What this category's data looks like
Vocational college program setup filing is mainly based on three types of materials: the program setup filing form, talent demand justification, and faculty and equipment list. The sources are submissions from academic affairs, human resources, equipment, and departments/colleges, updated with each application batch. Fields include program name, code, program length, planned enrollment number, faculty count, equipment units/sets, justification date, and signature. When the three types of materials enter the knowledge base retrieval service, their roles differ: the filing form is the main record, the talent demand justification provides supporting paragraphs, and the faculty and equipment list supplements quantity and date fields. During retrieval, they serve respectively as sources of candidate snippets for subsequent comparison of field consistency.
Configuration and implementation checks
The knowledge base retrieval service (retrieval node and search test) only outputs candidate snippets and similarity scores (after mode, merge, rerank, filter, and upper-limit trimming). Material drafts that can be issued and handover records must be completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Program setup filing form | Fields, dates, and signatures are complete; the program name and code are consistent across all types of materials | In the search test, search once each by program name and code, check the name, date, and signature fields in the candidate snippets, record the number of candidates and scores, and confirm that searchMode matches the material form |
| Talent demand justification | The justification basis paragraphs can be recalled, and the date is synchronized with the filing form version | Search with justification keywords, turn usingReRank on and off to compare the ranking of the top few results, and check the date and version identifier in the candidates |
| Faculty and equipment list | The faculty count and equipment units/sets match across the two materials, and signatures and receipts are not missing | Search the faculty and equipment fields separately, check the quantity and dates in the candidate snippets, and check whether key snippets are still retained after similarity filtering |
| Handover record | The responsible person is clear, the version and date are consistent, and the next stage has signed for receipt | Draft one according to the template and complete one handover, check whether missing items are blocked, record the number of candidates and scores, and confirm that the citation upper limit is trimmed according to limit |
Three common errors
- Symptom: The same program name is written differently in the filing form and justification materials, and the quantities and dates do not match. Cause: Multi-source materials were not aligned according to the main record fields, and retrieval candidates were not compared for consistency.
- Symptom: The faculty and equipment list version is not synchronized, and old-version quantities are still recalled. Cause: The update cadence and retrieval index are not synchronized, and candidate snippets are not labeled with version and date.
- Symptom: The handover record lacks a signature or receipt, and the next stage cannot sign for receipt. Cause: Fields were not validated during drafting, and retrieval candidates did not cover the signature and receipt paragraphs.
How to verify the setup
Acceptance for this feature: test once each with synonym-rewritten questions, compare candidates and scores, then turn rerank on and off to observe the ranking of the top few results; record the number of candidates and scores, and confirm the effects of searchMode, similarity, limit, and usingReRank.
Business test: draft one according to the template and complete one handover, and see whether missing items are blocked.
- In the knowledge base search test, switch retrieval modes, compare candidate snippets and similarity scores, and record the number of candidates.
- Turn rerank on and off, compare the order of the top few candidates, and check the tokens and array length of the cited text.