What the documents look like
Parcel labels and handoff records sit at the point where a parcel leaves the warehouse, so the operative fact is which carrier scan has been recorded. The working documents for this category are electronic waybills, handoff lists, pickup scan records and exception item registers. Sources include the carrier system, WMS shipment records, the GB/T express service standard and carrier billing. Where a draft describes dispatch timing, pickup progress or exception handling, the wording has to correspond to a specific batch, event and status, and handoff and scan records are checked separately.
This affects the drafting task in a direct way. A marketing or product sentence about pickup completion is only as strong as the scan record behind it. A handoff list confirms that parcels were transferred; it does not confirm that the carrier scanned them. An exception register that shows one item still pending does not support a claim that no exceptions exist. The drafting step therefore depends on reading the source documents as separate evidence streams rather than as one combined status.
Constraints at the model connections step
Copy drafting in FastGPT runs through model connections. FastGPT connects to model providers through AI Proxy. Both the App and an Admin service that reuses App/Service configuration must set AIPROXY_API_ENDPOINT and a non-empty AIPROXY_API_TOKEN; missing either one causes the startup check to fail. AIPROXY_API_ENDPOINT takes the service root address and does not carry /v1. The proxy address in a model channel configuration likewise takes a BaseUrl, and the question of whether to add /v1 has to be judged rather than assumed. The system needs at least one language model and one index model to work normally. Model capabilities are checked against what the application actually sends, and multimodal input and tool calling are verified separately. Application parameters are set in the application or node configuration.
A hypothetical paper exercise sample follows. A warehouse dispatch team on 10 March 2026 prepares copy for one batch of waybills. The source material for checking consists of three sentences. The first sentence: "This batch of parcels completed handoff at 14:00 on 10 March 2026, handoff list number JL-20260310-07." The second sentence: "As of 18:00 on 10 March 2026, the pickup scan records for this batch show 42 items scanned." The third sentence: "The exception item register records 1 item held at the branch because the outer packaging was damaged, with status pending." The draft sentence to be judged is: "This batch of parcels has completed all pickup scans, with no exception items." The check concerns the support relationship between the three source sentences and this one draft sentence.
Configuration and implementation checks
The table lists only settings or implementation checks that the official documents describe and that bear on this page's task. The "hypothetical exercise starting value" column is a comparison value chosen for the paper check; it is not an official default.
| Configuration item | Official description | Hypothetical exercise starting value and comparison method |
|---|---|---|
AIPROXY_API_ENDPOINT | Required; service root address, no /v1 appended | Paper entry http://aiproxy:3000, compared character by character against the documented example format, confirming no /v1 suffix |
AIPROXY_API_TOKEN | Required; non-empty; must match the server configuration | Paper entry of a non-empty string, checked separately for non-emptiness and for consistency with the server configuration; the actual call remains untested |
| Model channel proxy address | Takes a BaseUrl; whether to add /v1 must be judged | Paper entry of a vendor BaseUrl, compared against the documented instruction not to enter a full model request address, confirming no concrete path is included |
| Model mapping | Maps a requested model to a specific model | Paper entry {"gpt-4o-test": "gpt-4o"}, compared against the documented example structure, confirming the key-to-value correspondence |
| Language model and index model | The system needs at least one of each | Paper list of enabled models, compared against the requirement of at least one language model and one index model, confirming both types are present |
| Model status probe | Applies only to enabled models; interval is an integer from 5 to 60 minutes | Paper entry of 15 minutes, compared against the documented range, confirming it falls between 5 and 60 and is an integer |
Three common errors
- Entering
AIPROXY_API_ENDPOINTwith/v1appended. The input is the address string written on paper; the comparison object is the documented instruction not to append/v1; the observable result is whether/v1appears at the end of the address. Its appearance means the entry does not conform. - Substituting an application API key for
AIPROXY_API_TOKEN. The input is the written description of where the token comes from; the comparison object is the documented requirement for an AI Proxy administrator token that must match the server configuration and must not be empty; the observable result is whether the token is declared to be an administrator token. A declaration that it is an application API key means the entry does not conform. - Entering a full request address as the model channel proxy address. The input is the proxy address written on paper; the comparison object is the documented instruction not to enter a full model request address and to enter a
BaseUrl; the observable result is whether the address contains a concrete path such as/chat/completions. Its presence means the entry does not conform.
How to verify the setup
- Against the first hypothetical source sentence, "This batch of parcels completed handoff at 14:00 on 10 March 2026, handoff list number JL-20260310-07," the input is this sentence together with the draft sentence "This batch of parcels has completed all pickup scans, with no exception items." The comparison object is whether "completed handoff" and "completed pickup scans" are the same field in the same status. The observable result is that the two point to different stages, and the draft sentence's claim that all pickup scans are complete is not supported by the first sentence.
- Against the second hypothetical source sentence, "As of 18:00 on 10 March 2026, the pickup scan records for this batch show 42 items scanned," the input is this sentence together with the draft sentence. The comparison object is whether "42 items" and "all" agree. The observable result is that the source gives a specific item count and does not give the total number of items in the batch, so the claim "all" cannot be judged to hold and that wording remains to be verified.
- Against the third hypothetical source sentence, "The exception item register records 1 item held at the branch because the outer packaging was damaged, with status pending," the input is this sentence together with the draft sentence "with no exception items." The comparison object is whether "1 exception item" and "no exception items" are mutually exclusive. The observable result is that the two take mutually exclusive values on the same field, and the draft sentence contradicts the source. All three checks describe expected outcomes of a paper comparison only; no test was actually run.