Reference only; verify against the target FastGPT release and deployment documentation.
What This Use Case’s Data Looks Like
The data for declaration materials and case review of state-owned asset management entities mainly comes from project declaration documents required by state-owned asset supervision regulations, internally archived property transaction cases, asset appraisal reports, and similar materials. Data updates follow project progress cycles. Periodic declaration materials are generated during a single project cycle, and bulk cases are archived quarterly or annually. Most documents include fixed-format approval pages, long main text chapters, and attachment materials in scanned form. Fields include project number, asset category, appraisal value, approval document number, and others. Units involved include ten thousand yuan, square meters, equity ratio, and similar units.
What Constraints These Characteristics Impose on Document Parsing and Chunking
The declaration materials and cases of state-owned asset management entities contain many structured fields, and the context logic of the approval process must be preserved. Therefore, core approval clauses and associated fields must not be split during parsing. There are many long main text paragraphs, so the hierarchical structure of different chapters must be adapted to prevent chunking from destroying content integrity. Some cases include multi-page scanned attachment materials, so OCR extraction must be handled to avoid default parsing ignoring image content. Periodically updated documents must match project metadata to ensure chunked content corresponds to project milestones. Bulk archived large files must also meet upload and parsing performance requirements to avoid task failure due to oversized files or excessive processing time.
How to Set Configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the long paragraph structure of declaration materials, avoids splitting core approval clauses and associated fields |
chunkSplitter | Split by paragraph + title | Matches the chapter hierarchy of declaration materials, retains the context logic of the approval process |
customPdfParse | Enable OCR mode | Processes scanned attachment materials in cases, extracts text content from images |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Meets the large file upload requirements for bulk case archiving |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Meets the time requirements for long document parsing, avoids interrupting parsing tasks due to timeout |
chunkSettingMode | Manual configuration | Adjust parameters for different types of documents such as asset appraisal reports and property transfer plans |
Three Common Mistakes
- Uploading a Java interface document with a modified file extension results in no parsing result. The cause is that the custom parsing rule for the corresponding format is not enabled. Modifying only the file extension cannot trigger the correct text extraction logic.
- Using chunk mode to call the pushdata API for upload results in long-term display of "indexing". The cause is that
chunkSizeis set too large, leading to excessive chunking time, orPARSE_FILE_TIMEOUT_SECONDSis set too low to complete full parsing. - Uploading a case PDF containing scanned attachments results in ignored image content. The cause is that the OCR mode of
customPdfParseis not enabled. The default parsing only extracts the text layer content of the PDF.
How to Confirm Configurations Are Set Correctly
- A single typical asset appraisal report PDF is uploaded, and the parsed text is reviewed to confirm it includes complete approval document numbers and asset value fields, ensuring structured content is not incorrectly split.
- Parsing task logs are examined to confirm the value of
PARSE_FILE_TIMEOUT_SECONDSmatches parsing duration, with no timeout errors occurring. - A case file containing scanned attachments is uploaded, and OCR-extracted image text is verified for completeness to confirm
customPdfParseis enabled. - The pushdata API is called to test chunk mode upload, and indexing status is checked to confirm it completes within a reasonable timeframe with no prolonged hanging.
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.