What the documents look like
Brand outlet discount operations exist to clear out-of-season stock. The category does not maintain a new-arrival full-price image, so the material behind any draft sentence comes from four document types: discount approvals, slow-moving stock lists, price tags, and gross-margin worksheets. The working data arrives from ERP stock age, POS sales, outlet mall campaigns, and historical discount records, refreshed weekly or monthly. Copy drafting draws on these documents, and the drafter first needs to confirm whether a given discount or slow-moving statement has matching material before writing it into a sentence.
In FastGPT v4.17.2, the form input node is a user interaction node. When triggered, the conversation enters an interaction state, the workflow state is recorded, and execution continues after the user completes the interaction. After submission, the node collects the entered form values and passes them to later nodes; it does not carry business database functions of its own. The user selection node also triggers the interaction state and executes a branch according to the user's choice. Both nodes handle interaction and value collection only. Field validation, business approval, write-back, and batch operations need separate implementation in external systems or later nodes.
Constraints at the interactive forms step
The fields a copy-drafting form must collect are decided by this task; the node itself does not prescribe a field set. Discount approvals, slow-moving lists, price tags, and gross-margin worksheets are distinct documents, and the form can only collect entered values. Whether material is genuine, whether authorization exists, and whether approval has passed all require checking in a later step.
Hypothetical paper exercise: on 10 March 2025, an outlet clearance group prepares a slow-moving list. The material text reads: "This batch of SKUs is 2023 autumn/winter styles, stock age over 18 months." "Discount approval number QH-2025-0310, approval status is pending review." "Price tags planned at 50% of list price, valid until 30 April 2025." The draft sentence to be judged is: "This batch has received 50% clearance approval and is executed across all channels." The subject, object, status, and validity scope needed for judgment are all present in the source text. The approval status is pending review, while the draft sentence states approval has been granted; the two values are mutually exclusive on the same approval number and the same discount magnitude, so the sentence is not supported by the material. The price tag validity date states only the period for tag execution; it does not indicate a supply period or that approval has passed.
Configuration and implementation checks
The table below lists implementation checks tied to the exercise above. Where a value appears, it is a hypothetical starting point for this exercise, used for comparison with the material text, and does not represent an official default.
| Configuration or check item | Hypothetical starting point | Comparison method and reason |
|---|---|---|
| Required form fields | Approval number, approval status, discount magnitude, validity period | Without any one of these, the draft sentence cannot be compared field by field against the material; these four cover the information the judgment needs |
| User selection branches | Material supported, material not supported, pending verification | Three values cover supported, missing, and contradictory situations, avoiding a two-value scheme that hides pending verification |
| Interaction state handling | Wait for submission after trigger, pass values after submission | Checked against the documented form input node behavior; no automatic submission or attempt-count control is set |
| Fields for sentence-to-material comparison | Same approval number, same discount magnitude, same validity period | Judgment requires the same subject, same period, and same field; comparison across different fields does not constitute a contradiction |
| Empty-result handling | Paper input [] checked against downstream design | Check the downstream design for empty values; do not claim that a fictional rule search necessarily returns [] |
| Approval action ownership | External system or later node | The node collects values; approval, write-back, and batch operations require separate implementation |
Three common errors
- Treating pending review as approved. The input is approval number QH-2025-0310 with status pending review, while the draft sentence says "has received 50% clearance approval." Compare the approval status field on the same approval number; the two values are mutually exclusive. The observable result is that the material status and the draft wording disagree, and the sentence is not supported. Pending review does not demonstrate a lack of operating permission, and it is not the same as approval passing.
- Treating the price tag validity date as a supply period. The input is "valid until 30 April 2025," while the draft sentence says "executed across all channels." The comparison object is the scope to which the validity period applies; the source text constrains only price tag execution and states no channel scope or supply arrangement. The observable result is that the draft sentence introduces a scope the source text does not state, so it is not supported.
- Using relevance or retrieval results in place of an authorization check. The input is a slow-moving list entry returned by one search, and the draft sentence says "authorized for publication." The comparison object is the source of authorization; retrieval results and similarity scores do not constitute authorization or entity isolation. The observable result is that the authorization field is missing from the material, recorded as an evidence gap, and not written as verified or as denied.
How to verify the setup
- Test the primary task with the exercise material and draft sentence. Input the three source sentences and "This batch has received 50% clearance approval and is executed across all channels," then compare the four fields: approval number, approval status, discount magnitude, and validity period. The expected observation is that the system or a human reviewer points out the mutually exclusive status field and the unstated channel scope, and the draft sentence is judged not supported. This is a paper expectation and was not actually run.
- Test the missing item and its completion. Input a form value that omits approval status, and check whether the values passed after submission are complete; then add the status and check again. The expected observation is that with the item missing, downstream cannot complete the same-field comparison, and after completion the comparison is possible while the conclusion still depends on the status value.
- Test empty-result handling. Input the paper empty value [] and check the downstream node's design for empty values. The expected observation is that downstream handles the empty branch and records the gap, produces no fabricated entry, and does not assert that a search necessarily returns empty. Libraries under different vector models cannot be associated together; check the identifier first, exclude those that do not match, and record the actual observation without inferring missed recall or redundancy from it.