Vector Models and Indexes for Grade Entry and Credit Recognition: Find Supporting Evidence

What this category's data looks like

This page addresses grade entry and credit recognition scenarios, and the materials include three types: transcripts, deferred examination applications, and make-up examination arrangements. Transcripts record the course name, credits, grade, and signature field; deferred examination applications record the application date, reason, and approval signature; make-up examination arrangements record the make-up examination time, location, and receipt. The materials are updated by semester batch, and the version is based on the current semester. When entering the business basis query function, transcripts serve as the primary record source, while deferred examination applications and make-up examination arrangements serve as supporting materials, each taking on the role of field comparison and date verification. The name, quantity, and date of the same record may not match across different materials, and versions may be out of sync or signature receipts may be missing, causing basis entries to be unable to point back to the original text. Field characteristics include course code, credit value, signature status, and receipt number, which need to be checked item by item against the original text.

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 point 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). The indexing model must be consistent with the vector model actually used by the knowledge base. The index field of a data row records multiple vector identifiers, and the rebuild task is triggered only for data that has already been indexed.

Three common errors

• Symptom: duplicate entries appear in retrieval results. Inspect the following: the same group of data corresponds to multiple vectors, and the highest score is not taken during merging, resulting in duplicate recall.
• Symptom: the name, quantity, and date of the same record do not match in the two materials. Inspect the following: the versions of the transcript and deferred examination application are not synchronized, and the fields are not checked item by item.
• Symptom: recall quality decreases after changing the model. Inspect the following: the rebuild task is triggered only for data that has already been indexed, and does not cover all data rows.

How to verify the setup

This feature acceptance: sample one data row and look at the index field, confirm the status is indexed and multiple vector identifiers are present; then retrieve once, and see whether data in the same group is merged and whether the vector score takes the highest score. The number of citations is the array length, the citation content budget depends on the citation text tokens, and the array count is recorded separately; check whether limit match the indexing model.

Business test: ask a basis question that spans materials, and see whether the answer gives a traceable source, whether the basis comes from this repository's materials, whether the version is current, and whether the entries match the original text.

• Check the indexing model in the knowledge base settings, confirm it is consistent with the vector model actually used by the knowledge base, and check the values of defaultToken and maxToken.
• Open a data row and check the index field, confirm the status is indexed and multiple vector identifiers are complete; view the coverage scope of the rebuild task record.