What the documents look like
Creator special sales campaigns depend on external creator fulfillment and commission settlement. The materials handled in product information and selection question answering include creator cooperation agreements, commission setting sheets, sample shipment sheets, and livestream performance reports. Brand business teams, MCN agencies, and creator managers maintain these materials separately. Updates follow a per-session or weekly rhythm.
The same product may appear across multiple materials with different commission rates, slot fees, sample status, and historical sales performance. Product specifications, commission conditions, and cooperation stage information need to sit in one knowledge base for question answering. Retrieval results present material content only. They do not constitute commission confirmation, authorization, or a settlement basis.
Constraints at the retrieval step
The knowledge base search node can associate one or more knowledge bases that use the same vector model. Knowledge bases with different vector models cannot be associated together. Check the vector model identifier first and exclude those that do not match.
Search modes include semantic retrieval, full-text retrieval, and hybrid retrieval. Hybrid retrieval merges multiple result paths through RRF. RRF is a fusion process, not an independent switch that can be turned on. The citation limit controls the amount of content cited per search in tokens. The minimum relevance score takes effect only under semantic retrieval or when result reranking is used. Question optimization handles coreference resolution and question expansion, and it adds an extra model call. Image retrieval requires image input and corresponding model capability. Reranking requires a model and must be enabled according to settings. These paths do not execute on every request.
The retrieval output is an array, and its length can be zero. The empty-result path still completes. A similarity score does not perform permission or entity isolation. Retrieval itself does not constitute authorization.
Configuration and implementation checks
The following configuration is set around the hypothetical drill sample on this page. The sample material text reads: "The sample shipment sheet dated March 2025 states that Creator A received two units of Product X on March 10, and the shipment purpose was special-session selection testing. The commission setting sheet for the same period states that the commission rate for Product X in Creator A's special session is fifteen percent, and this rate applies from March 15 to March 31." The draft sentence to be assessed is: "Creator A has confirmed the commission rate for Product X and completed the special-session cooperation." This draft sentence concerns whether cooperation was reached. The materials show only sample shipment and a commission rate range. They do not state creator confirmation or cooperation signing.
| Configuration item | Hypothetical starting point for this trial | Reason for selection and comparison method |
|---|---|---|
| Associated knowledge bases | One product information base and one cooperation agreement base | Both bases must share the same vector model identifier to be associated together; check identifiers one by one |
| Search mode | Hybrid retrieval | Commission rates contain numbers and proper nouns; compare hit differences against semantic recall |
| Citation limit | Start from complete presentation of longer materials | Test with materials of different lengths to see whether content is truncated, then adjust |
| Minimum relevance score | Label relevant and irrelevant samples first, then set a value | Takes effect only under semantic retrieval or reranking; compare results before and after filtering |
| Question optimization | Enabled for multi-turn follow-up scenarios | Compare the original question with the optimized question, and check whether the subject was altered |
| Result reranking | Enabled according to settings when a model is available | Compare ordering before and after reranking, and confirm whether candidates are sufficient |
Three common errors
- Treating a retrieval hit as cooperation confirmation. Input the draft sentence "Creator A has confirmed the commission rate for Product X and completed the special-session cooperation," and compare it with the material text. The text states only sample shipment and a commission rate range. It does not state creator confirmation or signing. The draft sentence is not supported.
- Forcing association of knowledge bases with different vector models. Input the vector model identifiers of the two bases, and compare whether they match. When the identifiers do not match, exclude the base. Do not run cross-model retrieval to compare missed recalls.
- Treating the minimum relevance score as a permission or entity filter. Input an unrelated business question, and compare the returned results. A relevance threshold does not exclude unauthorized records. Retrieval does not perform authorization judgment.
How to verify the setup
- Input the draft sentence "Creator A has confirmed the commission rate for Product X and completed the special-session cooperation." Compare it with the sample shipment date, the commission rate range, and the validity period in the material text. If the text does not state creator confirmation or cooperation signing, judge the sentence as unsupported and preserve the evidence gap.
- Input the vector model identifiers of the two knowledge bases to be associated. Compare whether the two identifiers match. Association is permitted when the identifiers match. When they do not match, record and exclude the base. Do not infer an operational result.
- Input a clearly paper-based empty-result downstream scenario, and check whether the downstream design accepts a citation array of length zero. The condition is met when the empty-result path completes. Do not claim that a hypothetical rule retrieval necessarily returns an empty array.