What this category's data looks like
This page addresses the student awards, grants, loans, and hardship determination scenario. Materials include three types: family economic hardship determination forms, award and grant application forms, and loan receipts. Determination form fields include household population, income sources, hardship level, and signature date; application form fields include student ID, award name, amount, review comments, and signature; receipt fields include loan contract number, disbursement date, amount received, and acknowledgment. Materials are updated by academic year and batch: determination forms are submitted centrally each academic year, application forms flow with award batches, and receipts are backfilled according to disbursement cycles. When entering the knowledge base retrieval and recall function, the determination form takes on the role of hardship level and review basis, the application form takes on the role of award and amount verification, and the receipt takes on the role of disbursement status and acknowledgment. Together, the three support a handover-ready deployment state and operation records with consistent versions, connections, and task records. Retrieval returns candidate snippets and similarity scores for subsequent nodes or manual generation of business deliverables.
Configuration and implementation checks
This step is executed by the knowledge base retrieval service (retrieval node and search test), outputting candidate snippets and similarity scores (after mode, merge, rerank, filter, and limit trimming). The handover-ready deployment state and operation records are then completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Family economic hardship determination form | Fields, dates, and signatures must be consistent with the application form; if the name, quantity, or date of the same record does not match across the two materials, it is deemed inconsistent | In the search test, test once each with synonymous questions, record the number of candidates and scores, and check searchMode and similarity |
| Award and grant application form | The award name, amount, and review date must correspond to the determination form's level; when versions are not synchronized, the latest signed version prevails | Switch between semantic, full-text, and hybrid modes to compare candidates, and enable usingReRank to observe changes in the ranking of the top few results |
| Loan receipt | The contract number, disbursement date, and amount received must be consistent with the application form records; if signatures or receipts are missing, it must not enter handover | Check the tokens and array length of the returned reference text, and separately record the number of candidates after limit trimming |
| Knowledge base retrieval service | Candidates and referenced content are not statistical tables, business documents, or authorization results; if the source does not match, first check hits and associations | Check item by item against the handover checklist whether versions, connections, and task records are consistent |
Three common errors
- Symptom: The hardship level, amount, and date for the same student do not match between the determination form and the application form. Cause: The two materials' versions are not synchronized, old signatures have not been withdrawn, and retrieval hits old snippets.
- Symptom: The loan receipt lacks a signature or the receipt has not been backfilled, yet the handover checklist still shows it as handover-ready. Cause: Task records do not cover receipt status, and connections and versions have not been checked item by item.
- Symptom: Changing the phrasing causes key passages to not be recalled; there are many candidates but the top-ranked ones are all irrelevant. Cause: The retrieval mode does not match the material form, and rerank and similarity filtering have not been enabled as needed.
How to verify the setup
Acceptance for this function: test once each with synonymously rewritten questions, compare candidates and scores, then toggle rerank on and off to observe the ranking of the top few results; record the number of candidates under searchMode, similarity, usingReRank, and limit.
Business test: check item by item against the handover checklist once to see which record is missing; check whether versions, connections, and task records are consistent; if signatures or receipts are missing, return for backfilling.
- In the knowledge base search test, switch between semantic, full-text, and hybrid modes, compare the number of candidates and scores, and confirm that the material form matches the mode.
- Compare the order of the top few results before and after enabling rerank, check the tokens and array length of the reference text, and check item by item against the handover checklist for versions, connections, and task records.