What the data for this category looks like
Core data for this category originates from code repositories, test framework reports, requirement specifications, and interface definitions. Updates occur daily or after each code commit. Daily generated documents include test cases, coverage statistics entries, Mock interface descriptions, defect tickets, and more. Test cases include execution steps and expected result fields. Mock descriptions include interface paths and return examples. Defect tickets include reproduction steps and associated commit records. Document structures are highly structured, with fields tightly linked to business test scenarios.
Constraints on knowledge base retrieval and recall
The high-frequency update of data requires the retrieval and recall workflow to support rapid synchronization of the latest test documents, to avoid recalling outdated test cases or defect records. Structured document fields allow retrieval to narrow scope via associated fields, but also require retrieval logic to adapt to query scenarios that mix technical keywords and natural language. Test data construction and locating failed test cases are time-consuming business processes. Retrieval and recall must accurately match documents for target scenarios, to reduce the filtering cost of invalid results. The risk of missed local logic defects requires recall results to have high accuracy, to avoid missing key test references.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Filter low-relevance test documents, adapt to retrieval matching accuracy for structured test cases |
limit | Top 8 entries | The volume of unit and component test documents is relatively manageable; 8 entries cover reference needs for core test scenarios |
usingReRank | Enabled | Test documents include many similar case descriptions. Re-ranking improves the ranking priority of highly relevant results |
datasetSearchUsingExtensionQuery | Enabled | Questions in test scenarios often include technical abbreviations and specialized terminology. Extended queries cover more associated documents |
collectionIds | Bind dedicated collections for test cases and Mock descriptions | Isolate business data to avoid retrieving irrelevant non-test documents |
Three Common Configuration Errors
- Symptom: Knowledge base documents are retrieved, but the large language model's generated answer is concise and does not display specific clause content from the documents. Cause:
datasetSearchUsingExtensionQueryis not enabled, and full document content is not included in the large language model context. - Symptom: Knowledge base search takes a long time, and timeouts occur in environments with hardware configuration of 8 cores, 64G memory, and RTX2070. Cause: Dedicated
collectionIdsare not bound, so full knowledge base data is retrieved, or thelimitparameter is not set reasonably to recall too many irrelevant documents. - Symptom: The
usingReRankconfiguration does not take effect, and re-ranked results show no changes. Cause: The parameter is not configured in FastGPT v4.17.0 or later, or a matching re-ranking model is not configured.
How to Verify Successful Configuration
- View retrieval logs to confirm that the number of recalled results matches the value set for the
limitparameter. - Submit a query containing test case keywords to verify whether highly relevant test documents after re-ranking are ranked in higher positions.
- Check the knowledge base binding configuration to confirm that only dedicated document collections such as test cases and Mock descriptions are associated.
- Review the large language model's generated answer to confirm that it includes specific execution steps or return example content from the recalled documents.
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.