What the documents look like
For product questions in Refund-only disputes, this page examines FastGPT retrieval. Compare the material received with the sources used in the answer. The examples are hypothetical.
Constraints at the retrieval step
A new refund-only application test says, “Missing accessory; specific model still required.” This is the applicant’s issue wording, without an identified accessory. Search the literal issue and a question about what product the available material establishes, preserving the missing-model qualification. An answer borrowing a product name from another application confuses the retrieval target. Compare the original application, cited record and question rather than treating related topics as the same item.
Configuration and implementation checks
| Documented item or implementation check | Material and condition | Evidence action |
|---|---|---|
| Source scope (implementation check) | Application, shipment wording and replacement request | Inspect their different source roles. |
| Search modes: semantic, full-text, hybrid | Missing-accessory and product-identity questions | Do not borrow an object from another application. |
| Citation token limit | Missing-model and request contexts | Trimming must not remove the unresolved state. |
| Minimum relevance | Check relevance under semantic search or reranking | Joint retrieval does not establish transaction linkage. |
| Question optimization | Follow-up about the replacement accessory | Do not convert an unknown object into a model. |
Documented conditions
| Documented item / implementation check | Condition |
|---|---|
Associated knowledge bases | Knowledge bases associated with one search node must use the same embedding model. |
Search mode | Semantic, full-text and hybrid modes are available. |
Citation limit | Citation limits are measured in tokens. |
Minimum relevance | Range 0–1; active for semantic search or reranking. |
Question optimization | Resolves references and expands questions. |
Three common errors
A proposed shipment record lists a main item and accessory bag, while an appeal requests a replacement accessory. Turning the request into an established shipment event misuses the source role. Trace the three pieces of text to their actual purposes. Token limits should preserve request and application-reason contexts, beyond isolated accessory nouns. Correspondence between records still requires transaction linkage; appearing in the same candidate set does not establish that relationship.
How to verify the setup
The model, contents list and transaction-linking evidence remain absent. They cannot be derived from main item or accessory bag. For a follow-up about which replacement to use, inspect whether optimization turns an unresolved object into a specified product, recording the source of additions. After obtaining corresponding material, repeat the question and trace new citations, separating candidate changes from settings changes. Check minimum relevance only under its applicable conditions; several hits do not establish product identity or resolve the dispute.
| Material context | Scope |
|---|---|
| Document types | Refund-Only Application, Proof of Shipment, Proof of Receipt and Appeal Materials |
| Candidate sources | Platform After-Sales Backend, Carrier Tracking, Customer Service Chat and Warehouse Outbound Records |
| Technical baseline | FastGPT v4.17.2 |