What the documents look like
Amazon FBM fulfillment means the seller handles delivery and sets the return address without relying on platform warehouse inspection. The working materials for this category are order pick lists, shipment confirmation sheets, return address settings, and logistics tracking sheets. They come from Amazon Seller Central, carrier systems, ERP orders, and customer service email. The product information and selection question-answering task needs to turn product attributes, carrier options, return addresses, and tracking numbers from those materials into a searchable, citable basis for answers. Records update per order and per day; this page does not infer batch synchronization, real-time triggering, or transaction-linked automation from that cadence.
A hypothetical drill sample is used below. The material text reads: "The pick list PO-0310-07 dated March 10, 2026 records product insulated bottle 500 ml, quantity 2, carrier A, and return address Shenzhen warehouse; the shipment confirmation sheet of the same date records tracking number SF0310A and handover date March 11, 2026; the return address setting record shows the Shenzhen warehouse was enabled on March 1, 2026, valid through December 31, 2026." The draft sentence to be judged is: "The return address for this order is the Shenzhen warehouse, tracking number SF0310A, handover date March 11, 2026." This sentence shares the same subject, period, and fields as the material and is judged supported.
Constraints at the model connections step
FastGPT connects to model providers 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 causes startup validation to fail. Legacy OpenAI environment variables or a single-model custom request address cannot replace these two settings. AIPROXYAPIENDPOINT takes the AI Proxy service root address without appending /v1. The system needs at least one language model and one index model to work normally. Model types are language model, index model, rerank model, speech synthesis, and speech recognition. The model ID is the value of the model field in the request body and is globally unique.
For this task, model connections serve product information and selection question answering. Retrieval is a recall process and does not carry permission or entity isolation. Libraries across different vector models cannot be jointly associated; identifiers must be checked first and non-matching entries excluded, with actual observations recorded. An empty result can be checked against an explicit paper input [] to inspect downstream design, without claiming that a fictional rule query must return []. Model capabilities are checked against the input the application requires; multimodal, tool calling, and thinking settings are verified separately. Application parameters are set in the application or node configuration, and same-named fields of a compatible chat interface cannot be assumed to override orchestration.
Configuration and implementation checks
| Configuration item | Hypothetical starting value for this drill | Reason and comparison method |
|---|---|---|
| AIPROXYAPIENDPOINT | http://aiproxy:3000 | Documented as the service root address without /v1; compare against the documented example format |
| AIPROXYAPITOKEN | Non-empty administrator token | Documented as consistent with the server side and non-empty; compare against the AI Proxy server configuration |
| Language model | Enabled and passing channel test | Documented requirement for at least one language model; compare against model test results |
| Index model | Enabled and passing channel test | Documented requirement for at least one index model; compare against model test results |
| Model ID | Consistent with the model in the channel | Documented as globally unique and consistent with the channel model; compare against channel configuration |
| Model status probe interval | 30 minutes | Documented as an integer between 5 and 60 minutes; compare against the allowed range |
The table gives hypothetical starting values for this drill, not official defaults. Material identities for this page remain the pick list, shipment confirmation sheet, return address setting, and logistics tracking sheet, with the drill sample above as the object and the March 2026 dates as the scope. The authorization boundary stays at configuration inspection and paper comparison; no deployment, approval, or business action is claimed.
Three common errors
- Filling AIPROXYAPIENDPOINT with a full address that includes
/v1. Input ishttp://aiproxy:3000/v1; comparison is the documented statement not to append/v1; the observable result is a recorded root-address format mismatch, with the actual request outcome still untested. - Using an application API key in place of AIPROXYAPITOKEN. Input is an application-side API key; comparison is the documented administrator-token requirement; the observable result is a recorded credential-purpose mismatch, with the authentication result still untested.
- Writing the handover date in the draft sentence as March 10, 2026. Input is the draft sentence and the material text; comparison is the shipment confirmation sheet's stated March 11, 2026; the observable result is a date-field inconsistency.
How to verify the setup
- Input the three material sentences and the draft sentence from the hypothetical sample. Compare the return address, tracking number, and handover date in the draft sentence against the same fields in the material. When all three match, the draft is judged supported; when any field differs from the material, it is recorded as unsupported. A field that is not stated, or for which only related-stage evidence exists, is recorded as unsupported or pending verification.
- Input the candidate AIPROXYAPIENDPOINT value and the candidate AIPROXYAPITOKEN value. Compare them against the documented requirements for a root address and a non-empty administrator token. When both are satisfied, only these two configuration checks are recorded as passing, and the call outcome remains untested; when either is missing, a configuration gap is recorded.
- Input the paper empty result
[]and the downstream question-answering design. Check whether the downstream design has an empty-result handling branch. This is a paper check only; the real query result remains to be observed, and no assumption is made that a fictional rule query returns[]. Retrieval does not establish authorization, and a relevance score is not a category, entity, or permission filter.