What this category of data looks like
This category of data originates from the national land and space planning one-map system, project approval and supervision platform, land use status database, and ecological protection red line data, with irregular update frequency. Daily processed documents include Land Use Pre-examination and Site Selection Opinion, Construction Land Planning Permit, Construction Engineering Planning Permit, red line drawings, and planning conditions. Core fields cover project subject, land use scope, control requirements, approval progress, and similar items. Most documents are official document formats with coordinates and tables, and some are scanned paper approval files.
What constraints do these characteristics impose on the "document parsing and chunking" link
Documents in this category come from scattered sources and have no fixed update cycle. The parsing process must support on-demand triggering and incremental updates to avoid repeated parsing of full datasets. Documents include tables with coordinates, official documents, and scanned files, so multi-format parsing and high-precision OCR are required to ensure complete extraction of key information such as red line ranges and approval conditions. The party building learning Q&A scenario requires that chunking preserves the contextual integrity of approval processes and compliance clauses, to avoid splitting that breaks logical connections. It also needs to adapt to typesetting differences across documents from different sources, to ensure that relevant content can be accurately matched during Q&A calls.
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 | This category of documents contains tables and coordinate information. This range preserves the logical integrity of single-block content and meets the precise call requirements of party building learning Q&A scenarios |
chunkSplitter | Split by paragraphs and headings | Land use pre-examination documents are mostly divided into chapters by approval links and compliance clauses. Splitting by paragraphs and headings avoids forced splitting of cross-link content |
customPdfParse | Enable high-precision OCR mode | This category of documents includes scanned approval materials. High-precision OCR can accurately identify key content such as coordinates and tables, meeting the information extraction needs of party building learning Q&A scenarios |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Some large-scale planning drawing documents take a long time to parse. This duration covers parsing needs for most scenarios and prevents parsing failures due to timeout |
UPLOAD_FILE_MAX_SIZE | 1000 MB | This category includes large planning drawing files. This upper limit meets conventional document upload requirements while preventing excessive files from occupying too many parsing resources |
chunkSettingMode | Manual configuration | The party building learning Q&A scenario requires adjusting chunking rules for core content such as approval processes and compliance clauses. Manual configuration allows flexible adaptation to business needs |
Three common mistakes
- Symptom: After uploading a scanned planning red line drawing, the interface displays an OCR Error alert. Cause: The high-precision OCR mode of
customPdfParseis not enabled, or the OCR engine configuration is not adapted to complex typeset documents with coordinates, resulting in failure to identify key information. - Symptom: After updating the FastGPT version, the configuration entry for the document parsing function cannot be found. Cause: The configuration entry location has been adjusted during version iteration, and the entry was not searched in the advanced settings of the corresponding module, resulting in failure to enable the parsing function.
- Symptom: When uploading a Construction Land Planning Permit in simple mode, the automatic parsing process is not triggered in some scenarios. Cause: The parsing trigger conditions are not clearly configured, or the default configuration of the selected large model does not include file parsing logic, resulting in the large model not determining that document parsing is required.
How to confirm that the configuration is complete
- Upload a typical scanned Land Use Pre-examination and Site Selection Opinion, check whether the parsed text content includes complete core fields such as project number and land use scope, and confirm that the OCR and chunking logic meet the configuration requirements.
- Enter the parsing configuration module, check whether the values of configuration items such as
customPdfParseandchunkSizematch the preset plan, and confirm that the configuration has taken effect. - Upload a single document whose size is within the
UPLOAD_FILE_MAX_SIZElimit, wait for the duration specified byPARSE_FILE_TIMEOUT_SECONDS, and check whether the parsing task status shows as successful. - For test questions related to party building learning Q&A, such as "What are the compliance requirements for land use pre-examination of a certain project", check whether the returned result references the parsed document content, and confirm that the chunked content can be correctly called.
The values provided here 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.