What Data for This Category Looks Like
Unit and component testing related data primarily comes from code repositories, test framework reports, requirement specifications, and interface definitions. Data is updated daily or after each code commit. Daily documentation includes structured test cases, coverage reports, Mock interface descriptions, defect tracking tickets, and more. Data fields include case ID, covered code scope, Mock address, defect status, reproduction steps, and others. The length of individual data entries ranges from hundreds to thousands of characters. Some report-style documents include section titles and jump links.
Constraints Imposed on Knowledge Base Retrieval and Recall
The high-frequency updates of test data require retrieval processes to support incremental synchronization and rapid refresh, to avoid recalling outdated cases or reports. Documents contain section structures and associated information, so the integrity of grouped data must be preserved to prevent splitting associated test steps and Mock descriptions. Test cases have many boundary conditions and high field precision requirements. Retrieval processes must accurately match keywords, while avoiding irrelevant low-relevance results that interfere with test engineers’ positioning work.
Configuration Recommendations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Test cases have high requirements for logical rigor. This interval filters low-relevance non-target cases and avoids interfering with boundary condition positioning |
limit | Top 8 entries | Component testing has a moderate number of associated test cases. This value controls context length while covering major test scenarios |
searchMode | Precise Matching | Test document fields are clearly defined. Precise matching quickly retrieves corresponding test cases, Mock descriptions, or defect tickets |
usingReRank | Enabled | Test documents contain numerous boundary conditions and parameter details. Reranking improves the sorting priority of highly relevant results |
datasetSearchUsingExtensionQuery | Enabled | Test documents include interface definitions and Mock descriptions. Extended queries cover associated parameters and dependencies |
Three Common Mistakes
- Phenomenon: When creating a plain text knowledge base collection, incorrect configuration of the
trainingTypeparameter causes test cases to be split incorrectly. Cause: The optional values oftrainingTypefor this category and the matching rules for test document structures are not clarified, and parameters unsuitable for structured documents are misused. - Phenomenon: When uploading test documents with section titles, associated test steps and Mock descriptions in the same group are split. Cause: Correct chunking rules to preserve section associations are not configured, resulting in failure to match complete test scenarios during retrieval.
- Phenomenon: After enabling the citation display function, no citation markers for the corresponding documents appear at the end of responses. Cause: The citation display switch is not enabled in knowledge base settings, or retrieval results are not correctly associated with the corresponding dataset.
How to Confirm Configuration is Complete
- Check the App/Admin server environment variables for FastGPT v4.17.0, confirm that
AIPROXY_API_ENDPOINT(root address of the AI Proxy service) andAIPROXY_API_TOKEN(AI Proxy administrator Token) are correctly configured and not empty. - Upload a structured test case document, verify that associated data in the same group is fully preserved and not split incorrectly.
- Initiate a retrieval request, verify that the number of returned results matches the
limitconfiguration, and the similarity falls within the presetsimilarityinterval. - After enabling the citation display function, initiate a query related to test cases, verify that citation information for the corresponding documents appears at the end of responses.
- Simulate high-frequency test data updates, verify that retrieval results can be refreshed promptly to obtain the latest test reports and case information.
The configuration values provided are common starting points and should be evaluated against relevant 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.