What the data for this category looks like
The core data for this category comes from the national medical security information platform, hospital HIS and medical insurance interfaces, medical insurance catalog and coding database, and cross-regional medical treatment settlement platform. Daily processed documents include medical insurance settlement lists, cost details, medical record homepages, prescriptions, rejection lists, and settlement reconciliation statements. Data is updated daily. Document fields include diagnosis and treatment item codes, medical supplies codes, cost amounts, insured person information, settlement dates, handling agency identifiers, and more. Units include yuan, person-times, standard code numbers, and others. Document structures vary significantly. Settlement lists are mostly tabular details, while prescriptions are itemized records.
What constraints these characteristics impose on the document parsing and chunking link
Daily updated batch documents require the parsing link to have efficient processing capabilities. This avoids parsing delays that disrupt settlement audit workflows. Multiple document types require chunking rules to adapt to different formats. The rules must preserve the integrity of key fields such as codes and amounts. This prevents policy basis matching errors after chunking. Metadata such as settlement dates and handling agencies is an important association basis for policy Q&A. Chunking must extract and bind this metadata to the corresponding text block. Without this, accurate policy basis association is impossible. Parsing large batch files requires controlling timeout risks. This avoids partial settlement data going unprocessed due to interruptions.
How to Configure Settings
The following configurations apply to FastGPT v4.17.0
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the length of single settlement records or policy clauses in medical insurance documents. Prevents splitting of key code and amount fields |
chunkSplitter | Mixed paragraph + fixed length | Preserves complete context for policy basis. Adapts to structural differences across documents. Ensures chunked content can independently associate with policies |
customPdfParse | Enable image text extraction mode | Adapts to stamped and detailed tabular images common in medical insurance documents. Prevents unretrievable image content |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Supports large batch uploads of settlement reconciliation statement files. Meets daily batch processing requirements |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Reserves sufficient time for parsing large settlement lists. Avoids parsing interruptions due to timeout |
chunkSettingMode | Manual mode | Adjusts chunking rules for different document types such as prescriptions and rejection lists. Meets precise retrieval needs for policy basis Q&A |
Three Common Configuration Mistakes
- Symptom: After uploading a PDF-format settlement list, the parsing result does not display the detailed tabular image content from the original document. Cause: The image extraction configuration of
customPdfParseis not enabled. Only plain text content is extracted, resulting in loss of policy basis associated with images. - Symptom: When searching for policy questions, the returned chunked results do not include metadata such as settlement order numbers and insured person information. This makes it impossible to associate specific business scenarios. Cause: Metadata fields in the document header are not retained during chunking, or metadata binding rules are not configured. This leads to retrieval results that cannot match business context.
- Symptom: After uploading an Excel-format cost detail file, the system prompts that the file type is not supported, and parsing fails. Cause:
customPdfParseis not configured to adapt to Excel table parsing, or the corresponding format parsing switch is not enabled. This prevents extraction of cost detail content from tables.
How to Confirm Configurations Are Correct
- Upload a single-page PDF-format medical insurance settlement list. Check if the parsing result includes the tabular image text content from the original document. Confirm that the
customPdfParseconfiguration is active. - Upload a single Excel-format cost detail file. Verify that the system parses the file normally and generates chunked data. Confirm that the file format adaptation configuration is correct.
- Run a test policy question search. Check if the returned results include metadata for the corresponding chunks, such as settlement dates and handling agency identifiers. Confirm that the chunking rules retain metadata information.
- Upload a settlement reconciliation statement file of approximately 500 MB. Verify that the parsing process does not trigger a timeout error. Confirm that the
PARSE_FILE_TIMEOUT_SECONDSconfiguration is appropriate.
The values provided here are common starting points. They should be evaluated against samples specific to the actual deployment scenario.
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.