What the data for this category looks like
The data sources for this category are the low-income population dynamic monitoring platform, resident family economic status verification system, and cross-department shared data from public security, human resources and social security, medical insurance, housing and urban-rural development, tax, and other departments. Daily processed documents include minimum living security application forms, family economic status verification reports, household survey forms, democratic review records, public notice materials, and assistance approval forms. Most document structures include fields such as family member information, income and property details, and approval process records. The update frequency is monthly, with a core focus on family economic status and assistance eligibility.
What constraints these characteristics impose on document parsing and chunking
Cross-department shared data has inconsistent formats, including structured tables, scattered text, and attached images. This requires the parsing link to support multiple file formats and mixed content. The monthly update frequency means high-frequency real-time parsing is not needed. However, when processing multiple approval and verification documents in batch, the parsing duration per file must be controlled. Most documents contain sensitive and closely related fields such as family member information and income and property details. When chunking, content that spans related fields must not be split. At the same time, complete extraction of key supporting materials must be ensured to prevent core information loss during chunking, which would affect subsequent verification and approval processes.
How to configure the settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Documents in this category mostly contain closely related family information and property details. This range avoids splitting core related fields, while ensuring the retrieval granularity fits the question matching needs of business process guidance. |
chunkSplitter | Chunk by paragraph + title | Daily processed documents include materials like minimum living security application forms and verification reports with hierarchical titles. Chunking by title preserves the logical structure of the document, making it easier to match corresponding processing links later. |
customPdfParse | Enable mixed image-text parsing | Daily documents include images such as household survey photos and scanned public notices. Enabling this setting extracts text embedded in images and preserves the association between images and text, preventing loss of image content. |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Single batch approval materials may include multi-page scanned documents and attachments. This upper limit covers conventional batch processing needs and prevents uploads from being blocked. |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Parsing cross-department shared data may involve multi-format merging processing. This duration prevents mid-process timeouts during batch document parsing and ensures complete parsing. |
chunkSettingMode | Manual configuration | Documents in this category have low uniformity in format and field characteristics. Manual configuration allows precise matching of business needs and adapts to specific document structures not covered by the general mode. |
Three common configuration mistakes
- Symptom: After uploading a docx file containing images, the returned result shows invalid image identifiers or missing image content. Cause: The
customPdfParseconfiguration is not enabled, or the parsing mode does not adapt to mixed image-text documents. - Symptom: Parsing duration for single batch documents exceeds expectations, and a timeout status code is returned. Cause: The value set for
PARSE_FILE_TIMEOUT_SECONDSis too low, failing to match the duration requirements of cross-department data merging parsing. - Symptom: Fields in excel-format family income details are split chaotically after parsing, and a single income record is split into multiple independent chunks. Cause: The
chunkSplitterconfiguration is not adjusted for structured tables, and table rows are mistakenly split into independent content, breaking the association of income details.
How to verify the configuration is correct
- Upload a single docx test document containing images. Verify that the parsed result includes embedded image text and the original document structure, confirming the mixed image-text parsing configuration is active.
- Upload batch multi-format documents, monitor parsing task duration, and adjust the value of
PARSE_FILE_TIMEOUT_SECONDSuntil tasks complete normally. - Upload an excel-format family income details document. Verify that parsed chunks preserve the association of table rows, confirming the chunking configuration adapts to structured data.
- Upload a single test document close to the
UPLOAD_FILE_MAX_SIZEupper limit. Confirm that the upload and parsing process does not trigger size limit blocking.
The values provided here are common starting points, and should be evaluated 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.