What data for this category looks like
The data for this category primarily comes from requirement specification documents, test environment operation logs, customer business sample data, and defect system ledgers. Daily generated documents include UAT test cases, acceptance issue tickets, acceptance reports, and signed confirmation forms. Each document typically includes fields such as business scenario description, test steps, problem reproduction steps, and acceptance conclusions. There are no uniform mandatory unit requirements. Data updates align with project phases or version iterations. After each acceptance phase or system version update, relevant documents and data are synchronized to ensure retrieved content matches the current project status.
What constraints do these characteristics impose on knowledge base retrieval and recall
Dispersed data sources require cross-multiple data source association matching during retrieval. Configure rules to support joint recall across multiple datasets. Unstructured document formats require the retrieval model to adapt to semantic matching for non-standard text, to avoid recall failures caused by format differences. The update rhythm tied to phases and versions requires regular synchronization of dataset versions in the knowledge base, to prevent retrieval of expired acceptance test cases or issue tickets. For business scenario-focused retrieval needs, prioritize recalling content matching the current UAT phase, to reduce interference from irrelevant results on acceptance judgments.
How to set the configurations
The following configurations apply to FastGPT v4.17.0.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Adapts to semantic matching accuracy for UAT test cases and acceptance issue tickets, filters low-relevance non-business results |
limit | Top 8 entries | Matches the volume of single-scenario test documents for the UAT phase, avoids overly long context affecting acceptance conclusion judgments |
searchMode | semantic | Prioritizes recalling documents related to the current UAT business scenario based on semantics, adapts to non-standard acceptance issue descriptions |
usingReRank | Enabled | Performs secondary ranking on initial recall results, improves the ranking priority of highly relevant acceptance issue tickets and reproduction steps |
datasetSearchUsingExtensionQuery | Enabled | Automatically generates extended query terms based on user-input acceptance issues, covering implicit scenario descriptions in customer business samples |
collectionIds | Bound to current project’s UAT test case and acceptance issue ticket datasets | Restricts retrieval scope to only acceptance-related data for the current project, avoids recalling defect records from unrelated systems |
Three common mistakes
- Symptom: After migrating the acceptance knowledge base to a newly deployed environment, retrieval returns empty results. Cause: The bound
collectionIdsdataset configuration was not migrated synchronously, or the index for the corresponding dataset was not rebuilt, causing the new environment to fail to recognize the original dataset. - Symptom: Executing knowledge base retrieval returns a
text index required for $text queryerror. Cause: After upgrading the version, the full-text index for the corresponding dataset was not rebuilt, making semantic retrieval unsupported. - Symptom: When passing knowledge base variables via an API in a workflow, the current UAT test case cannot be matched. Cause:
datasetSearchUsingExtensionQuerywas not enabled, and no extended query terms adapted to the business scenario were generated, leading to overly narrow retrieval scope.
How to confirm the configuration is correct
- Manually input a current project acceptance issue with a length of 500 characters, wait 10 seconds, then verify whether the corresponding UAT test case or acceptance issue ticket appears in the retrieval results.
- Check whether
collectionIdsin the configuration panel is only bound to acceptance-related datasets for the current project, with no IDs of unrelated datasets. - View the retrieval log to confirm whether the relevant logic for
usingReRankanddatasetSearchUsingExtensionQuerywas triggered. - Simulate the retrieval scenario after a version upgrade, and verify that the
text index required for $text queryerror does not occur.
The values provided are common starting points and should be measured against the corresponding project’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.