What the data for this category looks like
Data for this category comes from the construction project approval management system, construction drawing review system, enterprise qualification database, and credit platform. The update frequency is daily. Commonly processed documents include construction permit application materials, construction drawing review qualification certificates, quality and safety supervision procedures, and completion acceptance record forms. Most documents use fixed form formats, with some attached high-definition drawing attachments. Fields include project unique identifier, participating unit name, approval process records, filing number, etc. There are no complex custom unit fields.
What constraints these characteristics impose on the document parsing and chunking link
Data sources are scattered and updated daily. Parsing timeliness must be ensured to prevent expired data from being used. Fixed form documents have a structured layout. Some PDFs with attached drawings may contain non-text elements. Multi-type document parsing support is required. Fields such as participating unit names and approval nodes are core business fields. Key information must not be truncated during chunking. Filing documents such as completion acceptance record forms must fully retain filing numbers and approval conclusions. Chunk boundaries must avoid splitting core business fields. Format differences across different documents require parsing configurations to cover both form and attachment scenarios. Field omissions or parsing errors must be prevented.
How to set the configurations
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the long-text structure of engineering documents, preventing core approval fields from being split |
chunkSplitter | Split by paragraph + heading | Matches the structured layout of fixed forms, retaining contextual association of approval processes |
customPdfParse | Enable OCR enhancement mode | Processes PDF documents with attached construction drawings, resolving parsing issues for non-text elements |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Adapts to scenarios with large construction drawing PDFs and batch upload of multiple materials |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Matches the parsing time required for large documents, preventing timeout interruptions |
chunkSettingMode | Manual configuration mode | For fixed formats of engineering documents, customize chunking rules without relying on general templates |
Three Common Misconfigurations
- Phenomenon: A knowledge base platform fails to sync PPT and PDF documents. A sync failure message is displayed on the interface. Cause: Non-text file support for the
customPdfParseconfiguration is not enabled. The default parsing rule only supports general text format documents. - Phenomenon: After deploying the latest v2 version of marker, logs show an OCR error. Parsing tasks are interrupted. Cause: The OCR enhancement mode is not configured, or the uploaded PDF contains encrypted layers leading to OCR recognition failure.
- Phenomenon: Claude-series models throw errors after connecting to file parsing. Normal chat flows work without issues. Cause: The
chunkSizeconfiguration does not meet the document length requirements, leading to context exceeding the adaptive range during model calls.
How to confirm configurations are set correctly
- A single large construction drawing PDF can be uploaded. After parsing completes, system logs can be checked to confirm there are no timeout errors. This verifies the rationality of the
PARSE_FILE_TIMEOUT_SECONDSconfiguration. - A construction permit application document can be imported. Chunking results can be checked to confirm full retention of core fields such as project number and participating unit names. This verifies the adaptability of the
chunkSizeandchunkSplitterconfigurations. - PPT-format approval materials can be tested for upload. It can be confirmed that they sync to the knowledge base normally. This verifies that the
customPdfParseconfiguration is enabled. - The total size of uploaded files can be checked. It can be confirmed that sizes do not exceed the
UPLOAD_FILE_MAX_SIZElimit. This verifies the matching of upload configurations.
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.