What the documents look like
Member private-domain operations centers on customer data and repeat-purchase outreach rather than store scheduling. The working materials are member tag tables, outreach plans, coupon rules and privacy authorization records. These materials are updated daily or weekly and are used for operating reviews and exception post-mortems, so a retrieval result needs to trace back to a specific field and a specific period.
The material types are mixed, which creates a direct constraint. Coupon codes, tag names and identifiers are suited to keyword matching, while natural-language descriptions are suited to semantic recall. A single query can involve both types, and one search mode may not cover every phrasing. A retrieval step that returns references does not by itself confirm that a tag was active, that a coupon was issued, or that an outreach action was authorized.
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 using different vector models cannot be associated together. When checking, first read the vector model identifier, exclude those that do not match, and record the actual observation. Do not run a cross-model retrieval to compare missed recalls.
Retrieval operates on the whole knowledge base. A collection is used for classification management and does not affect search results. After vector recall, the original data is retrieved from MongoDB by vector ID, and when records map to the same original data they are merged with the highest score retained. Hybrid retrieval merges multiple result paths through RRF; RRF is a fusion process, not a switch. The minimum relevance value takes effect only in semantic retrieval or when result reranking is used, and it filters low-relevance results. It does not perform permission or entity isolation. The reference limit is measured in tokens. The output reference is an array whose length can be 0, and the downstream path still runs when nothing is found.
A hypothetical paper exercise sample: a brand CRM team keeps a fragment of a member tag table whose source text reads, "Tag 'High repeat purchase - 30 days' definition: completed 2 or more paid orders in the last 30 days; effective period 2025-06-01 to 2025-06-30; applicable channel is the mini program." A draft sentence to be judged reads, "This tag remains continuously effective in July 2025." The check compares the effective period in the tag definition with the month stated in the draft sentence. July does not fall inside the stated period, so the draft sentence is not supported.
Configuration and implementation checks
| Configuration item | Hypothetical starting point for this exercise | Reason and comparison method |
|---|---|---|
| Associated knowledge bases | Associate only bases using the same vector model | Read the vector model identifier first, exclude inconsistent ones, then associate |
| Search mode | Run semantic retrieval and full-text retrieval once each | Compare the two result paths on the same test question, then decide whether to use hybrid |
| Reference limit | Set a token value that covers a complete coupon rule | Compare actual material fragments of different lengths to see whether content is truncated |
| Minimum relevance | Label relevant and irrelevant samples first, then set a threshold | Takes effect only in semantic retrieval or result reranking; compare samples before and after filtering |
| Question optimization | Enable in multi-turn follow-up and short-question scenarios | Compare the original question with the optimized question and check whether the subject was altered |
| Result reranking | Enable when candidates are numerous and the question is clear | Requires a rerank model enabled by configuration; compare ordering before and after reranking |
The values in the table are hypothetical starting points for this exercise and do not represent official defaults. Image retrieval requires corresponding image input and model capability. The optional paths shown in the documented flow diagram do not mean every request executes all of them.
Three common errors
- Treating minimum relevance as a permission or entity filter. The input is the candidate results for the same question, the comparison object is the relevance score against the authorization scope, and the observable result is that low-scoring results are filtered while out-of-scope records may still appear. Relevance does not perform isolation, and retrieval does not constitute authorization.
- Forcing knowledge bases with different vector models into a shared association. The input is the vector model identifier of each base to be associated, the comparison object is whether the identifiers match, and the observable result is that mismatched bases are excluded and recorded. Do not infer missed recalls or abnormal behavior from this.
- Treating a pending review or a quoted price as already effective. The input is the status field in the handover material, the comparison object is the value of the same field in the same period, and the observable result is that a pending-review status is judged as requiring verification. Signing, reaching cooperation and contract effectiveness are three distinct facts.
How to verify the setup
- Input the member tag table source text above and the draft sentence "This tag remains continuously effective in July 2025." Compare the effective period in the tag definition with the month stated in the draft sentence. The support condition is met only when the month stated in the draft falls inside the stated period; in this comparison the result is not supported, and the evidence gap remains.
- Input one precise query containing a coupon code and one natural-language query, and route them through full-text retrieval and semantic retrieval respectively. Compare the reference content arrays returned by the two paths. Observe whether either path returns an empty array; an empty result can be checked against a paper input of [] for downstream design, without claiming that a fictional rule retrieval must return [].
- Input a handover request: "Please synchronize the coupon package configuration from the June outreach plan to July." Compare the action stated in the request with the authorization status recorded in the materials. The execution condition is met only when an explicit authorization record exists and the action falls within the authorized scope; that authorization record is not provided here, so the judgment is pending verification and the synchronization is not executed.
All of the above is a paper check. No deployment, test or business action took place.