What this category of data looks like
Data sources for this category include the National Medical Security Information Platform, hospital HIS and medical insurance interfaces, the Medical Insurance Catalog Coding Database, and the Cross-provincial Medical Settlement Platform. The update frequency is daily. Daily processed documents include medical insurance settlement lists, expense details, medical record homepages, prescriptions, rejection payment lists, and settlement reconciliation statements. Documents contain fields such as diagnosis and treatment item codes, medical supplies codes, expense amounts, insured person identity identifiers, and settlement status. Some batch-processed documents have large individual file sizes.
What constraints do these characteristics impose on the document parsing and chunking stage
Since data is updated daily and sources include heterogeneous documents from multiple systems, the parsing stage must support medical insurance business documents in different formats. Batch-processed documents have large individual file sizes, so the chunking stage must support large file parsing to avoid timeouts or out-of-memory issues. Documents contain a large number of standardized coding fields. When chunking, the association between codes and corresponding expenses, settlement status must be retained to avoid losing contextual connections after splitting. Frequently updated data requires that the parsing and chunking process can adapt to incremental processing, ensuring accurate matching between the latest medical insurance catalog codes and settlement data.
How to set configurations
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value/Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Medical settlement documents include short fields such as codes and amounts, as well as long-text settlement descriptions. This range balances contextual completeness and retrieval efficiency |
chunkSplitter | Semantic chunking | Medical insurance documents have close associations between codes, expenses, and insured person information. Semantic chunking preserves business logical connections |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Adapts to large batch medical insurance settlement lists and expense detail documents |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Large file parsing requires sufficient time for format conversion and content extraction |
customPdfParse | Enabled | Medical insurance PDF documents are mostly structured forms. Enabling custom parsing preserves field structure |
chunkSettingMode | Manual configuration | Adapts to the fixed field structure of medical insurance documents. Manually adjusting chunking parameters allows for more precise configuration |
Three common mistakes
- Phenomenon: After importing spreadsheet documents, core fields such as codes and amounts are empty in the parsing results. Reason: Structured document parsing configuration is not enabled. The default general chunking cannot recognize the column association relationships of medical insurance spreadsheets.
- Phenomenon: After uploading a medical insurance PDF settlement list, the retrieval results cannot open the original document attachment. Reason: Document storage permissions or associated paths are not configured. The parsing stage did not retain the original file's associated information.
- Phenomenon: When calling the batch parsing interface, the interface returns a success status code but no corresponding chunk data is generated. Reason:
PARSE_FILE_TIMEOUT_SECONDSis set too small. The parsing process for large medical insurance documents is terminated before completion.
How to confirm the configuration is correct
- Upload a standard medical insurance settlement list PDF, check whether the parsing results include core fields such as diagnosis and treatment item codes and expense amounts.
- Check the file upload configuration items, confirm that the value of
UPLOAD_FILE_MAX_SIZEcovers the maximum document size processed daily. - Initiate a batch document parsing request, verify that the number of returned chunks matches the document content length.
- Check the knowledge base import configuration, confirm that the structured parsing switch for spreadsheet documents is enabled.
The values provided are common starting points and should be measured 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.