What this category's data looks like
In course selection, scheduling, and classroom resource allocation scenarios, materials typically include three categories: course offering plans, timetables, and course selection rules. Course offering plans are maintained by the academic affairs office or departments per semester, with fields including course name, course code, credits, offering unit, instructor, and signature field; timetables are generated by the scheduling system or academic affairs staff by week, with fields including class, classroom, period, start and end weeks, and version number; course selection rules are published by the academic affairs office, with fields including selection batch, capacity, priority, and receipt. The three categories of materials have different update rhythms: course offering plans are revised centrally before the semester, timetables are fine-tuned multiple times before and after course selection, and course selection rules are released on a rolling basis by batch. When entering the knowledge base retrieval node, they serve respectively as the retrieved original text sources, and the citation output returns fragments and source identifiers according to the hit paragraphs, for subsequent nodes to compare fields, write ledgers, and generate reports.
Configuration and implementation checks
This link is executed by the citation output of the knowledge base retrieval node, which outputs an array of cited content, each item containing an original text fragment and a source identifier, and the length can be 0. The "discrepancy list" is completed subsequently by AI nodes, code nodes, or manually.
| Object | Check | Verification action |
|---|---|---|
| Course offering plan | Course name, credits, instructor, and signature field are consistent in both materials | Use one real record to compare the course offering plan with the timetable, and check the cited fragments and source identifiers in responseData (non-streaming module results returned with detailed output enabled) |
| Timetable | Classroom, period, start and end weeks, and version number are consistent in the same record | Open the citation superscript to trace back to the original text, and check the fields in quoteList against the ledger item by item |
| Course selection rules | Selection batch, capacity, priority, and receipt date match | Check the number of items in the citation array against the hit materials, record the array length separately, and do not compare the number of items with quoteMaxToken for equality |
| Knowledge base retrieval node | The selected knowledge base uses the same vector model, and the search parameters are consistent with the knowledge base retrieval settings | Open the search parameter panel and check item by item, and run a test with a question that has no matching materials to confirm the output is empty while the pipeline continues |
Three common errors
- Symptom: The name or credits of the same course do not match between the course offering plan and the timetable. Cause: The two materials are maintained by different personnel at different times, versions are not synchronized, and no field-level comparison is performed.
- Symptom: The classroom or period in the timetable is inconsistent with the ledger. Cause: After a scheduling adjustment, only one place is updated, while the other still retains the old version, lacking a receipt or signature confirmation.
- Symptom: The answer has content but the citation array is empty, or the cited fragments do not match the original text. Cause: No materials were hit in this round, or a different record was hit; it should not be directly attributed to length truncation.
How to verify the setup
This feature acceptance: Ask a question with original text evidence, check whether the citation array and superscripts can trace back to the original text, and view the source identifier of each citation in responseData; then ask a question with no matching materials, confirm the array is empty while the pipeline runs through, record the array length separately, and do not compare it with quoteMaxToken for equality.
Business test: Use one real record to find fields inconsistent with the ledger, submit them to the reviewer for confirmation, and write them back.
- Open the citation superscript, check the original text fragments and source identifiers item by item against the source document, and confirm that each citation can trace back to the original text paragraph.
- Run a test with a question that has no matching materials, confirm the citation array is empty while the output pipeline continues, and record the array length and the
quoteMaxTokenbudget.