What the data for this category looks like
Core data for this category originates from the national land and spatial planning unified map, project approval and supervision platform, land use status database, and ecological protection redline data. Update frequency is irregular. Daily processed documents include land pre-examination and site selection opinions, construction land planning permits, construction engineering planning permits, redline drawings, and planning conditions. These include both standardized official documents and scattered approval attachments. Document structures include fixed modules such as project subject information, land scope, and planning control requirements. Fields cover project number, land scale, redline coordinates, and more. Some documents have formatted headers, footers, and official approval seals.
What constraints do these characteristics impose on document parsing and chunking?
Data sources are scattered and formats vary. The parsing process must adapt to both standardized official documents and scattered attachments with differing formats. Documents include key fields such as spatial coordinates and land scale. Chunking must avoid splitting associated information across fields. Splitting such information will impact subsequent planning control verification. Irregular update frequency means parsed document chunks must retain the original upload time and source system tag. This supports subsequent update verification. Some documents have redundant headers, footers, and approval seals. Extra filtering is required to avoid interference with key content recognition.
How to configure settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Single chunks of land pre-examination documents must contain complete project information blocks, to avoid splitting key fields such as project number and land scale |
chunkSplitter | Split by paragraph + heading | Standardized documents have fixed chapter structures. Splitting by paragraph and heading preserves contextual logic, and prevents cross-chapter control requirements from being split |
customPdfParse | Enabled | Most approval documents are standard PDFs with headers and footers. Enabling custom parsing filters redundant content and retains core body text |
UPLOAD_FILE_MAX_SIZE | 500 MB | Must support uploading large files such as CAD-exported PDFs of planning redline drawings and multi-page approval attachment packages |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Large planning documents require longer processing time, to avoid parsing failure due to timeout |
chunkSettingMode | Manual configuration | Parameters must be adjusted for different document types. Automatic mode cannot adapt to the fixed structure of standardized documents |
Three common mistakes
- Symptom: After server deployment, parsing uploaded files returns a 404 error, while parsing works normally in a local environment. Cause: The server has not correctly configured the storage path for uploaded files and the directory mapping for the parsing node, so the parsing node cannot access the uploaded files.
- Symptom: Key fields such as land scale and redline coordinates are empty in parsed document chunks. Cause:
customPdfParseis not enabled, and redundant content such as standardized PDF headers, footers, and approval seals is not filtered. This causes key text to be overwritten or truncated. - Symptom: Chunking results for multi-page planning condition documents include cross-chapter control requirements. Cause:
chunkSplitteris configured to split by fixed character count, and the document's heading hierarchy and paragraph boundaries are not recognized.
How to confirm correct configuration
- Upload a standard land pre-examination and site selection opinion PDF. Check if parsed text has filtered headers and footers, and if key fields are complete.
- Adjust
chunkSizeto different values, upload multi-page approval documents. Check if chunking results retain complete fields such as project number and land scale. - Upload large planning attachments under 500 MB. Check if parsing completes within 600 seconds with no timeout errors.
- View the uploaded file storage directory. Confirm that uploaded files after server deployment are synchronized normally with the parsing node's directory.
The values listed are common starting points, and should be evaluated against the relevant samples being used.
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.