What this category's data looks like
In the enrollment consultation conversion scenario at training institutions, materials fall into three categories: enrollment brochures, consultation records, and trial class schedules. Enrollment brochures contain course names, class types, start dates, fee items, and version numbers; consultation records contain visit dates, intended courses, key communication points, and signature fields; trial class schedules contain trial class times, instructors, classrooms, and receipt fields. The three types of materials are maintained separately by marketing and academic affairs, and their update cadence follows class-term and policy adjustments. When entering knowledge base retrieval, enrollment brochures serve as the source of course definitions, consultation records as the source of communication facts, and trial class schedules as the source of scheduling and receipts; each is split into candidate chunks that participate in recall and reranking.
Configuration and implementation checks
This step is executed by the knowledge base retrieval service (retrieval node and search test), which outputs candidate chunks and similarity scores (after mode, merge, rerank, filter, and limit truncation). The version ledger and change records are completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Enrollment brochure | Course name, class type, start date, and version number are consistent between the brochure and the consultation records | In the search test, run one search each with synonymous questions, check the number of candidates and the similarity scores, and switch searchMode to compare recall differences |
| Consultation record | Visit date, intended course, and signature field match between the record and the receipt | After changing one course name, search again to see whether the old-version chunk is still recalled, and check the candidates after similarity filtering |
| Trial class schedule | Trial class time, instructor, and classroom are consistent between the schedule and the receipt | Compare the top few rankings before and after enabling usingReRank, and confirm whether scheduling chunks remain stably near the top |
| Version ledger and change records | The current version is unique, old versions are traceable, and references are in sync | Check the number of references by limit, record the tokens of the referenced text separately, and verify whether the version number and references are updated in sync |
Three common errors
- Symptom: the name and start date of the same course do not match across two materials. Cause: after the brochure was revised, the consultation records were not updated in sync, and the old-version chunk is still recalled.
- Symptom: the consultation record lacks a signature or receipt, and the trial class schedule and record dates are inconsistent. Cause: materials were not checked against the version ledger when archived, and changes left no trace.
- Symptom: there are many candidates, but the ones ranked at the top are all irrelevant, and the answer lacks key passages. Cause: the retrieval mode does not match the material form, and reranking and similarity filtering were not performed.
How to verify the setup
Acceptance for this feature: test once each with synonymously rewritten questions, compare the number of candidates and the similarity scores, then toggle reranking on and off to inspect the top few rankings.
Business test: change one piece of content and see whether the version number and references are updated in sync.
- In the knowledge base search test, switch
searchModeand check whether the candidates include both semantic and full-text results. - Adjust
similarityandlimit, record the number of references and the tokens of the referenced text, and verify whether the version number and references are updated in sync.