What the documents look like
Weighing and sorting happen inside the warehouse fulfilment stage. The centre of the record is weight and specification, and the material set for operating reports and exception review is the sorting task sheet, the weighing label, the short-weight compensation sheet and the quality inspection record. Sources named in the category brief are WMS, electronic scales, the order system and inspection photographs. Short shipments lead to compensation, gross-margin loss and poor customer feedback, so a report or review that misstates a document status carries direct commercial consequences. These materials update per order and per day; a request that pulls them needs to keep the document subject, the document status and the valid scope separate.
Constraints at the HTTP integrations step
The HTTP module sends a request to a target address. Params, Headers and Body can each reference variables with {{}}, and the URL can also reference a variable. Variables come from global variables, system variables or the output of an earlier node. Return values are extracted with JSON path syntax, and when the output type is set to string the value at the corresponding key is converted to a JSON string.
The chat endpoint uses APIKey authentication. When calling chat/completions, passing body.appId in the request body is recommended; Authorization: Bearer <apiKey>-<appId> is also supported, in which case body.appId is not required. The appId priority order is body.appId, then <apiKey>-<appId>, then the appId associated with the apikey under the older compatibility path. Some SDKs need v1 added to BaseUrl and some do not; a 404 can be retried with v1 added. The model, temperature and similar fields passed in are ineffective because the orchestration decides them. Under test mode, chatId and responseChatItemId do not exist. A retrieval relevance score does not serve as a permission or entity filter, and retrieval on its own does not constitute authorization.
Hypothetical drill sample, clearly labelled: a front warehouse sorting task sheet A-0618-07 dated 18 June 2025 records two units of broccoli at the 500-gram specification; the sorter has weighed and labelled them, and the quality inspection supervisor has not yet reviewed. Short-weight compensation sheet B-0618-03 records a customer report of an 80-gram shortfall and its status is pending review. The draft sentence under review reads: "A-0618-07 has completed quality inspection review, and the compensation on B-0618-03 has been executed." The draft does not match the source. The task sheet records labelling only, so review is not complete, and the compensation sheet carries only a pending-review status.
Configuration and implementation checks
The table below concerns the sample above. It lists settings the official documents describe, or implementation checks, together with the starting value used in this hypothetical drill. The starting values are comparison points for this drill and are not official defaults.
| Configuration item | Hypothetical starting value in this drill | How to check it |
|---|---|---|
| Request address and method | Address and method taken from the external interface document | Compare the address and method against the documented contract |
| Params and Headers | Authorization references a credential variable | Check the request header value against the documented authentication method |
| Body | JSON structure written to the interface contract | Compare the field structure against the external interface |
| Variable references | Document number and status variables | Check each value actually passed in |
| Return value extraction | key configured by JSON path | Compare the extracted value against the response structure |
| Debug identity fields | chatId absent means test mode | Check whether the field exists; do not treat this as a business fault |
Three common errors
- Writing a pending-review status as executed. In the sample, B-0618-03 is pending review while the draft says the compensation has been executed. The input is the status field of the compensation sheet, the comparison object is the status stated in the source, and the observable result is that the two do not agree.
- Writing labelling as completed review. In the sample, A-0618-07 records labelling only, while the draft says quality inspection review is complete. The input is the task sheet record, the comparison object is the review field, and the observable result is that the source does not state a completed review.
- Treating retrieval or a form as authorized execution. In the sample, retrieving a document does not mean the reviewer holds review authority. The input is the retrieval result, the comparison object is the authorization boundary, and the observable result is the absence of an authorization record.
How to verify the setup
- Input the sample source text and the draft sentence. Compare the review conclusion in the draft against the record on task sheet A-0618-07. When the draft contains a statement of completed review that the source does not state, the conclusion is unsupported.
- Input the status field of short-weight compensation sheet B-0618-03. Compare "pending review" against "executed". When the two values are mutually exclusive for the same subject, period and field, the result is a contradiction and the item stays open for verification.
- Input an explicit paper empty result
[]. Compare how the downstream design handles the empty value. When the output does not treat the empty result as authorization or as successful execution, the behaviour matches the expectation.