What the documents look like
This category covers foreign-exchange receipt and settlement together with trade directory records. The working materials are receipt vouchers, trade contracts, invoices and foreign-exchange business registration records. The focus is on the authenticity of FX funds and the underlying trade background; a full letter-of-credit document set is not examined here. Materials arrive on two rhythms: per transaction and monthly. A single receipt corresponds to a voucher, a contract and an invoice; the monthly dimension covers the collection of FX business registration records.
For deployment and upgrade work, these materials exist in the system as files and data rows, and permission and handover states attach to specific records. A retrieval hit or a completed form does not by itself constitute authorization. Typical roles are trade finance staff, bank relationship managers and FX specialists. Errors at this stage can lead to bank rejection, abnormal FX classification or funds that cannot be credited.
Constraints at the upgrades step
Before an upgrade, check the current version against the target version and the corresponding upgrade notes. Do not infer current behavior from an old configuration name. The V4.17.2 upgrade guide lists two manually triggered scripts: the historical chunk training task migration (20261008_migrate_chunk_training) and the knowledge base rebuild status migration (20261009_migrate_dataset_rebuild_status). The second depends on the first being complete. Both are manual triggers that do not block service startup. The first should be run after confirming that all old-version FastGPT App and Pro instances in the cluster have fully stopped and old task leases have expired.
App and Admin must set AIPROXY_API_ENDPOINT and a non-empty AIPROXY_API_TOKEN. Missing either causes startup validation to fail. Since 4.15.0 the open-source edition no longer reads config.json; former fields are now environment variables, and the old volume mount for that file should be removed. FILE_TOKEN_KEY, AES256_SECRET_KEY and INVOKE_TOKEN_SECRET are required at runtime and should be random strong keys rather than example values. DEFAULT_TEAM_BASIC_PERMISSIONS_ENABLED affects newly created teams only and does not change existing teams; FastGPT and Pro Admin should keep the same value.
Hypothetical paper exercise sample. A trade finance specialist submits a handover request. The source materials read: "Receipt voucher: payer Company A, amount USD 12,000, value date 2026-03-05; Trade contract: seller Company B, buyer Company A, subject a batch of hardware fittings, signed 2026-02-20; Invoice: issuer Company B, amount USD 12,000, issued 2026-03-01." The proposed handover sentence is: "The three materials above have been checked and are consistent; the receiving entity, the contract buyer and the invoice payer are the same entity, so settlement handover can be completed." The question is whether that sentence is supported by the materials.
Configuration and implementation checks
The table below is a hypothetical trial starting point for this page's sample and primary task. It is not an official default.
| Configuration item | Hypothetical starting value | Reason and comparison method |
|---|---|---|
AIPROXY_API_ENDPOINT | Assume http://aiproxy:3000 | Documentation requires the root address without /v1; compare against the documented example |
AIPROXY_API_TOKEN | Assume a non-empty string | Documentation requires non-empty and matching the server; compare against the deployer's server value |
FILE_TOKEN_KEY | Assume a random strong key | Documentation requires at least 6 characters; compare against the length floor |
AES256_SECRET_KEY | Assume a random strong key | Documentation requires at least 6 characters; compare against the length floor |
INVOKE_TOKEN_SECRET | Assume a random strong key | Documentation requires at least 32 characters; compare against the length floor |
DEFAULT_TEAM_BASIC_PERMISSIONS_ENABLED | Assume false | Documentation states it does not affect existing teams; compare the value on both FastGPT and Pro Admin sides |
Three common errors
- Treating a retrieval hit as authorization. Input: the handover request above. Compare the entity fields in the source materials against the assertion in the proposed sentence. Finding a receipt voucher or contract shows the record exists; it does not show the submitter holds settlement operation permission.
- Treating a completed form as business execution. Input: the amount and entities entered in the handover request. Compare them against the corresponding fields in the voucher, contract and invoice. All three present and consistent supports the fact that materials were checked; whether settlement is executed requires a separate downstream implementation.
- Treating a pending-review state as missing permission. Input: the current state of the handover request. Compare against the authorization scope stated in the source materials. When the materials do not state an approval conclusion, record it as pending verification rather than writing it as a denial or a conflict.
How to verify the setup
- Input the source materials above and the proposed handover sentence. Compare the voucher payer, contract buyer and invoice payer fields to see whether they name the same entity. If all three point to Company A and the amounts match, the entity and amount assertions in the sentence are supported; if any field is not stated, record the evidence gap.
- Input the pre-upgrade environment variable list. Compare whether
AIPROXY_API_ENDPOINTandAIPROXY_API_TOKENare both set and non-empty. When both are present, the condition for startup validation passing is met; when either is missing, record it per documentation as a validation failure without inferring specific error text. - Input the upgrade script execution order. Compare the sequence between the historical chunk training task migration and the knowledge base rebuild status migration. The condition that the latter runs after the former is complete is met; if only the latter is run, record the dependency as unmet without claiming a specific data outcome.