What the documents look like
On a Shopify standalone store, the merchant configures payment, shipping, and fulfillment services directly; no platform-level automatic return process applies. Order fulfillment documents therefore live in several separate places: order details, shipping rate templates, shipping labels, and refund records. Order and refund records update per order, while shipping rate templates and shipping labels update daily. A copy drafting task that cites these materials must trace every promotional sentence back to a specific document, and must check which version of each cited material is currently effective. The subject, object, status, and validity scope of one document can differ from another, so each item is compared before drafting.
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 reference output is an array, and the length may be zero; an empty array does not mean the workflow has failed. Retrieval results do not constitute authorization and do not constitute business verification. The minimum relevance value is a filter threshold and does not enforce permission or entity isolation. RRF is a fusion step for mixed results, not a standalone switch. The reference cap is measured in tokens, and the thinking budget belongs to model thinking configuration; the two are set in different configuration areas and are not interchangeable.
Hypothetical paper exercise sample: A standalone store merchant holds a shipping rate template effective 1 March 2025, stating "US domestic orders over USD 49 receive free standard shipping; standard shipping is USD 5; estimated delivery is 5 to 7 business days." A separate refund record dated 20 February 2025 states "Order A-1042 refunded due to wrong size; refund amount USD 32; status completed." The draft sentence to be judged is: "US orders over USD 49 ship free, and refunds arrive within one business day."
Configuration and implementation checks
| Configuration or check item | Hypothetical starting point for this exercise | Reason and comparison method |
|---|---|---|
| Associated knowledge bases | Select knowledge bases containing the shipping rate template and the refund record | Compare the vector model identifier of each base; only matching identifiers can be associated together |
| Search mode | Hybrid search | The draft sentence contains an amount and a time expression, so semantic and full-text recall each contribute |
| Reference cap | Hypothetical starting point 3000 tokens | Compare with the length of the sample source texts to confirm both materials fit in full |
| Minimum relevance | Hypothetical starting point 0.3 | Compare with retrieval scores; too low admits irrelevant fragments, too high drops materials |
| Reference output handoff | Pass the reference array to the answering step | Check each answer sentence against the cited source text item by item |
| Material version check | Use the 1 March 2025 version as effective | Compare with the 20 February 2025 version and confirm the effective version is used |
Three common errors
- Treating "refunds arrive within one business day" in the draft sentence as supported by the refund record. The refund record states only the refund amount and the completed status; it does not state a time to arrival. The check inputs the draft sentence and the refund record text, compares whether any wording about time to arrival exists, and records that this expression lacks source support.
- Mixing the reference cap with the thinking budget. The reference cap controls the maximum tokens cited per search; the thinking budget controls model thinking intensity. The check inputs both configured values, compares which configuration area each belongs to, and confirms they are set separately.
- Treating minimum relevance as a permission or entity filter. This value only filters low-relevance results; it does not exclude unrelated businesses or unauthorized records. The check inputs a query unrelated to the sample, compares whether returned results may still contain other business content, and records the actual observation.
How to verify the setup
- Input the shipping rate template text from the hypothetical sample and the draft sentence. Compare whether "US orders over USD 49 ship free" has corresponding wording. When the shipping rate template states that condition, the expression is treated as having source support; when it does not, record the missing source.
- Input the refund record text from the hypothetical sample and the draft sentence. Compare whether "within one business day" has corresponding wording. When the refund record does not state a time to arrival, that expression remains unsupported.
- Input a paper empty array [] to check the downstream handoff design, confirm whether the downstream path still runs with empty references, and record the actual behavior. Do not claim that a fictional rule search necessarily returns an empty array.
This page addresses only Shopify standalone order fulfillment / product and marketing copy drafting / citation tracing. Whether a specific promotional sentence in product copy is supported by materials cannot be converted into an enterprise identity certification tutorial.