What the documents look like
On-demand retail fulfillment processes online orders by the minute. Store inventory, picking lists, out-of-stock reports and delivery exception forms update within each order rather than on a periodic stocktake cycle. Product information and selection questions must connect these minute-level materials to a model so that pickers can ask about specifications, substitutes and stock locations. The sources sit across platform backends and store systems, so model connection and configuration must let answers rest on checkable evidence for substitution and picking paths.
The materials this page addresses are the immediate order, the picking list, the out-of-stock report and the delivery exception form. Each carries a business stage and an authorization state. A submitted handover request is not an approved one, and an offered option is not an executed action.
Constraints at the model connections step
From V4.17.0, AI Proxy is a required dependency. The App and any Admin service that reuses App/Service configuration must both set AIPROXY_API_ENDPOINT and a non-empty AIPROXY_API_TOKEN; missing either item causes startup validation to fail. AIPROXY_API_ENDPOINT takes the service root address without a trailing /v1. Legacy OpenAI environment variables or a single-model custom request address cannot replace these two settings. The system needs at least one language model and one index model to operate normally. Model capabilities are checked against the inputs the application requires, with multimodal, tool calling and thinking settings verified separately. Application parameters are set in the application or node configuration, and same-named fields on a compatible chat interface are not assumed to override orchestration.
A relevance score is not a category, entity or permission filter, and retrieval does not grant authorization. Same-model retrieval association retains its documented conditions; libraries built on different vector models cannot be jointly associated, and the check is to confirm identifiers and exclude non-matching libraries rather than to run a cross-model search.
Hypothetical drill sample, clearly labelled and not drawn from real records: a store duty supervisor submits a handover request. The attached material text reads: "Store picking list for today, line 12: fresh milk 950 ml, stock 3 units, location A3-2; out-of-stock report item 4: no substitute arrival for the same specification before 14:00 today." The proposed determination sentence is: "This store can commit same-day delivery of a substitute for fresh milk 950 ml today." The request is pending review and has not been authorized for execution.
Configuration and implementation checks
| Configuration item | Hypothetical starting value for this page | Comparison method and reason |
|---|---|---|
AIPROXY_API_ENDPOINT | http://aiproxy:3000 | Documented root-address example; confirm no /v1 is appended and compare against startup validation |
AIPROXY_API_TOKEN | Empty, required | Must match the server configuration and be non-empty; an application API key does not satisfy this |
| Language model | One enabled | The system needs at least one language model; confirm enabled state on the model provider page |
| Index model | One enabled | The system needs at least one index model; confirm the model ID matches the channel model |
| Model capability | Checked against this task's inputs | Verify multimodal and tool calling separately; same-named fields are not assumed to override orchestration |
| Application parameters | Set in the application or node configuration | Application-side parameters are independent of compatible chat interface fields |
These are implementation checks tied to this page's materials, not official defaults beyond the documented example address.
Three common errors
- Entering
AIPROXY_API_ENDPOINTas a full request address that includes/v1. Comparison against startup validation shows the endpoint does not match the required root form. - Using an application API key in place of
AIPROXY_API_TOKEN. Comparison against the server configuration shows the token differs or is empty. - Enabling only a language model while the index model stays disabled. Comparison of the enabled model list against the requirement for at least one of each type shows a missing type, and startup validation fails.
How to verify the setup
- Enter the hypothetical sample material text and the proposed determination sentence, then compare whether the sentence exceeds what the material states. The material records stock of 3 units and no substitute arrival before 14:00; it does not record a same-day delivery commitment. The sentence is unsupported and remains pending verification.
- Enter
AIPROXY_API_ENDPOINTandAIPROXY_API_TOKEN, then compare whether the endpoint is a root address and whether the token is non-empty and consistent with the server configuration. Both conditions must hold for startup validation to pass. - Enter the enabled model list and compare whether at least one language model and one index model are present. A missing type fails startup validation and is recorded as not passed.
Where a document field is not supplied, the value stays unknown. No test, deployment, approval, refund or other business action is claimed to have occurred.