What the documents look like
Near-expiry markdown handling works on the remaining sellable shelf life of shelf stock. It does not weigh individual items the way weighing and sorting does. The daily document surface for this category includes the near-expiry list, the markdown approval, the damage report, and the promotional price tag. The data sources named for this category are POS stock age, store inspection, ERP inventory, and quality inspection records. The three most time-consuming steps are identifying near-expiry batches, approving markdowns, and isolating unsellable goods. Copy drafting for products and promotions sits inside this chain: the copy basis comes from the near-expiry list and the markdown approval record, and the routing state of a draft is carried by workflow nodes. Data updates on a daily or shift basis. This page does not derive batch synchronization or real-time trigger mechanisms from that update frequency.
Constraints at the workflow routing step
The Form Input node is a user interaction node. When it triggers, the conversation enters an interaction state, the workflow state is recorded, and after the user fills in required information and submits, the node collects the form information and passes it to subsequent nodes. The User Selection node is also a user interaction node. When it triggers, the conversation enters an interaction state, the workflow state is recorded, and when the user makes a selection, the node judges that selection and executes the "yes" branch. Copy drafting orchestration therefore turns on these two interaction nodes: Form Input collects the fields needed for drafting, such as batch, markdown range, and price tag wording, and User Selection confirms the branch direction. The nodes only collect and pass information. Order writes, document backfill, and external state synchronization need a separate business interface or a later process. The nodes do not perform those actions themselves.
Hypothetical paper exercise sample: the source text reads, "Near-expiry list entry: the object is one batch of low-temperature yogurt in a local store cold cabinet, the state is two days remaining before the shelf-life deadline, and the scope is all on-shelf quantity of that batch listed in the same day's inspection record." The proposed draft sentence reads, "This batch of yogurt has passed markdown approval and may be labeled externally at the discounted price." The comparison checks the proposed sentence against the source text. The source text records only the subject, object, state, and scope of the list entry. It does not record a markdown approval conclusion. The proposed sentence remains pending verification.
Configuration and implementation checks
| Configuration item | Value or check item on this page | Basis and explanation |
|---|---|---|
| Form Input node | Collect batch identifier, remaining shelf-life days, on-shelf quantity | The document states that it collects required information and passes it to subsequent nodes |
| User Selection node | Set "continue drafting" and "return for supplementation" branches | The document states that the "yes" branch executes according to the user's selection; verify branch connections item by item |
| Interaction state check | After triggering, confirm the conversation enters the interaction state and the state is recorded | Recording state is behavior of the interaction node; no separate global state switch is set |
| Output handoff check | Explicitly pass form fields to subsequent retrieval or judgment nodes | Compare node output and input item by item for consistency |
| Copy basis field | Markdown approval conclusion field | For this exercise the hypothetical starting value is "pending approval," compared with the "has passed" claim in the proposed sentence, because this field determines whether the copy may cite a markdown conclusion |
| External operation division | Document backfill and state synchronization | A separate business interface or later process is required; the node does not carry this |
Three common errors
- Treating fields collected by Form Input as verified business conclusions. The input is a value filled in by the user. Business verification and approval need downstream implementation. Check method: input the hypothetical sample value "markdown approval conclusion is pending approval," compare it with the proposed sentence "has passed markdown approval," and if the two values are mutually exclusive, judge the proposed sentence unsupported.
- Treating execution of a User Selection branch as completion of an approval action. Branch execution by the node indicates only the process direction. Check method: input the selection result "continue drafting," compare whether an independent approval record exists in the system, and if only a branch direction record exists, the approval state remains unknown.
- Treating retrieval results as an authorization or entity isolation measure. Relevance does not perform permission filtering. Check method: input a query directed at another store's batch, compare the returned records with the authorized scope for this task, and if returned records fall outside the scope, record that observation without inferring that the system necessarily blocks them.
How to verify the setup
- Verification one: input the source text and the proposed draft sentence from the hypothetical sample, and compare whether the proposed sentence cites a markdown approval conclusion not recorded in the source text. If the proposed sentence contains "has passed" while the source text records only the list state, the sentence is judged unsupported and must be returned for supplementation of the approval basis. The expected result is that the sentence does not enter external copy.
- Verification two: input the set of fields submitted by the Form Input node, and compare it item by item with the fields actually received by subsequent nodes. If a field is missing during handoff, the output handoff check fails. The expected result is a one-to-one correspondence between field names and values.
- Verification three: input an empty-result handoff scenario, using the explicit paper input [] to check the downstream design, and compare whether the downstream node can still distinguish "no record" from "not queried" when there is no retrieval result. Do not claim that retrieval under a fictional rule necessarily returns []. The expected result is that the downstream node retains an evidence-gap marker and does not generate affirmative copy.