What the documents look like
For product questions in Brand member operations, 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 membership-product rule test says, “The member gift set contains a trial pack; the full-size item is selected separately.” Inspect candidates for trial and full-size terms before comparing modes with the complete contents question. The membership label supplies a hypothetical context, without real personal data. A product-keyword match cannot convert the optional full-size item into established gift-set contents.
Configuration and implementation checks
| Documented item or implementation check | Material and condition | Evidence action |
|---|---|---|
| Source scope (implementation check) | Separate gift-set rules and display wording | A displayed item does not replace the contents list. |
| Search modes: semantic, full-text, hybrid | Trial-pack wording and the full contents question | Compare actual support under each mode. |
| Citation token limit | Full-size separately selected; display not configuration | Retain exclusion and optional-item conditions. |
| Minimum relevance | Inspect filtering under semantic search or reranking | Related results do not supply trial-pack specifications. |
| Question optimization | Follow-up about the set specification | Do not transfer full-size attributes to the trial pack. |
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 outreach plan says, “Full-size display illustration; this does not describe the gift-set configuration.” It is ordinary text without an actual image. Using a displayed item as the set’s contents misuses that source. Retain the rule, plan wording and actual citation together. Inspect token trimming for a lost exclusion or separate-selection condition, rather than an isolated full-size noun. Joint retrieval of an outreach plan and a product rule does not supply a new gift-set list. A third error extends a full-size display in one contact plan to every member activity. Check the plan activity and period against this bundle rule. A single display neither covers other campaigns nor supplies the current contents list.
How to verify the setup
No trial-pack specification was supplied. The rule and display wording cannot provide capacity or material. For a size follow-up, inspect whether optimization retains the trial pack instead of importing full-size specifications. After obtaining the corresponding product explanation, repeat the question and trace the new citation, recording source and retrieval-setting changes separately. Check minimum relevance under semantic or reranking conditions. A score does not establish complete gift attributes or any real member’s purchase or authorization state.
| Material context | Scope |
|---|---|
| Document types | Member Tagging Sheet, Engagement Plan, Coupon Rules and Privacy Authorization Records |
| Candidate sources | CRM System, WeCom, POS Consumption Records and Mini Program Behavior Data |
| Technical baseline | FastGPT v4.17.2 |