What the documents look like
Complaints and regulatory replies are directed at platforms or supervisory authorities, so the material carries formal evidence and rectification wording rather than ordinary after-sales chat. The working records come from the 12315 platform, platform complaint centres, customer service conversations, and order systems. Day-to-day documents include complaint tickets, regulatory inquiry letters, order evidence, and rectification reports. Records update irregularly, case by case.
For operating reports and exception reviews in this category, each conclusion has to be traceable to a specific ticket, letter, or piece of evidence text. When the cited source and the original text disagree, the report figure cannot be adopted directly.
A citation returned by knowledge base search is not an authorization, and it is not proof that material has passed review. Retrieval and approval are separate facts.
Constraints at the citation tracing step
The knowledge base search node can associate one or more knowledge bases that share the same vector model. Its output is an array of citations whose length may be zero; when nothing is found, that output path still runs through. Retrieved results pass through question optimization, semantic or full-text retrieval, RRF fusion, ReRank reordering, similarity filtering, and citation upper-limit trimming before being returned. The minimum relevance value is a filter threshold and takes effect only in semantic retrieval or when result reordering is used; it does not perform permission or entity isolation.
Hypothetical paper exercise sample. A shop receives a regulatory inquiry letter. The letter body states: "Please submit a written reply together with the corresponding order evidence within ten working days from the date of receipt of this letter." The issue date shown is 3 March 2025. The letter does not state a date of receipt. A ticket record contains a draft sentence: "This shop received the inquiry letter on 3 March and now submits the rectification report." The question to decide is whether the "date of receipt" in that draft sentence has source support.
The letter body and the draft sentence are the only materials supplied for this exercise. No signing record, delivery receipt, or acknowledgement is provided, so the receipt date remains unsupported. The draft sentence does not become a verified fact because it appears in a ticket.
Configuration and implementation checks
| Configuration or check item | Use on this page | Condition to apply |
|---|---|---|
| Associated knowledge bases | Limit the retrieval scope | Check the selected knowledge bases and material versions first; only knowledge bases with the same vector model may be associated together |
| Citation content output | Carry retrieval results forward | Output is an array; length may be zero; an empty array does not mean the workflow has errored out |
| Citation input handover | Connect to the answering step | After retrieval output is passed to the answering step, compare each answer item against the cited original text |
| Citation display check | Verify source evidence | Use a traceable sample to check the displayed original text; a fluent answer does not replace source evidence |
| Material timeliness check | Separate versions | Test sources with the current version and a historical version; whether a citation is adopted still depends on review against the effective version |
| Minimum relevance | Filter low-relevance results | Takes effect only in semantic retrieval or when result reordering is used; it is not a permission filter |
For this page, the associated knowledge bases must be checked for the same vector model identifier before they are linked. The citation upper limit is expressed in tokens, not in a top-k item count. RRF is the fusion step for mixed retrieval results, not a toggle that is switched on. Thinking budget is a reasoning budget and is not the citation token upper limit.
Three common errors
- Treating a retrieval hit as an approval. Input: the citation array returned by retrieval. Comparison: the authorization status field on the ticket or letter. If a citation exists but the authorization status field is empty, the record can only be marked as pending verification; it cannot be written as approved.
- Treating a date in a draft sentence as a verified fact. Input: the draft sentence above and the inquiry letter body. Comparison: the "date of receipt" against the letter's issue date. These are different fields. Because the letter does not state a receipt date, the draft sentence is unsupported. The issue date and the receipt date are not the same field, and only mutually exclusive values in the same field would count as a contradiction.
- Forcing association of knowledge bases with different vector models. Input: the vector model identifiers of the knowledge bases to be associated. Comparison: whether those identifiers match. If they differ, exclude them. Do not run a cross-model retrieval to infer missed recalls or redundant results; record only the actual observation.
How to verify the setup
- For the draft sentence in the sample, input the inquiry letter body and the draft sentence. Compare the "date of receipt" against the letter's issue date and check whether they are the same field. The draft sentence is supported only if the letter body states a receipt date that matches the draft. In this sample no receipt date is stated, so the item is recorded as pending verification.
- For empty-citation handover, the paper input is a citation array of
[]. Compare whether the downstream answering step still executes and does not report an error. Empty-result handover is satisfied only if the downstream step continues as designed. This is a paper check of the downstream design; it does not claim that a real query returns an empty array. - For knowledge base association, input the vector model identifiers of two knowledge bases to be associated. Compare whether the identifiers match. Only matching identifiers allow joint association. When identifiers differ, exclude them and record the actual observation without inferring runtime behaviour.
These checks describe what to compare and what remains unknown. No test, deployment, approval, refund, or other business action has taken place. Where the letter does not state a receipt date, the evidence gap stays open.