What this category of data includes
Data for this category comes from construction project approval management systems, construction drawing review systems, enterprise qualification databases, and credit platforms. Updates occur daily. Commonly processed documents include construction permit applications, construction drawing review certificates, quality and safety supervision procedures, and completion acceptance record forms. Most documents are a mix of structured forms and compliance files. They contain fields such as participating construction unit qualifications, personnel information, and approval nodes. Some files include stamped scans and electronic signatures, with no unified fixed format.
Constraints on Document Parsing and Chunking
The mix of structured forms and unstructured compliance files requires chunking to balance accurate field extraction and retention of paragraph logic. Splitting that breaks relational links between approval materials must be avoided. The daily updated data volume requires the parsing process to have stable batch processing capabilities and reduce abnormal interruptions. The need to accurately identify core fields such as participating construction unit qualifications and personnel information limits the minimum chunk granularity. This prevents key information from being split apart. Large-volume construction drawing review files are prone to content shift. This puts higher requirements on upload and parsing stability.
How to Configure Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the length of form paragraphs and compliance descriptions in construction approval documents, avoids splitting key approval node descriptions |
chunkSplitter | Split by paragraph + title | Matches the multi-title, multi-paragraph structure of approval documents, retains the logical hierarchy of materials |
customPdfParse | Enabled | Adapts to construction drawing review files with electronic signatures and scanned copies, improves parsing accuracy for non-standard PDFs |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Covers the storage volume requirements of complete completion acceptance record files, adapts to large-file upload scenarios |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Reserves sufficient parsing time for large-volume approval documents, avoids parsing failures due to timeout |
Reference only; verify against the target FastGPT release and deployment documentation.
Three Common Configuration Mistakes
- Uploading a large-volume PDF file to the knowledge base triggers an "offset out of range" error when progress reaches 90%. This occurs because the
UPLOAD_FILE_MAX_SIZEandPARSE_FILE_TIMEOUT_SECONDSconfigurations are not adjusted, leading to file volume exceeding limits or insufficient parsing time. - The same file uploaded via the create file collection API and directly via the platform produces inconsistent chunking results. This happens because the
chunkSettingModeand chunking parameter configurations are not synchronized during API calls, resulting in inconsistent chunking rules. - After configuring a custom URL document parsing tool, no data is returned. This is due to failure to verify that the custom tool's interface return format matches the parsing result structure required by FastGPT, or that the interface does not correctly return document text content.
How to Confirm Configurations Are Correctly Set
Uploading a single completion acceptance record file larger than 100 MB enables verification that parsing progress completes without errors. Adjust UPLOAD_FILE_MAX_SIZE and PARSE_FILE_TIMEOUT_SECONDS to values matching the file volume. Extracting the participating construction unit field from a construction permit application document allows checking that chunking results retain complete field information. Adjust chunkSize and chunkSplitter configurations to match the document structure. Comparing the same file uploaded via the create file collection API and directly via the platform confirms consistent chunking results. Synchronize the chunkSettingMode and chunking parameter configurations. Calling the custom URL parsing tool allows verification that the interface returned text content matches the original document. Confirm the configured custom tool functions correctly.
The values provided here are common starting points and should be evaluated against available 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.