic, full-text, or hybrid, compare whether the candidates include both semantic and full-text results, and check the signature and receipt fields |
| Channel training schedule | The course names, start dates, and attendance receipts must be consistent with the product manuals and sales scripts; if the quantities and dates of the same record do not match across two materials, it is considered a version not synchronized | Enable and disable usingReRank to compare the ranking of the top few items, check the tokens of the cited text after similarity filtering, record the array count separately, and check whether key fragments are retained after limit trimming |
Three common errors
- Symptom: The same product has mismatched names and dates across two materials. Cause: The version is not synchronized; the update only changed one place, while the other remains the old version.
- Symptom: The number of sales script entries is inconsistent with the number of training attendees in the schedule. Cause: Missing signatures or receipts, and the original table was not written back during backfilling.
- Symptom: The same question cannot be recalled when rephrased. Cause: The retrieval mode does not match the material form, and the candidates do not cover both the semantic and full-text paths.
How to verify the setup
This feature acceptance: test once each with synonymous rephrased questions, compare candidates and similarity scores, then toggle rerank to observe the ranking of the top few items; at the same time switch searchMode and adjust similarity and limit, record the number of candidates, scores, and cited text tokens, and record the array count separately.
Business test: access once with a restricted identity, check whether the visible scope matches the setting, and ensure unauthorized content is not visible.
- In the knowledge base search test, issue synonymous questions separately for the product manual, sales script, and channel training schedule, record the number of candidates and similarity scores, and check whether the hit fragments come from the same version.
- Enable and disable
usingReRankto compare the ranking of the top few items, check the tokens of the cited text aftersimilarityfiltering, record the array count separately, and confirm that key fragments are still retained afterlimittrimming.