What the data for this category looks like
Data sources for this category include the national medical insurance information platform, hospital HIS and medical insurance interfaces, medical insurance catalog and coding library, and cross-regional medical treatment settlement platform. Daily processed documents include medical insurance settlement lists, expense details, medical record homepages, prescriptions, rejection lists, and settlement statements. Data update frequency is daily. Data fields cover diagnosis item codes, pharmaceutical and consumable codes, expense amounts, insured person identity identifiers, medical record associated information, and other standard medical insurance fields required for business operations and their corresponding units. No extra non-standard statistical content is included.
What constraints these characteristics impose on knowledge base retrieval and recall
The daily data update requirement means the knowledge base must support incremental synchronization to avoid retrieving outdated medical insurance catalogs, settlement rules, or latest settlement data. Documents contain standardized coding fields. Retrieval must match both text content and coding information to ensure the accuracy of audit results. Document lengths vary widely for single or batch processing. Some settlement lists or detail texts may exceed conventional segment length thresholds, so long text segmentation and retrieval must be supported. Rejection lists and settlement statements have associated relationships. Retrieval must support cross-document associated matching to complete reconciliation audits for rejection appeals.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Adapts to the matching accuracy requirements for medical insurance coding and business text, filters low-correlation retrieval results |
limit | Top 8 entries | Balances retrieval result coverage and context length, adapts to the processing rhythm of single or batch settlement audits |
searchMode | Hybrid retrieval | Supports both semantic matching and precise coding matching, covering retrieval needs for different audit scenarios |
usingReRank | Enabled | Re-ranks initial retrieval results, prioritizes returning high-matching medical insurance details, coding, and rule content |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands synonymous queries related to coding, improving retrieval coverage for medical insurance coding matching |
collectionIds | Bind dedicated medical insurance knowledge base collections | Isolates dedicated data for this category, avoiding confusion with retrieval results from other business knowledge bases |
Three Common Mistakes
- Knowledge base index status remains stuck on "Indexing" with no completion prompt. Cause: No incremental synchronization trigger rule is configured, and no reasonable chunked indexing parameters are set during full synchronization of large-volume medical insurance data, resulting in index task timeout or blocking.
- Retrieved content is adjusted or rewritten, and does not strictly match the original source document. Cause: The automatic question optimization function is not disabled, and the system performs unnecessary semantic reorganization on retrieved medical insurance detail content, deviating from the accurate wording of the source document.
- After custom delimiter configuration is completed, paragraph merging or single-paragraph truncation still occurs. Cause: The set chunk length does not adapt to the single-line detail length of medical insurance settlement lists, and the standard line break separation logic in documents is not correctly matched, resulting in the system automatically adjusting segmentation results.
How to Confirm Configuration is Complete
- Access the FastGPT v4.17.0 knowledge base management page. Check the sync status of the target medical insurance knowledge base. Confirm that the most recent sync task has completed and has no error logs.
- Submit a test query containing a medical insurance code. Verify that the matching content of retrieval results covers the original text of specified documents such as medical insurance catalogs and settlement lists.
- Check the values of retrieval configuration items. Confirm that settings such as
similarityandusingReRankmatch the preset plan, and have not been temporarily modified. - Import a test medical insurance settlement detail document. Set custom delimiters and chunk parameters. Manually trigger segment preview. Confirm that the segmentation results conform to the actual structure of the document.
Note: The configuration values listed above are common starting points. They should be evaluated against the user’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.