What this category's data looks like
This page addresses the enrollment consultation conversion scenario at training institutions, with materials drawn from three sources: enrollment brochures, consultation records, and trial class schedules. Enrollment brochures provide course names, class types, class hours, and fee definitions, and serve as the baseline source for the terminology glossary; consultation records carry the original wording, objections, and commitments of parents and students, and serve as the primary basis for translated names referring back to the source text; trial class schedules provide date, time slot, teacher, and classroom fields, and are used to verify whether the dates and signatures of the same record are consistent across different materials. Materials are updated in batches: brochures are revised by quarter, consultation records accumulate daily, and trial class schedules are adjusted weekly. Field characteristics include mixed multilingual text, a high repetition rate of terminology, multiple translated spellings for the same course, and frequent cases of unsynchronized versions, missing signatures, or missing receipts. When the three types of materials enter this feature, brochures and trial class schedules serve as the terminology and date baseline, while consultation records serve as the source text for the comparison draft, jointly supporting the generation of a unified terminology glossary and comparison draft.
Configuration and implementation checks
This step is executed by the indexing service and vector database, outputting vectors and index status (whether recall is possible and whether data in the same group is merged). The "unified terminology glossary and comparison draft" is then completed by subsequent AI nodes, code nodes, or manual work (retrieval and citation trimming only take supporting materials). Retrieval and citation trimming only take supporting materials and do not generate a terminology glossary, comparison draft, or version ledger.
| Object | Check | Verification action |
|---|---|---|
| Enrollment brochure | Whether the course name, class type, class hours, and fee fields are written consistently in the brochure and consultation records, and whether dates and signatures are complete | In the knowledge base settings, check the index model and confirm it matches the vector model actually used by the knowledge base; open a brochure data row and check the index fields to confirm that multiple vector identifiers for one data record are all present, and record the values of defaultToken and maxToken |
| Consultation record | Whether the name, quantity, and date of the same record match between the consultation record and the trial class schedule, and whether the receipt is missing | Run a retrieval once to see whether data in the same group is merged and whether recall is possible; after changing the model, check the data scope covered by the rebuild task to confirm that it is triggered only for data that has already been indexed, and check the status value indexed |
| Trial class schedule | Whether the date, time slot, teacher, and classroom fields are consistent with the brochure and consultation records, and whether signatures or receipts are complete | Check whether the data row status is indexed, and view the rebuild task record; confirm that the multiple vector identifiers recorded in the index fields correspond to the business database data row, and record the budgets of defaultToken and maxToken |
Three common errors
- Symptom: the name, quantity, and date of the same course do not match between the brochure and consultation records. Inspect the following: the two materials were imported separately and their versions were not synchronized.
- Symptom: the trial class schedule is missing signatures or receipts, and the externally published draft cannot refer back to the source text. Inspect the following: the signature field was not verified when the materials entered the knowledge base, and retrieval only takes supporting materials and does not add signatures.
- Symptom: after changing the model, recall quality decreases and duplicate entries appear in retrieval results. Inspect the following: the rebuild task only covers data that has already been indexed, and data rows that have not completed indexing do not participate in recall, so indexing must be completed before rebuilding.
How to verify the setup
Feature acceptance: sample one data row and check the index fields to confirm that multiple vector identifiers for one data record are all present, then run a retrieval once to see whether data in the same group is merged and whether recall is possible; for status values, only indexed and the rebuild task record are recognized, and check the budgets of defaultToken and maxToken.
Business test: sample three terms and check whether their spellings are consistent throughout the entire text, whether translated names can refer back to the source text, and whether dates and signatures are complete; the number of citations is the array length, and quoteMaxToken is the citation content budget. The two are not compared for equality; the citation upper limit only checks the tokens of the citation text, and the array length is recorded separately.
- In the knowledge base settings, check the index model and confirm it matches the vector model actually used by the knowledge base; open a data row and check the index fields, and record the values of
defaultTokenandmaxToken. - View the data scope covered by the rebuild task, confirm that after changing the model it is triggered only for data that has already been indexed, and record the rebuild task record and the status value indexed.