What this category of data looks like
Relevant data sources for integration and API testing include API management platforms, test environment logs, databases, and message queues. Daily maintained documents include API test cases, commissioning issue tickets, environment descriptions, and test reports. Data is updated daily or when a new version is released. Data fields include API paths, request parameters, response status codes, error stacks, case numbers, environment identifiers, and more. Data formats vary across different sources. Some log data contains unstructured text, while test cases are structured configuration entries.
What constraints do these characteristics impose on reranking and hybrid retrieval workflows
The fact that data is scattered across multiple systems means hybrid retrieval must integrate information from different sources, to avoid missing information from only searching a single platform. The high update frequency means retrieval logic must adapt to real-time or near-real-time content changes, to avoid returning expired commissioning configurations or logs. The coexistence of multiple structured and unstructured fields means the reranking step must match both structured case parameters and unstructured error text, to improve the accuracy of relevance judgment. The need for quick positioning in operational on-call scenarios means reranked results must prioritize content most relevant to the current ticket, to reduce manual filtering time.
How to configure settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
usingReRank | Enabled | Integration testing scenarios require distinguishing the relevance of retrieval recall results, to prioritize displaying the most matching commissioning issues and logs |
similarity | 0.75–0.85 | Adapts to text similarity matching for API test cases and logs, to avoid recalling irrelevant cross-module configurations |
limit | Top 10 entries | Controls the volume of basic results returned per retrieval, to avoid information overload in operational on-call scenarios |
searchMode | hybrid | Performs hybrid retrieval of API data and logs from different sources, to cover all information needed for commissioning issue positioning |
Three common configuration errors
- Symptom: The reranking switch shows as enabled, but returned results are not reranked by relevance. Cause: The
AIPROXY_API_ENDPOINT(root address of the AI Proxy service, for example http://aiproxy:3000; do not append paths such as /v1 or /chat/completions, and do not configure as a single model address such as VLLM, Ollama, or Xinference) andAIPROXY_API_TOKEN(AI Proxy administrator token, must match the AI Proxy server configuration and cannot be empty; do not use other types of credentials such as application API keys or model vendor keys) for the FastGPT v4.17.0 version App/Admin server are not configured or are configured incorrectly, causing the reranking service to fail to call normally. - Symptom: The number of retrieval return results does not match the configured
limit. Cause: The quantity parameters for recall and reranking were not adjusted synchronously, causing intermediate results to be truncated early. - Symptom: The reranking function fails to start on arm architecture deployments. Cause: A reranking model image that only supports amd architecture was used, and the arm architecture deployment package was not adapted.
How to confirm configuration is complete
- Review the environment variables of the App/Admin server to confirm that
AIPROXY_API_ENDPOINT(root address of the AI Proxy service; do not append paths such as /v1 or /chat/completions, and do not configure as a single model address such as VLLM, Ollama, or Xinference) andAIPROXY_API_TOKEN(AI Proxy administrator token, must match the AI Proxy server configuration and cannot be empty; do not use other types of credentials such as application API keys or model vendor keys) are correctly configured and not empty. - Initiate a test retrieval, compare the returned results when
usingReRankis enabled and disabled, to confirm that the sorting logic has changed. - View the logs of the AI Proxy service, to confirm there are no error messages about failed reranking service calls.
- Verify that hybrid retrieval returns relevant content from API cases, logs, and databases at the same time, to confirm that the
searchModeconfiguration is effective.
The values provided here are common starting points, and should be evaluated against deployment-specific 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.