What the documents look like
Group-buy presale collection takes orders first and organizes procurement and fulfillment afterward. Store service and internal training answer store questions from four document groups: the presale product table, collection results, the procurement plan, and the cut-off notice. Materials update daily and per session, so whether a retrieval result reflects the current cut-off wording directly shapes the answer given to a store.
Within one batch of materials, the product table describes specifications and presale status, collection results record orders already received, the procurement plan reflects the purchasing arrangement after cut-off, and the cut-off notice gives the time boundary. Store staff commonly ask whether a product can still be ordered and how a session is handled after cut-off. Retrieval needs to bring back the subject, the session, and the status together.
Constraints at the retrieval step
The knowledge base search node can associate one or more knowledge bases that use the same vector model. Knowledge bases built on different vector models cannot be associated together. Search mode is compared across semantic, full-text, and hybrid retrieval against the test question; no single mode is presumed suitable for all materials. The citation limit is measured in tokens. The minimum relevance value takes effect only under semantic retrieval or when result reranking is used. Question optimization performs coreference resolution and question expansion, and the subject must be checked for unintended change. The node outputs cited content as an array whose length can be 0; when nothing is retrieved, the output path still completes. The retrieval step does not perform authorization decisions. When operations handover material is pending review, a retrieval hit does not mean approval has been granted.
Hypothetical drill sample: the store training material text provided for checking reads "Entry A-07 in this session's presale product table: specification 500 grams, presale status is cut-off. Cut-off notice N-03 states this session's cut-off time is 12 March 18:00. Procurement plan P-02 records that the purchase quantity for this session is pending confirmation." The draft sentence to be judged reads "Stores may continue to add orders for A-07 after 12 March 18:00." No authorization document accompanies this request.
Configuration and implementation checks
| Configuration item | Value or check | Relationship to this page's sample |
|---|---|---|
| Associated knowledge bases | Select the presale product table, collection results, procurement plan, and cut-off notice bases that use the same vector model | Bases on different vector models cannot be associated together; verify identifiers first and exclude non-matching ones |
| Search mode | Use full-text or hybrid retrieval for specifications and session numbers; use semantic retrieval for natural-language phrasing | Number-type queries depend on keyword hits; natural-language phrasing depends on semantic closeness |
| Citation limit | For this paper exercise, set a starting value covering one product entry plus one notice in tokens, and compare against a shorter value for content completeness | This starting value is set for this paper exercise, not an official default |
| Minimum relevance | Label relevant and irrelevant samples first, then adjust the threshold | This value takes effect only under semantic retrieval or result reranking |
| Question optimization | Compare the original question with the optimized question and check the session and product subject | Coreference resolution can change the subject and must be checked sentence by sentence |
| Result reranking | Confirm the rerank model is available and enabled per configuration | Reranked scores can be used for filtering; when not enabled this path is not taken |
Three common errors
- Treating a retrieval hit as authorization. In the sample, procurement plan P-02 records the purchase quantity as pending confirmation. Retrieving that entry only shows the material exists. The draft sentence claims orders can be added after cut-off, and the material does not state that authorization. This is unsupported.
- Associating knowledge bases that use different vector models. If a product table base and a notice base built on different vector models are associated together, verify identifiers first and exclude non-matching ones, then record the actual observation. Do not infer missed recalls or redundant results.
- Treating the cut-off time as a supply period. Cut-off notice N-03 gives only the cut-off moment. The draft sentence extends it into continued ordering after cut-off. The two have different subjects and fields, and the claim cannot be judged as holding on this basis.
How to verify the setup
- Input the hypothetical material text and the draft sentence, then compare the draft's claim of "adding orders after cut-off" against the original text's "presale status is cut-off" and "purchase quantity pending confirmation." If the material does not state that orders can be added after cut-off, judge the claim unsupported and record the evidence gap.
- Input the specification and session number for the same session, compare the cited entries returned by full-text retrieval and semantic retrieval, and observe whether the number-type query hits the corresponding entry. When it does not hit, record it as pending verification; do not guarantee a first-position hit.
- Use an explicit paper input of an empty array to check the downstream handling design. Compare whether the path continues when the node's cited content length is 0, and observe whether downstream treats it as an empty result. Do not claim that a fictional rule's retrieval necessarily returns an empty array.