What this category's data looks like
Core data for this category comes from the national basic catalog of government service items, administrative licensing item list, power and responsibility list, and local government service item databases, with a monthly update frequency. Daily processed documents include basic government service item catalogs, implementation lists, service guides, standard material directories, flowcharts, and frequently asked questions. Documents are primarily characterized by structured field combinations, including fixed fields such as item name, acceptance conditions, application materials, and handling procedures. Some historical archived documents are in scanned image format, requiring text extraction.
What constraints do these characteristics impose on the document parsing and chunking step
Full monthly updated item list files are usually large in size, so bulk parsing and chunking support is required. The structured fields of documents must retain their associated relationships. When chunking, business content such as acceptance conditions and corresponding application materials, handling procedures and corresponding items must not be separated. Some historical archived documents are in scanned image format, so text extraction and OCR processing support is needed. There are obvious structural differences between different types of documents, so differentiated chunking rules must be adapted to ensure field integrity for subsequent ledger summary and report generation.
How to set the configuration
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Fits the length range of single business fields in government documents, avoids splitting associated business content |
chunkSplitter | Chunk by title + paragraph + table | Government documents have many structured fields; splitting by document structure preserves the association between items, conditions, and materials |
customPdfParse | Enable OCR mode | Processes scanned historical service guide documents to fully extract text content |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Full item list files take a long time to parse; prevents parsing interruptions due to timeout |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Supports bulk import of full monthly updated government service item documents |
chunkSettingMode | Manual configuration | Adjust chunking parameters for different document types such as catalogs, service guides, and flowcharts to fit business needs |
Three common mistakes
- Unable to select table datasets as knowledge base materials: The phenomenon is that table files for standard material directories cannot be selected in the knowledge base configuration interface. The cause is that the structured table parsing mode is not enabled, so the system does not recognize table content.
- Single large file parsing timeout: The phenomenon is that after uploading a full item list, the system prompts parsing failure and returns a timeout status code. The cause is that the
PARSE_FILE_TIMEOUT_SECONDSparameter was not adjusted; the default timeout duration is insufficient for processing large-capacity files. - Associated content split after chunking: The phenomenon is that the acceptance conditions and corresponding application materials in the service guide are split into different chunks. The cause is that fixed-length chunking was used, and document structure-based splitting was not implemented, resulting in associated business content being separated.
How to confirm the configuration is correct
- Upload a single structured service guide PDF, check if the parsed text fully includes core fields such as item name, acceptance conditions, and application materials.
- Check the parsing log to confirm no timeout errors occurred, and the file upload size does not exceed the
UPLOAD_FILE_MAX_SIZElimit. - Compare the chunking results with the original document to confirm that associated business content (such as acceptance conditions and corresponding application materials) is not split into different chunks.
- Bulk upload monthly updated item list files, verify that the parsing progress is normal and the results have no obvious missing content.
The values given 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.