What this category's data looks like
This page's material comes from the kindergarten health care and meal management scenario: ingredient names and specifications, sample retention times and batches, disinfection time slots, absence cause classification, as well as menus, procurement documents, and shift handover records. When these materials enter the knowledge base retrieval node, each plays a different role—ingredient and sample retention entries provide the basis for batch verification, disinfection time slot entries provide the basis for time slot completeness, absence cause classification provides the classification criteria, and menus and procurement categories provide the basis for category consistency. Materials are ingested as document paragraphs, and source identifiers point to the original paragraphs. The update cadence follows daily menus, per-meal sample retention, each disinfection, and daily absence registration; changes in batches and time slots trigger the ingestion of new paragraphs. The field characteristics are short text plus timestamps and batch numbers, suitable for item-by-item reference back to the original text.
Configuration and implementation checks
This step is executed by the citation output of the knowledge base retrieval node, which outputs an array of cited content, each entry containing the original text fragment and a source identifier, and the length can be 0. The "handover-ready deployment state and operation records" are completed subsequently by AI nodes, code nodes, or manual work. The number of array entries is the number of citations, and quoteMaxToken is the citation content budget; the two are not compared for equality.
| Object | Check | Verification action |
|---|---|---|
| Associated knowledge base | The selected collections must be knowledge bases of the same vector model, consistent with the vector model registered at deployment | In the node, check the selected collections and their vector models, and record the collection list and model identifiers |
| Citation content output | The number of returned entries matches the hit materials; being empty is also normal, and the chain continues to run through | Run a trial with a question that has no matching materials, and see that the output is empty while the chain continues; record the array length in responseData (non-streaming module results returned with detailed output enabled) |
| Citation-to-original correspondence | Each citation can be traced back to the original paragraph; ingredient names and specifications, sample retention times and batches, disinfection time slots, and absence cause classification are each verifiable | In the conversation, open the citation markers, compare against the source documents, and check whether batches and time slots are consistent with the documents |
| Search parameters | Consistent with the values set in the knowledge base retrieval section; the tokens of the cited text are checked against the quoteMaxToken budget | Open the search parameters panel and check item by item, view the hit materials in quoteList, and record the array entry count separately |
Three common errors
- Symptom: documents and sample retention batches do not match. Cause: the ingested paragraph records only the ingredient name and not the batch, retrieval hits an old batch paragraph, and the citation cannot be traced back to the current meal's sample retention.
- Symptom: menus and procurement categories are inconsistent. Cause: menu paragraphs and procurement document paragraphs belong to different knowledge bases, retrieval hits only one side, and the citation array lacks the source from the other side.
- Symptom: disinfection records are missing time slots. Cause: disinfection time slot paragraphs are not split by meal, retrieval hits a whole-day entry, and the citation fragment cannot cover the missing time slot.
How to verify the setup
Acceptance of this feature: ask a question with original-text support and check whether the citation array and markers can be traced back to the original text; then ask a question with no matching materials, confirm that the array is empty while the chain runs through, and record the array length in responseData and the quoteMaxToken budget values separately.
Business testing: check item by item once according to the handover sheet, and see which record is missing—ingredient names and specifications, sample retention times and batches, disinfection time slots, absence cause classification—comparing each item against the citation source identifiers.
- Open the search parameters panel and check item by item, confirm that the selected collections are of the same vector model, and record the collection list and model identifiers.
- In the conversation, open the citation markers, compare against the source documents, and check whether ingredient names and specifications, sample retention times and batches, disinfection time slots, and absence cause classification are consistent with the documents.