What this category's data looks like
This page focuses on the business basis query for undergraduate program revision. The materials come from this institution's program, course syllabi, credit structure tables, and accompanying receipts. The program is revised by academic year, course syllabi are updated as courses are adjusted, and the credit structure table is checked at the beginning of the semester; the current signed draft is the authoritative version. When the three types of materials enter this function, the program provides the overall framework and revision notes, the course syllabi support course objectives and content basis, the credit structure table supports credit and contact-hour basis, and the receipts are used to confirm version and signature status. Field characteristics include course name, credit count, revision date, signature field, and version number. It is common for the name, count, or date of the same record to be inconsistent across different materials, or for the version to be out of sync, or for signatures or receipts to be missing.
Configuration and implementation checks
This step is executed by the indexing service and vector database, outputting vectors and index status (whether recall is possible, whether data in the same group is merged). "Basis entries that can be traced back to the original text" are then completed by subsequent AI nodes, code nodes, or manual work (retrieval and citation trimming only take supporting materials).
| Object | Check | Verification action |
|---|---|---|
| Program | Course name, credit count, revision date, signature field, and version number must be consistent with the current signed draft | Open a data row and check the index fields, confirm that multiple vector identifiers are present, and check that the status is indexed |
| Course syllabus | Course objectives, content, and contact-hour fields must match the name, count, and date of the corresponding record in the program | In the knowledge base settings, check that the indexing model is consistent with the vector model actually used by the knowledge base |
| Credit structure table | Credit totals, module proportions, and dates must be compared item by item with the program and course syllabi; missing signatures or receipts must be flagged | Check the data scope covered by the rebuild task, and confirm that after changing the model, it is triggered only for data that has already been indexed |
| Data row | Multiple vector identifiers for one data row are all present; after retrieval, data in the same group is merged and the highest score is taken | Check limit for the citation text budget; the number of array entries is recorded separately |
Three common errors
- Symptom: duplicate entries appear in retrieval results. Inspect the following: the same group of data corresponds to multiple vector identifiers, and the highest score was not taken during merging, or index fields are missing.
- Symptom: recall quality drops after changing the model. Inspect the following: the rebuild task did not cover all data that has already been indexed, and the version is out of sync.
- Symptom: a data row has content but cannot be retrieved. Inspect the following: the index status has not reached indexed, or the course name, credit count, or date does not match between the two materials.
How to verify the setup
This function's acceptance check: sample one data row and look at the index fields, confirm that multiple vector identifiers are present and the status is indexed, then run a retrieval once to see whether the same group is merged and the highest score is taken.
Business test: ask a basis question that spans materials, see whether the answer gives a traceable source, and check that the number of citations and the quoteMaxToken budget are recorded separately.
- In the knowledge base settings, check that the indexing model is consistent with the vector model actually used by the knowledge base, and check that the data row index fields are complete.
- Check the data scope covered by the rebuild task records, confirm that after changing the model, it is triggered only for data that has already been indexed, and separately flag materials with missing signatures or receipts.