What the documents look like
Brand after-sales repair work turns on warranty responsibility and physical repair locations. The working materials are warranty cards, repair work orders, inspection reports and replacement approvals. These differ from platform refund flows, which is why copy drafting here has to trace back to specific fields rather than to a refund status. Source material comes from the after-sales system, store orders, repair location feedback and the national three-guarantee catalogue.
Copy that mentions warranty scope, repair turnaround or replacement conditions must be traceable to a field in those materials. The model connection step therefore has to keep channel configuration, model capability and application parameters consistent, so that drafted text can be checked against the source wording. Update frequency does not change this page's function.
Constraints at the model connections step
From V4.17.0, AI Proxy is a required dependency. The App service and any Admin service that reuses App/Service configuration must set AIPROXY_API_ENDPOINT and a non-empty AIPROXY_API_TOKEN. A missing value in either one causes startup validation to fail. AIPROXY_API_ENDPOINT holds the AI Proxy service root address and does not carry /v1. Legacy OpenAI environment variables or a single-model custom request address do not substitute for these two settings. The system needs at least one language model and one index model to operate. Model capability is checked against the input the application actually needs, with multimodal input and tool calling verified separately. Application parameters are set in the application or node configuration; a same-named field on a compatible chat interface does not override orchestration.
Hypothetical paper exercise sample: a brand after-sales centre agent drafts product copy from one repair work order. The work order text reads: "Work order WX-2026-0417, accepted 17 April 2026, product is Model A water purifier, fault described as odour in output water, inspection conclusion is filter element expired and requires replacement, warranty status is whole-unit warranty in force, filter element is a consumable and outside warranty scope." The sentence to be assessed reads: "The Model A water purifier whole unit is within warranty, and filter element replacement falls within warranty scope and can be replaced free of charge." Subject, object, period and field can all be compared against the work order text.
Configuration and implementation checks
| Configuration item | Use on this page | How to check |
|---|---|---|
AIPROXY_API_ENDPOINT | Supplies the AI Proxy root address for copy drafting model calls | Compare the entered value against the documented root address form; confirm it does not contain /v1 or /chat/completions |
AIPROXY_API_TOKEN | Server-side administrator token | Compare against the server-side configuration; confirm it matches and is non-empty |
AGENT_ENGINE | Selects the Agent engine, fastAgent or piAgent | Compare against the engine this page's drafting flow requires |
CHAT_MAX_QPM | Chat QPM limit | Compare against any plan limit; the plan limit takes precedence |
LLM_REQUEST_TRACKING_RETENTION_HOURS | LLM request tracking retention | Compare the retention window against troubleshooting needs |
SYSTEM_MAX_STRING_LENGTH_M | Maximum characters for synchronous string handling | Compare the concatenated work order and warranty card text against the configured range |
The values shown for comparison in this table are hypothetical starting points for this paper exercise only. They were chosen to make comparison against documented values possible and are not official recommended values.
Three common errors
- Entering a full model request address into
AIPROXY_API_ENDPOINTwith/v1appended. Input: the hypothetical sample work order text and the drafted sentence, plus the entered endpoint value. Comparison: the entered value against the documented root address form. Observable result: the entered value contains/v1or/chat/completions, which the documented form does not. - Substituting a legacy OpenAI environment variable for
AIPROXY_API_TOKEN. Input: the server-side administrator token and the configuredAIPROXY_API_TOKEN. Comparison: whether the two match and whether the token is non-empty. Observable result: startup validation passes or fails accordingly. - Treating the warranty-scope wording in the drafted sentence as if it were the work order conclusion. Input: the work order text "filter element is a consumable and outside warranty scope" and the drafted sentence "filter element replacement falls within warranty scope". Comparison: the values of the same field. Observable result: the two values are mutually exclusive on that field.
How to verify the setup
- Input the hypothetical sample work order text and the drafted sentence. Compare the drafted sentence "filter element replacement falls within warranty scope" against the source wording "filter element is a consumable and outside warranty scope". If the same field carries mutually exclusive values, the drafted sentence is not supported and returns for revision. Expected result: this sentence does not pass. Not yet tested.
- Input the entered
AIPROXY_API_ENDPOINTvalue and the documented root address form. Compare whether/v1is appended. If it is appended, the configuration is wrong. Expected result: the value must be corrected to the root address. Not yet tested. - Input
AIPROXY_API_TOKENand the server-side administrator token. Compare whether the two match and whether the token is non-empty. If they differ or the token is empty, startup validation fails. Expected result: the configuration must be aligned. Not yet tested.
Where the work order does not state a field, that field remains unknown. A pending request is not an approval, and an offered option is not an executed action.