What this category of data looks like
Data for this category comes from the China Disease Prevention and Control Information System, hospital HIS systems, infectious disease reporting systems, and laboratory testing systems. The update frequency is daily. Daily documents include infectious disease reporting cards, epidemic analysis reports, case correction records, clustered epidemic reports, and epidemiological investigation reports. Data fields include case unique ID, diagnosis date, reporting unit, correction flag, test results, and other fields. There are no custom non-standard units, and all fields follow the standardized format for statutory infectious disease reporting.
Constraints on knowledge base retrieval and recall
Daily updated data requires the retrieval system to support scheduled incremental index updates, avoiding full rebuilds that take excessive time. Fixed-format documents and dedicated fields require retrieval to accurately match core fields such as case ID and diagnosis time window, preventing irrelevant results from entering the case file review process. Multi-source data requires retrieval to cover cross-system associated documents, ensuring complete case-related records can be retrieved during review. Additionally, the review link has very high accuracy requirements for results. Any missed reports or incorrect recalls may cause epidemic alert delays or regulatory notifications, so retrieval recall precision and stability must strictly align with business compliance requirements.
How to set configurations
| Configuration Item | Recommended Value | Basis for This Setting |
|---|---|---|
similarity | 0.75–0.85 | Must accurately match core fields such as case ID and diagnosis date, to prevent low-relevance non-business documents from being recalled |
limit | Top 10 entries | Case file review requires checking multiple historical case records, while controlling context length to avoid exceeding model capacity limits |
searchMode | vector | For structured and semi-structured infectious disease reporting documents, vector retrieval can accurately match core business fields |
usingReRank | Enabled | Filter out interference items such as duplicate corrected cards and incorrect cards, to increase the proportion of valid results during review |
datasetSearchUsingExtensionQuery | Disabled | This category of documents has a fixed structure, query terms do not require additional expansion, to avoid introducing retrieval results unrelated to infectious disease reports |
collectionIds | Bound to dedicated knowledge base collections for disease prevention and control reports | Isolate other categories of medical data, ensuring the retrieval scope only covers dedicated documents for this scenario |
Three common mistakes
- Switching a knowledge base index triggers a 60-second timeout. The cause is forced index switching before index rebuilding is complete. The index volume of this category's daily updated data is large, leading to loading time exceeding the system's default timeout limit.
- The knowledge base remains in the training or rebuilding state for a long time and cannot be switched. The cause is that no scheduled incremental update task is configured. Full rebuilding of all daily updated data takes too long.
- The online conversation interface continues to search the knowledge base without responding. The cause is that the
similaritythreshold is set too low, leading to too many retrieval results exceeding the context capacity limit, orusingReRankis not enabled, causing a backlog of invalid results.
How to confirm the configuration is correct
- Manually retrieve a known case unique ID, confirm that the retrieved documents include category-specific content such as the corresponding case's reporting card and correction records.
- Check the FastGPT v4.17.0 index update log, confirm that the daily incremental update task is completed on time, with no timeout error records.
- Test the knowledge base index switching operation, confirm that the loading time is within 10 seconds, with no timeout prompt pop-up.
- Verify that the retrieval results do not include irrelevant daily medical treatment records, only include documents for this scenario such as infectious disease reports and epidemic analysis, and the relevance meets business review requirements.
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 2026-09-28.