What the documents look like
A Shopify standalone store leaves payment configuration, shipping rates and fulfillment services to the merchant. The platform does not run a single automatic return process across all orders. For contract and transaction material comparison, this difference matters because the comparison has to read merchant-side order details, shipping templates, shipping labels and refund records together, and these materials do not share one platform-wide naming or status convention.
The data sources named in the category brief are Shopify Admin, payment gateways, carrier systems and inventory location data. Updates arrive per order and daily. Order details, shipping templates, labels and refund records are produced at different stages, so field names and status values can differ between sources. A comparison workflow needs to keep the original wording of each source rather than normalize it away.
Constraints at the workflow routing step
Workflow routing in this task handles material collection, interactive confirmation and result handoff. When a form input node triggers, the conversation enters an interaction state and the workflow state is recorded. After the user fills required information and submits, the node passes that information to later nodes. A user selection node, when the user makes a choice, executes the corresponding branch. The interaction state record is node behavior; it does not act as a global state switch. Retrieval results do not constitute authorization, and a relevance score does not filter by subject, object or permission. Order writes, document backfill and external status synchronization require a separate business interface or later process. This page does not assume hidden triggers or automatic synchronization.
Hypothetical paper test case: the order detail text supplied for comparison reads "Order H-2041, order date 2025-03-04, item A quantity 2, payment status paid, fulfillment status unshipped." The shipping template text reads "Template T-07 applies to East China, first weight 1 kg, continued weight charged per 0.5 kg." The refund record text reads "Refund R-118, related order H-2041, refund amount 80 yuan, status pending review." The draft sentence to be judged is "Order H-2041 has completed the refund and terminated fulfillment." This draft sentence shares the same subject and the same related order as the refund record, but the status field is pending review. Signing, approval and execution are three distinct facts.
Configuration and implementation checks
Configuration below follows the comparison chain in the sample and lists only documented node behavior and implementation check items. The numeric values are starting points for this hypothetical trial, used to explain the comparison method; they are not official default values.
| Configuration item | Setting or check content | Relation to the sample |
|---|---|---|
| Form input node | Set order number and material type as required fields; pass submitted values to later nodes | Collects H-2041 and the material type for comparison |
| User selection node | Set branches for "materials consistent," "materials inconsistent," "insufficient evidence"; verify each branch connection one by one | Gives the draft sentence a judgment branch |
| Interaction state check | Confirm that triggering enters the interaction state and records workflow state | Check waiting for input and continuing after submission |
| Output handoff check | Explicitly pass form information to later retrieval, judgment or interface nodes | Compare output fields against input fields item by item |
| Retrieval scope check | Check that the target retrieval scope contains orders, templates and refund records; record missing items | Missing material records are evidence gaps |
| External operation separation | Order writes and external status synchronization connect to a separate business interface or later process | Do not treat the comparison conclusion as an execution result |
Three common errors
- Treating the refund record status "pending review" as a completed refund. The input is the status field of refund record R-118; the comparison object is "has completed the refund" in the draft sentence; the observable result is that the status values differ, so the judgment is unsupported.
- Treating a retrieval hit as authorization. The input is the order material returned by retrieval; the comparison object is the subject and object referenced by the draft sentence; the observable result is that retrieval returns text only and lacks an approval field, so the judgment remains to be verified.
- Treating a user selection branch as business execution. The input is the "materials consistent" choice made at the user selection node; the comparison object is the order fulfillment status field; the observable result is that the selection action does not change the order status, so the judgment is a comparison conclusion and does not constitute fulfillment execution.
How to verify the setup
- Using the sample text and the draft sentence, input order number H-2041, refund R-118 and the status field; compare "has completed the refund" in the draft sentence against the refund record status "pending review"; the observable result is that the same subject and same field hold mutually exclusive values, so the judgment is unsupported.
- Using the sample text, input shipping template T-07 and the fulfillment status of order H-2041; compare the template's applicable region against the order fulfillment status field; the observable result is that the template does not contain a fulfillment status, so the judgment is insufficient evidence and the gap is recorded.
- Using a clearly labelled paper input of an empty result, check whether the downstream design accepts an empty value; the observable result is that the downstream node receives empty input and keeps a to-be-verified marker, so the handoff is correct. This does not claim that a fictional rule retrieval necessarily returns an empty result. Libraries built on different vector models cannot be associated together; check the identifier first and exclude those that do not match.