What this category's data looks like
Data for this category primarily comes from the national medical insurance information platform, hospital HIS and medical insurance interfaces, medical insurance catalog coding database, and cross-regional medical settlement platform. Daily processed documents include medical insurance settlement lists, expense details, medical record homepages, prescriptions, rejection lists, and settlement reconciliation statements. Fields cover settlement numbers, expense amounts, diagnosis and treatment item codes, insured person identity identifiers, and more. The update frequency is daily. Data is associated with entities such as designated hospitals, medical insurance handling institutions, and medical insurance information platform operators, covering ledgers and audit vouchers related to single or bulk medical expense payments.
Constraints Imposed on Knowledge Base Retrieval and Recall
Dispersed data sources with daily updates require the knowledge base to support multi-source data synchronization access and incremental updates, to avoid retrieving expired medical insurance catalog or settlement data. Document structures vary, including structured expense detail tables and unstructured medical record text, requiring retrieval and recall adapted to mixed formats. Fields include coding and amount content; retrieval must accurately match diagnosis and treatment item, drug and consumable codes, while verifying consistency between expenses and documents, with higher relevance requirements for recall results. In bulk settlement audit scenarios, multiple documents must be processed at once, requiring the retrieval link to support bulk calls to the knowledge base to avoid delays from single-item processing.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the accuracy requirements for medical insurance codes and expense details. Too low will introduce irrelevant settlement records, too high will miss valid matching items. Adapts to the similarity calculation logic of FastGPT v4.17.0 |
limit | Top 10 entries | The volume of documents to be processed in a single settlement audit is moderate. Too many will exceed the context window, too few will fail to cover associated rejection and reconciliation records |
searchMode | hybrid | Balances retrieval for structured coding fields and unstructured text, adapting to mixed-format medical insurance documents |
usingReRank | Enabled | Re-ranks recall results to improve the accuracy of code matching and expense consistency verification |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands retrieval terms, supplements synonyms and code aliases for medical insurance catalogs, improving matching coverage |
collectionIds | Bind knowledge base collections corresponding to medical insurance catalogs, settlement details, and rejection lists | Precise limitation of the retrieval scope to avoid interference from irrelevant data |
Three Common Misconfigurations
- Phenomenon: After uploading a single medical insurance settlement document, the number of segments displayed by the system does not match expectations, and duplicate index entries appear. Cause: The default chunking-related triggering logic is not adjusted. Repeated headers or fixed-format text in the document are repeatedly identified as independent chunking units.
- Phenomenon: Retrieval results do not hit the specified medical insurance catalog code matching items, only returning irrelevant general medical documents. Cause: The
collectionIdsparameter is not configured correctly, and the knowledge base collection corresponding to the medical insurance catalog is not bound. - Phenomenon: When calling the knowledge base for retrieval, the number of returned results does not match the set
limitparameter, with extra or missing entries. Cause: AfterdatasetSearchUsingExtensionQueryis enabled, automatically expanded retrieval terms introduce additional associated documents, and the value range oflimitis not adjusted synchronously.
How to Verify Proper Configuration
- Upload a standard medical insurance settlement list document, check the chunking results, confirm that the number of segments matches expectations, and there are no duplicate chunking entries.
- Manually enter a retrieval term for a diagnosis and treatment item code, check whether the similarity score of the recall results falls within the preset 0.75–0.85 range.
- Switch the retrieval mode between hybrid and pure text mode, compare the recall results, confirm that mixed-format documents can be accurately matched.
- Bind multiple specified knowledge base collections, test with retrieval terms, confirm that only document content within the corresponding collections is returned.
The values provided here 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.