What the documents look like
A fresh-food concession counter in a hypermarket is commonly run by a supplier who stations staff on site and shares revenue with the store. Copy material for products and marketing is therefore scattered across four document types: the concession contract, the daily sales report, the shrinkage record, and the food purchase inspection record. The concession contract defines the scope of cooperation that may be described publicly and the settlement basis. The daily sales report provides the actual sales figures for a given day. The shrinkage record captures write-offs and their allocation. The food purchase inspection record carries batch, supplier, and inspection status.
Whether a proposed promotional sentence holds depends on whether these materials support it with the same subject, the same period, and the same field. Data updates daily and per batch; that frequency describes the stated rhythm of the source materials and does not imply any batch synchronization, real-time trigger, or full refresh mechanism.
Constraints at the model connections step
Model connection requires AI Proxy as a mandatory dependency. Both App and Admin must set AIPROXY_API_ENDPOINT and a non-empty AIPROXY_API_TOKEN; missing either item causes startup validation to fail, and legacy OpenAI environment variables or a single-model custom request address cannot substitute for these two settings. Model configuration is performed on the administrator model provider page. A channel takes a BaseUrl, and model mapping maps the requested model to the actual upstream model.
Model probing applies only to currently enabled system models. After a first failure the system retries at most 3 times, so a single round makes at most 4 attempts; only when all 4 attempts fail is the model judged abnormal. The probe interval accepts integers from 5 to 60 minutes. The alert Webhook pushes one notification per model when the status changes at an edge. A retrieval library's vector model identifier must match the query side; across different vector models there is no shared association condition.
Hypothetical exercise sample. Source text one: "Party A and Party B sign a concession contract for the fresh-food counter; Party B operates on site and shares revenue by sales amount; the contract term runs to a stated year and month." Source text two: "The daily sales report for a stated date shows a stated sales amount for this counter." Source text three: "The food purchase inspection record for this batch states the supplier and the inspection status." The proposed sentence to be judged: "The fresh food at this counter is supplied directly by the contracted supplier, and the same-day sales figure can be checked." Judging requires checking whether subject, period, and field agree.
Configuration and implementation checks
| Configuration item | Hypothetical trial starting value | Reason and comparison object |
|---|---|---|
AIPROXY_API_ENDPOINT | Enter the AI Proxy service root address without appending /v1 | Compare against the documented root-address format and confirm no extra path |
AIPROXY_API_TOKEN | Enter a non-empty administrator token consistent with the server | Compare against the documented non-empty consistency condition and confirm it is not replaced by an application API key |
| Model mapping | Map the requested model to the actual upstream model | Compare against the channel's available model list and confirm the mapping target exists |
| Probe interval | Hypothetical trial starting value of 30 minutes | Falls inside the documented 5 to 60 minute integer range; compare against both bounds |
| Alert Webhook | Enter a receiving address and use the test action | Compare against the documented simulated error and simulated recovery test messages |
| Retrieval library vector model identifier | Record the vector model identifier used when the library was built | Compare against the query-side identifier and exclude the library when the identifiers differ |
Three common errors
- Writing
AIPROXY_API_ENDPOINTas a full request address that includes/v1or/chat/completions. Check: read the variable value, compare it against the documented service root address format, and treat any appended path as not satisfying the condition. - Using an application API key in place of
AIPROXY_API_TOKEN. Check: read the variable, compare it against the documented non-empty administrator token condition, and treat an empty value or a different source as not satisfying the condition. - Treating the phrase "supplied directly by the contracted supplier" in the proposed sentence as already supported by the materials. Check: compare the proposed sentence field by field against the three hypothetical source texts; when the materials do not state a direct-supply field, the wording remains unsupported and the evidence gap is preserved.
How to verify the setup
- Input the three hypothetical source texts and the proposed sentence, then compare subject, period, and field for agreement. The sentence is satisfied only when the three materials share the same subject, the period covers the day named in the proposed sentence, and the direct-supply field is explicitly recorded; when the field is not stated, record it as unsupported.
- Input the retrieval library's vector model identifier and the query-side identifier, then compare whether the two agree. When the identifiers agree, the library may take part in retrieval; when they differ, exclude the library and record the actual observation, without running a cross-model retrieval to compare missed recalls.
- Input the probe interval value and the Webhook address, then compare whether the interval falls inside the 5 to 60 minute integer range and whether the Webhook test receives both the simulated error and simulated recovery messages. When both conditions hold, the configuration condition is met; when the messages are not received, record it as pending verification and do not infer a failure cause.
This page addresses only one scope: fresh-food concession counters in hypermarkets, product and marketing copy drafting, and model connection and configuration. Whether a specific promotional sentence in product copy is supported by the materials must not be converted into an enterprise identity certification tutorial. The hypothetical sample above is a clearly labelled paper exercise; no test, deployment, approval, or other business action is claimed to have occurred.