What data in this category looks like
Data in this category primarily originates from defect management systems, test reports, code commit records, and user feedback. Daily generated documents include defect tickets, regression test checklists, repair verification records, and defect statistical reports. Data fields include unique defect identifiers, priority levels, reproduction steps, associated version numbers, verification status, and more. Data update frequency follows daily or per-version release cycles, and changes dynamically alongside defect repairs, test execution, and version iterations.
What constraints do these characteristics impose on reranking and hybrid retrieval
Data in this category is layered by version, defect priority, and repair status, and is updated frequently with version iterations. Hybrid retrieval must match both semantic relevance and metadata features, to avoid high-priority defects being obscured by low-priority results when relying solely on semantic similarity. The length of different document types varies widely, ranging from hundreds of characters for a single defect reproduction step to thousands of characters for summary reports. The reranking model must adapt to a wide range of input lengths. The data update frequency means retrieval must support dynamic refreshing of associated defects and test data to ensure retrieval results match the current version state.
Configuration Settings
These settings apply to FastGPT v4.17.0 and later versions.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
usingReRank | Enabled | This category of data requires differentiation of defect priority and relevance; reranking optimizes the relevance ranking of retrieval results |
similarity | 0.75–0.85 | Filters low-semantic-relevance defects and test data to avoid irrelevant results interfering with retrieval |
limit | Top 10 entries | Controls the number of basic results returned per retrieval, adapting to the single-batch processing scale of defect management data |
searchMode | hybrid | Combines semantic similarity and metadata features to match the mixed defect and test document types in this category |
| Reranking model timeout threshold | 30 seconds | Prevents retrieval process blocking due to reranking model response delays, adapting to the real-time requirements of defect troubleshooting |
Three Common Configuration Mistakes
- Phenomenon: The result reranking option cannot be checked when creating an application. Cause: The server has not configured
AIPROXY_API_ENDPOINT(the root address of the AI Proxy service, do not append paths such as /v1) andAIPROXY_API_TOKEN(the administrator token consistent with the AI Proxy server configuration, cannot be empty), causing service startup verification to fail, failing to connect to the AI Proxy, failing to load the reranking model function dependent on AI Proxy, and thus unable to check the result reranking option. - Phenomenon: The order of retrieval results does not change after configuring the reranking model. Cause: The
usingReRankparameter is not set to enabled, or thesimilaritythreshold is set too high, filtering all basic retrieval results. - Phenomenon: The reranking model call returns a timeout error. Cause: The reranking model timeout threshold is set too short, or the reranking model deployment resources are insufficient to complete sorting calculations within the specified time.
How to Confirm Proper Configuration
- Check that the server environment variables have been configured with
AIPROXY_API_ENDPOINT(the root address of the AI Proxy service, for example http://aiproxy:3000, do not append paths such as /v1) andAIPROXY_API_TOKEN(the administrator token consistent with the AI Proxy server configuration, cannot be empty), and that the parameter values match the AI Proxy server configuration. - Enter the application's retrieval configuration page, confirm that the
usingReRankswitch is in the enabled state, and that thesimilarityandlimitparameters meet current business requirements. - Submit a test query containing a defect ID and version number, and check whether the order of retrieval results matches the defect priority and relevance.
- Verify the reranking model call logs to confirm there are no timeout or authentication failure error messages.
The values provided here are common starting points and should be evaluated 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.