What the documents look like
Closure settlement handles exit and asset recovery, the reverse direction from head-office site selection and store opening. Source records come from POS inventory, fixed-asset systems, lease contracts and financial ledgers. The working documents for this category are closure approvals, inventory clearance sheets, lease termination agreements and asset count sheets. They are updated irregularly, per store.
For copy drafting, the practical consequence is that product names, quantities, condition labels and dates in a draft must be traceable to those source documents. An approval record, a clearance sheet and a termination agreement each carry different fields, and a signed status does not follow from a return date. Authorization boundaries and termination conclusions are checked separately, and fields the source does not state remain unknown.
Constraints at the model connections step
Model connections run through AI Proxy. From V4.17.0, AI Proxy is a required dependency. Both the App and any Admin service that reuses App/Service configuration must set AIPROXYAPIENDPOINT and a non-empty AIPROXYAPITOKEN; missing either one fails startup validation. AIPROXYAPIENDPOINT takes the service root address and does not append /v1. Legacy OpenAI environment variables and single-model custom request addresses do not substitute for these two settings.
Model configuration is performed on the administrator model provider page. Channel configuration includes channel name, protocol type, models, model mapping, proxy address and API key. The proxy address takes a BaseUrl, and whether to add /v1 must be confirmed against the provider. The model ID is the value of the model field in the request body and is globally unique; the model name is for display and can be customized. A system needs at least one language model and one index model to work normally. Model capabilities are checked against the inputs an application requires, with multimodal input and tool calling verified separately. Application parameters are set in the application or node configuration; same-named fields on a compatible chat interface cannot be assumed to override orchestration.
Hypothetical paper exercise
The following is paper exercise material and does not involve a real store or entity.
Material one (inventory clearance sheet excerpt): "Store number H-07. Clearance goods: folding storage boxes, 200 units, condition listed as sellable, clearance deadline 2026-11-30."
Material two (lease termination agreement excerpt): "Landlord and tenant intend to process lease termination for store H-07. The agreement lists a handover date of 2026-11-15. Signing status is pending signature by both parties."
Draft sentence to be assessed: "Store H-07 has 200 folding storage boxes, sellable, clearance completed by 30 November, and the lease has been terminated."
Configuration and implementation checks
The table below lists configuration items or check items tied to this page's materials and the copy drafting task. Where a value is marked as a hypothetical starting point for this exercise, the reason for choosing it and the comparison object are stated.
| Configuration item or check item | Value or status in this exercise | Notes and comparison method |
|---|---|---|
| AIPROXYAPIENDPOINT | Hypothetical starting point http://aiproxy:3000 | Documentation requires the root address with no /v1 appended; compare against a form containing /v1 to confirm path level |
| AIPROXYAPITOKEN | Non-empty administrator token | Documentation requires it to match the server side and be non-empty; distinguish it from an application API key and check the source |
| Language model and index model | At least one of each enabled | Documentation states this as the minimum system requirement; check the enabled list covers both types |
| Model mapping | Hypothetical starting point gpt-4o-test mapped to gpt-4o | Documented example form; check whether the request identifier points upstream to the mapped target |
| Channel proxy address | BaseUrl entered, /v1 confirmed per provider | Documentation requires a BaseUrl rather than a full request address; compare against a full-path form |
| Draft field source check | Sentence-by-sentence against the clearance sheet and termination agreement | Check whether subject, object, condition and date range appear in the materials |
Three common errors
- Writing AIPROXYAPIENDPOINT as a full request address containing /v1. Documentation requires the service root address with no /v1 appended. The input is an address containing /v1; the comparison object is the root address form described in the documentation; the observable result is whether startup validation passes and whether the request path is duplicated. The actual validation outcome remains to be observed.
- Using an application API key as AIPROXYAPITOKEN. Documentation requires an administrator token that is non-empty. The input is an application-side key; the comparison object is the administrator token configured on the server side; first check the credential's purpose and server-side consistency, and the actual authentication result remains to be tested.
- Writing a pending signature as a completed termination in the draft sentence. Material two states the signing status is pending signature by both parties. The input is that draft sentence; the comparison object is the signing status field in material two; the observable result is whether the sentence exceeds what the material supports. The source gives no termination effective condition or completion record, so the termination status remains to be verified.
How to verify the setup
The three checks below all use the hypothetical paper exercise above and do not indicate that anything was actually executed.
- Address configuration check. The input is the hypothetical starting point http://aiproxy:3000 together with a second form containing /v1. The comparison object is the documented path requirement for AIPROXYAPIENDPOINT. The condition is met when the root address form matches the documentation description and the form containing /v1 is identified as non-conforming; the startup validation result must be observed and recorded.
- Model mapping check. The input is the hypothetical starting point mapping gpt-4o-test to gpt-4o. The comparison object is the documented distinction between the model ID and the actual upstream model. The condition is met when the request-side model ID and the upstream model are separately identifiable and not merged into one field; the actual channel test output must be observed and recorded.
- Draft field check. The input is material one, material two and the draft sentence to be assessed. The comparison object is the goods name, quantity, condition, clearance deadline and signing status stated in the materials. The condition is met when the quantity of 200 units, the condition sellable and the deadline 2026-11-30 match the materials; the signing stage and termination effectiveness are different fields, and the source gives only pending signature and a handover date with no termination completion record, so "the lease has been terminated" remains to be verified. No conclusion is drawn in either direction.