What the documents look like
This page covers four record families used in platform store ratings and experience-score maintenance: negative-review handling records, experience-score detail sheets, service diagnostic reports, and rectification ledgers. Public industry material describes the domain as consumer feedback after post-sale fulfilment, drawn from sources such as Douyin store experience scores, Tmall service indicators, JD after-sales reviews, and customer-service systems. Errors in this area can be associated with lower search weight, restricted campaign registration, and reduced store conversion. Update frequency appears as daily or weekly. This page does not derive any batch-sync, full-sync, real-time, or transaction-triggered mechanism from that frequency.
The operational task here is permission and handover management at the deployment and upgrade step. The relevant distinction is between behaviour a node already has, fields that are actually configurable, manual reconciliation actions, and business actions that require a separate downstream implementation. Retrieval and forms collect or return values; neither constitutes authorization. A pending review state does not demonstrate the absence of an operating permission.
Constraints at the upgrades step
Upgrades change how back-office pages render and how interfaces behave. In v4.17.2, the agent workspace, skills, MCP, system tools, template marketplace, evaluation pages, account centre modules, the admin back office, and the pricing page were migrated to client-side rendering with shallow routing. The release adds a ClientRouteReadyGate route guard component and a useRequiredQueryParam hook, used to intercept state while route parameters are not yet ready and to sequence dialog initialisation. For ratings handover, this means a handover checklist has to separate what the node already does from what an operator must reconcile by hand.
The release also lists two manually triggered upgrade scripts, viewable and executable under the admin version-upgrade area: 20261008_migrate_chunk_training and 20261009_migrate_dataset_rebuild_status. Both are described as not blocking service start. The first should be run only after confirming that all old-version FastGPT App and Pro instances in the cluster have fully stopped and old task leases have expired. The second depends on the first having completed. Executing a script is not a zero-interruption commitment.
Hypothetical paper exercise sample. The source text to be checked reads: "Store operator A submitted a negative-review handling record on 9 October 2026, record number A-001, concerning one one-star review on order 20261001, status pending customer-service supervisor review, authorization scope limited to viewing and supplementing evidence, excluding review deletion." The handover request to be assessed reads: "Operator A hands over the deletion permission for this record to B as well." The source text states viewing and supplementing evidence only. It does not state deletion authorization. The request is therefore not supported, and the evidence gap must be recorded.
Configuration and implementation checks
The table below stays within this page's documents and the handover task. Values marked as hypothetical starting points are paper assumptions for the exercise, chosen for a stated reason and compared against a stated reference. They are not official defaults.
| Item | Hypothetical starting point for this exercise | Reason and comparison method |
|---|---|---|
AIPROXY_API_ENDPOINT | Assume http://aiproxy:3000 | Official text requires the root address without appending /v1; compare the format against the documented example |
AIPROXY_API_TOKEN | Assume a non-empty admin token | Official text requires non-empty and consistent with the server side; compare character by character against the AI Proxy configuration |
FE_DOMAIN | Assume http://localhost:3000 | Official text uses this to complete resource paths; compare against the address actually used by the browser |
FILE_TOKEN_KEY | Assume a random strong key | Official text marks it runtime-required with a minimum length of 6; compare length and randomness only, not the value |
AES256_SECRET_KEY | Assume a random strong key | Official text marks it runtime-required with a minimum length of 6; same comparison approach |
INVOKE_TOKEN_SECRET | Assume a random strong key | Official text marks it runtime-required with a minimum length of 32; same comparison approach |
The two upgrade scripts remain manually triggered tasks. Their execution state is a node behaviour; it does not by itself establish that any ratings record was handed over, approved, or deleted.
Three common errors
- Treating a retrieval hit as an authorization basis. Input: a retrieval hit on record A-001 from the sample. Compare the hit content against the authorization scope stated in the source text. If the hit contains only viewing and supplementing evidence, deletion permission is not supported.
- Reading a pending-review status as the absence of an operating permission. Input: the source phrase "status pending customer-service supervisor review." Compare the status field against the permission field. These are different fields and cannot negate each other.
- Treating a form submission as completed execution. Input: a deletion request form submitted by B. Compare the form values against the authorization scope in the source text. A form collects values; deletion execution requires a downstream implementation, and where none is stated the outcome is unknown.
How to verify the setup
- Input the three source sentences from the sample and the draft sentence "Operator A hands over the deletion permission for this record to B as well." Compare the authorization asserted by the draft against the authorization scope stated in the source. The condition is satisfied only if the source contains a deletion-authorization statement; otherwise record it as not supported.
- Input the record number, object, status, and effective scope of record A-001 from the sample. Compare the handover list for the same subject, the same period, and the same field. Record a contradiction only when a field has explicitly mutually exclusive values; where a value is not stated, record it as pending verification.
- Input the post-upgrade page route and the
useRequiredQueryParamrelated parameters. Compare whether an interception state appears while parameters are not ready. Interpret the interaction wait and the record state as the document describes; do not infer automatic submission counts or snapshot guarantees.