What data in this category looks like
Data sources for this category include databases, data dictionaries, data masking tools, and access approval systems. Update frequency is per batch or per project. Daily produced documents include masking plans, field mapping tables, masking verification reports, and approval records. Most document fields include sensitive field identifiers, masking rule codes, verification results, approval statuses, and similar fields. There are no universal standard units, and corresponding field definitions must be matched based on specific project scenarios.
What constraints these characteristics impose on knowledge base retrieval and recall
The non-standardized fields of this category of data and project-bound update rhythms require retrieval to support filtering recall results by project dimension, to avoid cross-project sensitive data leaks. The presence of sensitive fields requires that sensitive information identification be completed in advance during retrieval, to avoid exposing raw data directly during the recall phase. Differences in data formats across multiple sources require retrieval to be compatible with parsing results of different document types, and to limit the number of single-recall results to control compliance risks. At the same time, the association between masking verification reports and approval records requires retrieval to support cross-document associated recall, ensuring that technical support and pre-sales Q&A can obtain complete compliance chain information.
The following configurations apply to FastGPT v4.17.0.
How to set configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches semantic similarity for professional documents such as masking plans and field mapping tables, filtering out low-correlation non-masking content |
limit | Top 8 entries | Controls the number of single-recall results, reduces compliance risks of batch leakage of sensitive information, and adapts to the information density requirements of pre-sales Q&A |
searchMode | vector | Prioritizes vector retrieval to match semantic associations for structured masking rule documents and field mapping tables, improving retrieval accuracy |
usingReRank | Enabled | Re-ranks recall results, prioritizes returning highly relevant content related to masking compliance, and improves the efficiency of technical support and pre-sales Q&A |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands user technical support questions by adding professional terms related to data masking, improving the coverage and accuracy of recall |
collectionIds | Specify collections by project | Filters masking-related documents for the current project, preventing accidental recall of sensitive data across projects |
Three common misconfigurations
- The symptom is that unmasked raw sensitive field content appears in retrieval results. The cause is that the pre-sensitive field identification configuration is not enabled, and raw document data is recalled directly.
- The symptom is that masking configuration content corresponding to Feishu multi-dimensional documents cannot be recalled. The cause is that parsing indexes supporting multi-dimensional document formats are not configured, causing relevant documents to not be included in the recall scope.
- The symptom is that the number of results returned by a single retrieval exceeds expectations. The cause is that the
limitparameter is not configured correctly, orcollectionIdsis not specified, leading to excessive document recall across projects.
How to confirm the configuration is correctly applied
- Upload a test masking plan document, retrieve for questions containing "sensitive field identification", and check whether the returned results include this document and rank it highly.
- Check whether the
collectionIdsparameter is bound to the masking-related dataset for the current project, try to retrieve documents from other projects, and confirm that irrelevant content cannot be recalled. - After enabling
usingReRank, compare the result ranking before and after enabling to confirm that the priority of compliance-related documents has been improved. - Enter a question containing extended terms, such as "data masking field mapping", check whether documents related to field mapping tables are recalled, and confirm that
datasetSearchUsingExtensionQueryis functional.
The values provided are common starting points and should be measured against the reader's own samples.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-10-19.