What the Data for This Category Looks Like
This category's data comes from the China Disease Prevention and Control Information System, medical institution HIS and infectious disease reporting systems, and laboratory testing systems. Daily documents include infectious disease reporting cards, epidemic analysis reports, case correction records, clustered outbreak reports, and epidemiological investigation reports, covering structured forms and analytical texts. The data update frequency is daily. Fields involve core business content such as case diagnosis, reporting time limits, correction information, and monitoring signals.
Constraints Imposed on Knowledge Base Retrieval and Recall
Daily updated data requires the retrieval pipeline to support real-time incremental synchronization, ensuring policy basis always matches the latest reporting rules. Heterogeneous data from multiple systems requires retrieval to associate different knowledge base sets to enable cross-system information integration. Documents include short forms and long-text analysis reports, requiring retrieval to adapt to both short-field precise matching and long-context recall. Core fields involve statutory reporting rules, so recall result accuracy must meet regulatory requirements, avoiding missed or delayed reporting of critical information.
Configuration Settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Basis for This Setting |
|---|---|---|
similarity | 0.75–0.85 | This category's core fields involve statutory reporting rules. Balance recall accuracy and coverage to avoid missing critical case information |
limit | Top 8–12 entries | Daily documents include short forms and long analysis reports. Recall enough relevant content to cover policy basis while avoiding context overload |
searchMode | Vector retrieval first. Use hybrid retrieval when integrating multiple systems. Set timeout threshold based on actual testing | This category's data updates daily. Hybrid retrieval requires matching multiple features and has high time cost. Adjust based on deployment environment |
usingReRank | Enabled | Sort recall results by relevance to prioritize core fields such as statutory reporting time limits and case diagnosis, meeting business verification needs |
datasetSearchUsingExtensionQuery | Enabled | This category's policy basis Q&A requires expanding query synonyms, such as matching "case correction" with "card reprocessing", to improve recall coverage |
collectionIds | Configure corresponding collection IDs grouped by data source | Data comes from multiple systems. Associate knowledge base sets for disease control information systems, medical institution reporting systems, etc., to enable precise retrieval |
Three Common Errors
- Phenomenon: Hybrid retrieval calls take over 10 seconds, and some requests trigger timeouts. Cause: This category's data includes multi-system heterogeneous documents. Hybrid retrieval needs to match both vector and keyword features, and the trigger conditions for
searchModewere not adjusted based on deployment environment, leading to excessive resource usage. - Phenomenon: After uploading a PDF document, search test results are empty, and no parsing completion status prompt is displayed. Cause: Parsing rules for the corresponding document type were not configured, or the knowledge base synchronization task was not triggered at the daily update frequency, so data indexing was not completed.
- Phenomenon: The answer does not call knowledge base content and modifies policy basis on its own. Cause: Forced knowledge base recall configuration was not enabled, so the model did not associate category-specific statutory reporting rule documents, generating content that does not meet business requirements.
How to Confirm Proper Configuration
- Execute single queries related to infectious disease reporting and policy basis, and check if retrieval results include category-specific document types such as infectious disease reporting cards and epidemic analysis reports.
- View retrieval time statistics, confirm vector retrieval time meets expectations, and adjust
searchModetrigger conditions based on deployment environment for hybrid retrieval time. - Check knowledge base synchronization logs, confirm daily updated documents have completed indexing, with no parsing failures or synchronization exceptions.
- Verify the source of answer content, confirm all policy basis-related answers come from uploaded category-specific knowledge base documents, with no self-generated non-compliant content.
The values provided 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 2026-09-28.