What this category's data looks like
This page corresponds to the category of school safety hazard inspection and campus incident reporting. The materials include safety inspection checklists, hazard rectification ledgers, and emergency incident reports. Safety inspection checklists are filled out weekly or monthly by the responsible position, with fields including inspection date, location, hazard description, and inspector signature; hazard rectification ledgers are recorded on a per-instance basis, with fields including hazard name, rectification deadline, responsible person, review date, and review signature; emergency incident reports are formed per incident, with fields including incident name, occurrence time, number of people involved, reporter, and receipt acknowledgment. When these three types of materials enter this function, safety inspection checklists and ledgers serve as sources of basis entries, and emergency incident reports serve as sources for incident description and receipt verification. Model integration and configuration only determine whether dialogue and retrieval can invoke the corresponding models, and do not change the fields or signature status of the materials themselves.
Configuration and implementation checks
This step is executed by the server-side model channel configuration, which outputs model binding relationships and channel responses. "Basis entries that can be traced back to the original text" are subsequently completed by AI nodes, code nodes, or manual work. Language models are used for dialogue, indexing models are used for text chunk retrieval, reranking models are used to optimize retrieval ranking, and speech synthesis and speech recognition handle voice output and input respectively. The model types bound at each step must correspond to the step in which they are used.
| Object | Check | Verification action |
|---|---|---|
| Model type | The five categories—language, indexing, reranking, speech synthesis, and speech recognition—correspond to the steps in which they are used | Check the type labels in the model configuration, and confirm that the dialogue step is bound to a language model and the retrieval step is bound to an indexing model |
| Safety inspection checklist | Inspection date, location, hazard description, and inspector signature must match the hazard name, rectification deadline, and other items in the ledger one by one | Extract the same hazard, compare the name, quantity, date, and signature in the two materials, and record discrepancies |
| Hazard rectification ledger | Rectification deadline, review date, and review signature must be consistent with the timeline in the safety inspection checklist and emergency incident report | Check whether the review signature and receipt are complete, and confirm that the version is the current version |
| Emergency incident report | Incident name, occurrence time, number of people involved, reporter, and receipt acknowledgment must be mutually referable with the ledger and checklist | Compare the name and quantity of the same incident across the three materials, and check whether the receipt has any missing items |
Three common errors
- Phenomenon: The name, quantity, and date of the same hazard do not match between the safety inspection checklist and the rectification ledger. Cause: The two materials are filled out separately, without cross-material comparison and unified naming.
- Phenomenon: Cited entries point to old versions of materials, and the answer is inconsistent with the current version. Cause: Material versions are not synchronized, and retrieval still hits historical versions.
- Phenomenon: The emergency incident report lacks a receipt acknowledgment or review signature, and the basis entry cannot be traced back. Cause: The signature and receipt steps are not closed-loop, and the materials are not completely archived.
How to verify the setup
This function's acceptance: run one dialogue and one retrieval, check that the model identifier in the response is consistent with the model name returned by the channel, and check the values of maxContext, maxTokens, and quoteMaxToken in the model configuration; record the number of cited entries and the tokens of the cited text separately. quoteMaxToken is only the budget for cited content and should not be compared for equality with the number of array entries.
Business test: ask a cross-material basis question and see whether the answer provides a traceable source.
- In the model configuration, check that the model type corresponds to the step in which it is used, confirm that the channel configuration and key source are consistent, and that the model identifier is consistent with the name listed in the documentation.
- Extract the same hazard or incident, compare the name, quantity, date, and signature in the safety inspection checklist, hazard rectification ledger, and emergency incident report, and confirm that the version is the current version and the receipt is complete.