What this category's data looks like
This page addresses the university laboratory safety grading and classification scenario. Materials include laboratory grading and classification tables, hazard source lists, safety responsibility statements, as well as draft materials that can be issued and handover records. The grading and classification table is organized by room, hazard source category, and level fields; the hazard source list records name, quantity, storage location, and control measures; the safety responsibility statement contains responsible persons, responsibility clauses, signatures, and dates. When these materials enter the knowledge base retrieval service, they serve only as sources of candidate snippets. The retrieval node outputs candidate snippets and similarity scores, and does not produce statistical tables, business documents, or authorization results. Drafts and handover records serve as output templates pending verification, to be generated by subsequent nodes or manually based on them. The material update cadence adjusts with laboratory renovations, increases or decreases in hazard sources, and changes in responsible persons; versions and dates must be synchronized. During retrieval, select the retrieval mode according to the material form so that candidate snippets cover fields, dates, and signature information.
Configuration and implementation checks
The knowledge base retrieval service (retrieval node and search test) outputs only candidate snippets and similarity scores (after mode, merging, reranking, filtering, and upper-limit trimming). Draft materials that can be issued and handover records must be completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Laboratory grading and classification table | Whether fields, dates, and signatures are consistent with the hazard source list | In the search test, retrieve with synonymous questions, check the names, quantities, and dates in candidate snippets, record the number of candidates and scores, switch searchMode to compare semantic, full-text, and hybrid results |
| Hazard source list | Whether the name, quantity, and date of the same record match across the two materials | Retrieve the hazard source name and quantity, check whether candidate snippets contain both semantic and full-text results, enable usingReRank to compare the ranking of the top few, and check the quoted text after similarity filtering |
| Safety responsibility statement | Whether the responsible person, responsibility clauses, signature, and receipt are complete | Retrieve the responsible person's name and signature field, check whether candidate snippets retain key original text snippets, record the number of returned quotes and scores, and check whether dates and versions are synchronized |
| Draft materials that can be issued and handover records | Complete fields, clear responsible person, consistent version and date | Draft one according to the template and complete one handover, see whether missing items are blocked; when retrieving draft fields, record the number of candidates and scores, record the number of quotes separately by array length, and when the quote limit is involved, check the tokens of the quoted text and limit |
Three common errors
- Phenomenon: The name, quantity, and date of the same hazard source do not match across the two materials. Cause: Drafting and handover use different candidate snippets, without checking hits and associations.
- Phenomenon: The grading and classification table has been updated, but the responsibility statement still uses the old version. Cause: Versions are not synchronized, and retrieval candidates do not cover the latest date field.
- Phenomenon: The handover record lacks a signature or receipt. Cause: Candidate snippets did not recall the signature field, and subsequent nodes did not fill it in according to the template.
How to verify the setup
Acceptance of this feature: test once each with synonymously rewritten questions, compare candidates and scores, then toggle reranking on and off to see the ranking of the top few; record the number of candidates and scores, check similarity filtering and quoted text tokens, and record array counts separately.
Business test: draft one according to the template and complete one handover, see whether missing items are blocked.
- In the knowledge base search test, switch
searchModeto compare the number of candidates and scores across semantic, full-text, and hybrid modes. - Compare the ranking of the top few before and after enabling
usingReRank, and check whetherlimitmatches the quoted text tokens.