What the data for this category consists of
Core data for this category comes from the Chinese Center for Disease Control and Prevention Information System, medical institution HIS systems, infectious disease reporting systems, and laboratory testing systems. The update frequency is daily. Daily generated business documents include infectious disease report cards, epidemic analysis reports, case correction records, clustered epidemic reports, and epidemiological investigation reports. Data fields include case unique ID, diagnosis time, reporting unit, statutory infectious disease category, correction time, correction reason, and others. Some documents include structured statistical fields and unstructured epidemiological investigation text. Data uses individual case records or single reports as the smallest storage unit.
What constraints these characteristics impose on knowledge base retrieval and recall
Daily updated data requires the knowledge base to support incremental synchronization. This avoids performance loss caused by full reindexing. Multi-source, heterogeneous data structures require retrieval to cover both structured fields (such as case ID, correction time) and unstructured text (such as epidemiological investigation descriptions). This ensures accurate recall of relevant case records. Statutory infectious disease reporting compliance requirements mean retrieval results must prioritize matching fields related to statutory categories and reporting time limits. The system must also support quick recall of historical correction records to check for duplicate and incorrect report cards. Individual records have strong connections. Retrieval must avoid returning irrelevant cross-case documents. This reduces screening costs for subsequent ledger summary.
Configuration Settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.72–0.80 | Must accurately match core fields such as case ID and diagnosis time. This avoids filtering valid correction reports or introducing irrelevant documents |
limit | Top 10 entries | Ledger summary requires comparing multiple historical reports. Too many entries increase context redundancy. Too few entries fail to cover the complete correction record chain |
searchMode | vector+keyword hybrid mode | Balances precise matching of structured fields and semantic recall of unstructured text. This adapts to multi-source, heterogeneous data structures |
usingReRank | Enabled | Re-ranks recall results by case correction time. Prioritizes returning the latest correction records. This simplifies ledger verification workflows |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands query keywords. For example, expands "duplicate report cards" to "duplicate report cards, duplicate case records". This improves recall coverage |
collectionIds | Specify dedicated knowledge base collection IDs for disease control | Isolates documents from other categories. This avoids retrieving irrelevant daily medical records. Ensures retrieval results focus on data from this category |
Three Common Configuration Mistakes
- Phenomenon: Retrieval results do not include image content in uploaded Word documents. Cause: Retrieval and recall configuration for in-document multimedia content is not enabled. Only text content is recalled.
- Phenomenon: The number of returned retrieval results is far lower than expected, with only 2-3 entries returned. Cause: The
similaritythreshold is set too high. This filters eligible correction reports and historical epidemic records. - Phenomenon: The number of selectable knowledge bases is fewer than 30 after local deployment. Cause: The default knowledge base limit from the public version of FastGPT is incorrectly retained. The open-source version’s configuration parameters are not modified.
How to Verify Successful Configuration
- Upload a Word document of an infectious disease report card that includes case photos. Initiate a query containing "case photos" and "correction record photos". Confirm that retrieval results include image content.
- Initiate a query for "duplicate report cards corrected in June 2024". Check that the number of returned results matches the
limitsetting. Confirm that results are sorted by most recent correction time first. - View FastGPT retrieval logs. Confirm that
searchModeis set to vector+keyword hybrid mode, andusingReRankis enabled. - Verify the source of retrieved documents. Confirm that only documents from the dedicated disease control knowledge base collection are included. No irrelevant content from other categories is present.
The values listed here are common starting points. They should be evaluated against relevant sample datasets.
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.