What the documents look like
Fruit-store procurement runs on four recurring documents: purchase orders, acceptance sheets, pesticide-residue test reports and loss records. The underlying data comes from wholesale-market quotations, supplier quotes, store sales and test reports, and it refreshes daily or per batch. Grade acceptance depends on on-site judgement at the wholesale market, test reports travel with each batch, and loss records are registered batch by batch. These materials move between the store's fresh-produce supervisor, the fresh-produce buyer and the wholesale-market stall, and a handover must distinguish whether each item is still being filled in, awaiting review, signed, or already executed.
This page covers one task only: operations permission and handover management during deployment and upgrade. It asks which materials, actions and requests sit in which authorization state at handover, and it treats retrieval results and form submissions as neither authorization nor execution.
Constraints at the upgrades step
Upgrading to FastGPT v4.17.2 updates both the FastGPT app image tag and the FastGPT Pro image tag to v4.17.2. Since 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, and a missing item fails startup validation. This release contains two manually triggered upgrade scripts visible under the administrator version-upgrade entry: 20261008_migrate_chunk_training converts existing old-format chunk training tasks into a new mode tied to formal data rows, and it must be run after all old-version app and Pro instances have fully stopped and old task leases have expired; 20261009_migrate_dataset_rebuild_status depends on the first script completing. Both are manual and do not block service startup. Since 4.15.0 the open-source edition no longer reads config.json, so an old volume mount must be removed and its settings moved to environment variables.
Hypothetical drill sample, labelled as such and not drawn from any real store: on 9 October 2026 a fresh-produce supervisor, A, submits a handover request asking that the acceptance sheet and pesticide-residue test report for the 8 October apple batch be handed to buyer B, and that B be authorized to continue handling that batch's loss records after the upgrade. Two source sentences accompany the request. Acceptance sheet: "Acceptance sheet no. YS-1008-03, acceptance grade grade one, accepting person A, date 8 October 2026, status pending review." Test report: "Pesticide-residue test report no. JC-1008-03, conclusion pass, issue date 8 October 2026, validity limited to this batch." Draft sentence to be judged: "A has authorized B to handle this batch's loss records, and the handover is complete."
Configuration and implementation checks
| Configuration item | Hypothetical starting value for this drill | Reason for the choice and comparison method |
|---|---|---|
AIPROXY_API_ENDPOINT | Hypothetically http://aiproxy:3000 | The document requires the root address without /v1; compare character by character against the documented example format |
AIPROXY_API_TOKEN | Hypothetically a random string matching the server side | The document requires a non-empty value consistent with the AI Proxy server; compare against the AI Proxy server configuration |
AGENT_SANDBOX_OPENSANDBOX_VOLUME_NAME_PREFIX | Hypothetically the previous VM_VOLUME_NAME_PREFIX value | The document states the old value must be carried over on upgrade; compare against the pre-upgrade recorded value |
DEFAULT_TEAM_BASIC_PERMISSIONS_ENABLED | Hypothetically false on both app and Admin | The document requires both services to stay consistent; compare whether the two sides hold the same value |
SYSTEM_MIGRATION_DELAY_SECONDS | Hypothetically 0 | The document records 0 as no delay; compare against the required execution timing of the two manual scripts |
PRO_TOKEN | Hypothetically a 32-character-or-longer string matching Admin | Required when the app sets PRO_URL; compare against the Admin-side value |
Three common errors
- Treating "status pending review" on the acceptance sheet as an effective authorization. In the sample the sheet is pending review; pending review does not show that A holds the power to authorize B, and it does not show the handover is complete. Check: read the sheet's status field and compare it against explicit values such as reviewed or effective; when only pending review appears, record the claim as unsupported.
- Treating the test report's "pass" conclusion as a supply period or an authorization scope. The report's validity is limited to this batch; a pass conclusion states nothing about a supply period and nothing about B holding authority over loss records. Check: read the report's validity-scope field and compare it against the scope asserted in the draft sentence; when the scopes differ, record the claim as unsupported.
- Treating a retrieval result or a submitted form as executed authorization. Retrieval is not authorization, and a form collecting values is not business verification, approval, a database write or external synchronization. Check: look for a separate authorization record or approval-status field; when it is absent, record the evidence gap and do not assert that authorization took effect.
How to verify the setup
- Input: the sample acceptance-sheet wording "status pending review" together with the draft sentence "A has authorized B to handle this batch's loss records." Compare the sheet's status value against effective-type values. The authorization condition is met only when the status field explicitly reads effective and the authorized party matches B. The current sample does not meet this, so the claim is recorded as unsupported.
- Input: the sample test-report wording "validity limited to this batch" together with the request to continue handling that batch's loss records after the upgrade. Compare the report's validity scope against the scope asserted in the request. The scope-consistency condition is met only when both point to the same batch and the report has not expired. The current sample does not state any authority over loss records, so this point is recorded as pending verification.
- Input: the hypothetical values of
AIPROXY_API_ENDPOINTandAIPROXY_API_TOKENin the upgrade configuration. Compare whether both are non-empty and whether the format matches the documented example. The startup-validation precondition is met only when both are non-empty and the address carries no/v1. These are hypothetical starting points, not tested, so the result is recorded as an expected condition.
Across all three checks, the same subject, the same period and the same field must be compared. Only explicitly mutually exclusive values in one field count as a contradiction. Where a field is unstated, or only related-stage evidence exists, the finding stays unsupported or pending verification. Pending review does not show that an operating permission is absent, and signing, reaching cooperation and a contract taking effect are three separate facts. The materials needed for these checks must sit inside an accessible target retrieval scope or an explicit manual input; where they are missing, the gap is recorded.