What the documents look like
Brand after-sales repair work turns on warranty responsibility and physical repair outlets. Records come from the after-sales system, store orders, repair outlet feedback, and the national three-guarantees catalogue. The materials handled day to day include warranty cards, repair work orders, inspection reports, and replacement approvals. These materials carry different authorisation states during an operations handover. A warranty card is the consumer's evidence for claiming warranty service. A repair work order is the outlet's record of the job. An inspection report is the technical basis for a determination. A replacement approval is an application submitted by a store or after-sales centre. The existence of a material does not mean an approval passed. A submitted application does not mean a replacement was executed.
Constraints at the upgrades step
Deployment and upgrade work covers image updates, execution of upgrade scripts, and configuration migration. The version upgrade notes list two manually triggered scripts: the historical chunk training task migration and the knowledge base rebuild status migration. Neither blocks service startup, and the second depends on the first being complete. Handover management needs to separate four things: behaviour a node already has, fields that can actually be configured, manual comparison actions, and business actions that require a separate system implementation. Retrieval capability does not carry authorisation duties, and a similarity score is not a permission or entity filter.
Hypothetical drill sample: a brand after-sales centre plans to upgrade from an older version to v4.17.2. The source material reads as follows. The warranty card states "full-unit warranty for twelve months, counted from the invoice issue date." The repair work order states "fault description: screen flicker; inspection conclusion: non-human damage; recommendation: replace screen." The inspection report states "inspection date: 10 March 2026; inspection result: display module abnormality." One draft sentence to be assessed reads: "This work order has completed replacement approval, and a machine replacement is agreed." Checking that draft sentence requires comparing whether the materials contain the approving party, approval time, and approval conclusion for a replacement approval. The sample provides no replacement approval form, so the draft sentence remains unsupported.
Configuration and implementation checks
The settings below relate to the deployment and upgrade checks for this page's sample. Each is a setting or implementation check described in the official documents.
| Configuration item | Hypothetical starting point for this drill | Reason for the choice and comparison method |
|---|---|---|
AIPROXY_API_ENDPOINT | Assume the value http://aiproxy:3000 is entered | The document requires the AI Proxy root address without an appended /v1; compare against the documented example format |
AIPROXY_API_TOKEN | Assume a non-empty string is entered | The document requires a non-empty value consistent with the server side; compare against the AI Proxy server configuration |
FILE_TOKEN_KEY | Assume a random strong key is entered | The document marks this as required at runtime, at least 6 characters; compare against the required-item list |
AES256_SECRET_KEY | Assume a random strong key is entered | The document marks this as required at runtime, at least 6 characters; compare against the required-item list |
PRO_TOKEN | Assume a credential consistent with Pro is entered | The document requires this when App sets PRO_URL, at least 32 characters; compare against the Pro side configuration |
SYNC_INDEX | Assume the documented default true is kept | The document states missing indexes are created at startup; compare against the index state after upgrade |
Three common errors
- Treating "recommendation: replace screen" in the inspection report as a passed replacement approval. The input is the inspection report text; the comparison object is the replacement approval form; the observable result is whether the approval form carries an approving party, time, and conclusion. That field is absent in the sample, so the draft sentence is unsupported.
- Treating a similar work order returned by retrieval as a record inside the same authorisation scope. The input is the retrieval query and returned results; the comparison object is the outlet the work order belongs to and the authorisation boundary; the observable result is whether the returned record falls inside the current handover scope. A similarity score does not perform permission filtering.
- Treating the "manually triggered" state of an upgrade script as executed. The input is the script name and status shown on the version upgrade page; the comparison object is the script execution record; the observable result is whether a completion time and result exist. The sample provides no execution record, so the state remains unknown.
How to verify the setup
- Input the warranty card text from the hypothetical sample and the draft sentence "This work order has completed replacement approval, and a machine replacement is agreed." The comparison object is whether the materials contain the approving party, approval time, and approval conclusion for a replacement approval. The condition is met when every factual claim in the draft sentence has a corresponding field in the materials. The approval form is absent in the sample, so the assessment is unsupported.
- Input the repair work order and the inspection report from the hypothetical sample. The comparison object is whether the work order inspection conclusion and the report inspection result agree on the fault object, inspection date, and conclusion field. The condition is met when records for the same subject, same period, and same field align. Both point to display module abnormality in the sample, so the fields agree.
- Input the post-upgrade configuration list and check whether
AIPROXY_API_ENDPOINTandAIPROXY_API_TOKENare non-empty and match the documented format. The comparison object is the documented required-item list against the current entered values. The condition is met when both exist and are non-empty. If either is missing, startup validation fails as described in the document, and that behaviour needs observation in the actual environment.