What the documents look like
For after-sales questions in Certificates of origin, this page examines FastGPT multi-turn prompts. Preserve requested corrections, application stages and production evidence across turns without equating dispatch or acceptance with the outcome. The examples are hypothetical.
Constraints at the multi-turn prompts step
A proposed service case contains a customer email asking to correct a consignee name and a dispatch record showing that the original was sent. Asked whether it has been handled, the answer must separate the request, dispatch event and correction result. Dispatch alone does not establish amended content. If sent becomes corrected, compare each statement with actual history. Without a correction result, identify the unverifiable item instead of prompting a completed status into existence.
Configuration and implementation checks
| Service-conversation check | Certificate-correction material | Acceptance basis |
|---|---|---|
| Request scope | Email asks to amend the consignee name | No expansion into other amendments |
| Event separation | Dispatch, application and issuance | Original statements adopted at each turn |
| Reference handoff | Identify the new certificate and version | Compare original and optimized questions |
| Result gap | Unobtained versus omitted result | Collect the source or inspect input |
| Completion boundary | Inspect correction result and clearance requirements | Conversation does not confirm eligibility |
Documented conditions
| Documented item / implementation check | Condition |
|---|---|
Prompt and task | Define sources, the task and evidence-gap handling. |
Conversation history | Inspect the history actually submitted. |
Question optimization | Review reference resolution and question expansion. |
Model and length | Inspect context and reply length separately. |
Reply format | json_schema needs its format and parsing validation. |
Three common errors
- If dispatch of the original becomes a corrected consignee name, compare email, dispatch record and answer with the correction result. Mailing establishes no amended content.
- If the new document becomes an old certificate or another customer, retain questions and certificate identity. Inspect history and application version rather than substituting more context for correspondence.
- If submitted correction becomes issuance or clearance suitability, inspect actual results and customer requirements. Distinguish an obtained result omitted from input from an unobtained result; submission is no formal outcome.
How to verify the setup
Append the email, dispatch record and application progressively, checking preservation of their distinct stages. Recheck the same question after adding an actual result. Without production evidence or applicable origin rules, a service summary can list unresolved actions but cannot confirm origin eligibility. Record the original wording behind each conclusion and remaining owner confirmation. Verify actual issuance and customer requirements separately; the end of a conversation does not establish completion of the correction process.
| Material context | Scope |
|---|---|
| Document types | Certificate of Origin Application, Commercial Invoice, Packing List and Production Cost Data |
| Candidate sources | China International Trade Single Window, CCPIT System, Enterprise Production Records and Customer Customs Clearance Requirements |
| Technical baseline | FastGPT v4.17.2 |