What this category of data looks like
Data for this category comes from the national medical insurance information platform, hospital HIS and medical insurance interfaces, medical insurance catalog and coding library, and cross-regional medical settlement platform. The update frequency is daily. Commonly processed documents include medical insurance settlement lists, expense details, medical record homepages, prescriptions, rejection lists, and settlement reconciliation statements. Document fields cover diagnosis and treatment item codes, pharmaceutical and consumable codes, expense details, insured person identity information, settlement status, and more. The data volume of a single document varies with the number of settlement transactions, containing both structured coding fields and unstructured descriptive content.
Constraints on document parsing and chunking
These characteristics create clear constraints for the document parsing and chunking stage: Batch documents updated daily must align with high-frequency processing rhythms to avoid parsing timeouts. Documents include both structured coding fields and unstructured business descriptions. Chunking must preserve the link between codes and their corresponding expense details to prevent splitting from disrupting business logic. Settlement documents follow a defined audit workflow. Chunking must cover full information for a single settlement transaction, ensuring audits can match diagnosis and treatment items, pharmaceutical and consumable codes to expense data. Some documents have image attachments such as prescriptions and medical record pages. Parsing must process text and image association data together to avoid losing business-related information.
How to configure settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Matches the text length linked to single settlement details and codes in medical insurance settlement documents, avoiding splitting that breaks business associations |
chunkSplitter | Chunk by paragraph + semantic grouping | Adapts to structured and mixed content in medical insurance documents, retaining complete context for business nodes |
customPdfParse | Enable dedicated parsing template for medical insurance documents | Optimizes parsing logic for medical insurance-specific document formats such as settlement lists and prescriptions, improving extraction accuracy for structured fields |
PARSE_FILE_TIMEOUT_SECONDS | 120–180 seconds | Adapts to parsing times for batch medical insurance documents, preventing timeouts caused by oversized single documents or batch processing |
UPLOAD_FILE_MAX_SIZE | 50–100 MB | Matches the upload limit for single-batch bulk settlement documents, aligning with the scale of documents exported daily by handling institutions |
Three common configuration errors
Phenomenon 1
After importing a Word-format medical insurance expense detail document, image links in the parsed Markdown text lose their domain name.
Cause 1
No image path replacement rule is configured for the custom document parsing template. The default parsing only retains local relative paths and does not bind the internal access domain name of the handling institution.
Phenomenon 2
A timeout error is returned when calling the document parsing interface, or the parsing task status shows failure.
Cause 2
The PARSE_FILE_TIMEOUT_SECONDS parameter is not adjusted to a value suitable for batch medical insurance documents. Parsing time for a single large settlement list exceeds the default threshold.
Phenomenon 3
Only part of the content of a batch-imported medical insurance settlement list is parsed, with subsequent content missing.
Cause 3
The UPLOAD_FILE_MAX_SIZE parameter is not adjusted to a value suitable for the document scale, exceeding the default limit resulting in incomplete parsing.
How to confirm the configuration is correct
- Upload a single typical medical insurance settlement list document, review the parsed text structure to confirm coding fields and expense details are not incorrectly split.
- Batch upload multiple types of medical insurance documents such as prescriptions and rejection lists, verify all parsing tasks complete normally with no timeout or truncation prompts.
- Check parsing results for image-based documents, confirm image links include complete access paths, and adjust based on internal domain name configuration as needed.
- Trigger a chunking test, review chunked text blocks to confirm complete information for a single settlement transaction is included in a single or adjacent chunks.
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.