What this category of data looks like
Data for this category comes from the Chinese Center for Disease Control and Prevention Information System, medical institution HIS and infectious disease reporting systems, and laboratory testing systems. The update frequency is daily. Daily documents include infectious disease report cards, epidemic analysis reports, case correction records, clustered epidemic reports, and epidemiological investigation reports. Fields cover case unique identifiers, diagnosis time, reporting time limit, correction notes, test results, and more. Formats uniformly follow legal reporting specifications, with no additional custom fields.
What constraints these characteristics impose on the vector model and indexing link
Data from multiple systems has minor format differences. Complete standardized cleaning before entering the vector processing stage; this increases pre-verification costs. The daily update rhythm requires indexes to support incremental synchronization, avoiding excessive resource usage from full reconstruction. Multiple correction records exist for the same case. Retain version identifiers as chunk metadata to avoid confusion between different stages of report content during retrieval. Follow fixed fields required by legal reports when chunking; retain core identifiers. Avoid random truncation that causes loss of critical information, as this affects subsequent verification and early warning processes.
How to set configurations
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Chunk Length | 800–1200 characters | Matches the core content length of a single infectious disease report card, avoiding splitting critical information across documents |
| Overlap Length | 100–200 characters | Covers the context associated with case corrections, avoiding information breaks caused by split correction records |
| maxToken | 4096 | Adapts to the input limit of mainstream vector models, ensuring that long epidemic analysis reports can be fully vectorized |
| defaultToken | Calibrated via actual testing | Handles access permission verification for multi-system data, adapting to interface authentication rules of different data sources |
| Single Search Limit | Top 10 entries | Balances retrieval precision and query speed, meeting the needs of abnormal signal aggregation in disease control scenarios |
| Similarity Threshold | 0.75–0.85 | Distinguishes similar case reports from duplicate submissions, avoiding misjudgments of duplicate or incorrect report cards |
Three Common Misconfigurations
- A 401 Unauthorized error is returned when calling the vector model interface. Cause: The corresponding
defaultTokenpermission parameter is not configured for multi-system data sources, resulting in interception of cross-system data access. - Custom split documents are stored in the knowledge base, and duplicate content is automatically cleaned. This causes the preset index order to differ from the actual storage order. Cause: The case unique identifier is not added as metadata before chunking. The system uses content hash for deduplication by default, which destroys the original splitting logic.
- Vector model requests are incorrectly routed to large language model calls. Cause: The call paths for vector models and LLMs are not clearly distinguished in channel configuration. The transit platform mistakenly matches vector requests to general large model interfaces.
How to Verify Successful Configuration
- Import a single infectious disease report card document. Check that the chunked content retains the case unique identifier and core report fields, with no critical information truncated.
- Submit test data containing duplicate correction records. Confirm that the knowledge base retains all version contents, and does not automatically clean legitimate duplicate correction entries.
- Call the vector model interface. Check that the status code of the returned result is 200, and no permission interception error is triggered.
- Enable incremental index mode. Import the daily updated epidemic analysis report. Confirm that the index only synchronizes new content, and no full reconstruction is triggered.
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.