What the data for this category looks like
This category's data mainly comes from the national medical insurance information platform, hospital HIS and medical insurance interfaces, medical insurance catalog and coding database, and cross-provincial medical treatment settlement platform. The data updates daily. Daily processed documents include medical insurance settlement lists, expense details, medical record homepages, prescriptions, rejection lists, and settlement reconciliation statements. Most documents are structured tables or standard text with fixed fields, including core fields such as settlement number, diagnosis and treatment project code, expense detail items, and rejection judgment basis. Units include amount, number of visits, settlement days, and more.
What constraints do these characteristics impose on the "document parsing and chunking" link
Since most data in this category comes from tables and fixed-format text exported from structured systems, the document parsing link must retain associations between fields to avoid splitting that damages structured business information. Daily updated batch documents require the parsing link to adapt to high-frequency processing rhythms, and strictly limit the time taken to parse a single document. Medical insurance-related documents contain professional diagnosis and treatment project, drug and consumable coding fields. When chunking, retain the contextual association between codes and corresponding expenses and judgment basis, and do not split only by fixed character length. Documents such as rejection lists have multi-state associated fields. Chunking must avoid splitting the judgment basis of the same business into different chunks, which affects the integrity of subsequent audit logic.
How to configure the settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the length of structured fields in medical insurance documents, retains the contextual association between codes and expenses |
chunkSplitter | Split by paragraph + heading | Matches the fixed format structure of medical insurance documents, avoids splitting that damages business associations |
customPdfParse | Enable structured parsing mode | Adapts to table data in PDF formats such as medical insurance settlement lists, retains original field layout |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Adapts to the parsing time of large batch medical insurance documents, avoids timeout interruptions |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Adapts to batch uploaded medical insurance settlement document packages, avoids upload failures |
chunkSettingMode | Manual configuration | Adjust chunking rules for professional fields in medical insurance documents, adapts to business audit logic |
Three common mistakes
- Symptom: Calling the API to parse medical insurance documents shows a successful call, but the returned result contains no parsed text or field information. Cause: The
customPdfParseparameter is not configured correctly, or the correct parsing mode parameter is not carried when calling the interface. - Symptom: When uploading a batch of medical insurance document packages with multi-level subdirectories, only documents in the top-level directory are parsed and retrieved. Cause: Directory recursive parsing configuration is not enabled, and the chunking link does not traverse all document files in subdirectories.
- Symptom: Split-related error logs appear when parsing medical insurance settlement list PDFs, and the parsing task is interrupted. Cause: The document has damaged table nested structures, and the
chunkSplitterconfiguration does not adapt to complex table layouts.
How to confirm the configuration is correct
- Upload a single standard medical insurance settlement list PDF, check whether the parsed text retains the association between diagnosis and treatment project codes and corresponding expenses
- Call the parsing API to submit batch medical insurance documents, confirm that the number of parsed result entries returned exactly matches the number of uploaded files
- View the running logs of the parsing module, confirm that there are no timeout or split-related error messages
- Manually intercept complete data of a single medical insurance rejection business, verify that the judgment basis is not split into different paragraphs after chunking
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.