What the documents look like
For product questions in On-demand retail fulfillment, 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 instant-retail order test says, “Refill cleaning liquid; spray head selected separately.” A contents question needs the refill description and separate-selection condition together. Inspect candidates for literal product wording, then compare the three retrieval modes for spray-head inclusion. A head-term hit does not make it part of the order, and refill does not establish a complete container or accessory set. The input is proposed order wording only.
Configuration and implementation checks
| Documented item or implementation check | Material and condition | Evidence action |
|---|---|---|
| Source scope (implementation check) | Separate purchased contents and stock gaps | Still required does not mean included in the order. |
| Search modes: semantic, full-text, hybrid | Refill and spray-head inclusion questions | Inspect qualifications retrieved by each mode. |
| Citation token limit | Separately-selected and still-required conditions | Retain the actual source state in citations. |
| Minimum relevance | Inspect filtering under semantic search or reranking | Related bottled-product text supplies no capacity. |
| Question optimization | Follow-up about this bottle’s capacity | Retain order and product without adding a value. |
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 picking list says bottled product, while an out-of-stock note says the spray head is still required. Converting a stock gap into an included order item mixes source roles. Compare order, picking entry and stock note to identify the actual passage returned. Token trimming may also remove still-required or separately-selected conditions. Preserve complete wording beside the result. Inventory state and purchased contents are different targets, without a joint conclusion inferred from shared retrieval. A third error turns a pending spray-head replenishment note into definite current inability to fulfil. Check subsequent applicable replenishment or alternative arrangements for this item. An older inventory note and present fulfilment failure cannot be treated as equivalent.
How to verify the setup
No capacity or actual accessory detail was supplied. Bottled and refill cannot provide a numerical quantity. Inspect whether optimization of a size follow-up keeps the same order without switching product or adding a volume. After obtaining the corresponding explanation, repeat the question and trace new citations, recording source updates separately from retrieval settings. Check minimum relevance under semantic or reranking conditions. An unsupported capacity or accessory needs data-object inspection; a relevance score is not specification evidence.
| Material context | Scope |
|---|---|
| Document types | Instant Order, Picking List, Out-of-Stock Report and Delivery Exception Order |
| Candidate sources | Douyin Instant Retail, Meituan or JD Daojia Backend, Store Inventory and Rider Tracking |
| Technical baseline | FastGPT v4.17.2 |