What the data for this category looks like
Policy and regulatory basis data for building construction general contracting comes primarily from official public documents issued by housing and urban-rural development authorities, provincial and municipal housing regulatory departments, and industry associations. Updates are released irregularly alongside special policy announcements. Most documents are official PDF documents. Some supporting attachments are in Excel or CSV format. Documents include fields such as document number, effective date, clause number, applicable scenario, and pricing and measurement standards. Some files include specific engineering pricing values and their corresponding units.
Constraints imposed on document parsing and chunking
Official documents often include headers and footers with issuing authorities and document numbers. These elements interfere with the accuracy of plain text extraction, so redundant information must be removed in a targeted manner. Supporting attachments often contain multi-column pricing and project data. Cross-column associated parsing is required to avoid losing data relationships after chunking. Single policy documents are lengthy, with strict logical hierarchies between clauses. Chunking must retain clause context to avoid splitting complete policy clauses into meaningless fragments. Some files include specialized units for engineering measurement. Units must be bound to their corresponding clauses after chunking to avoid unit confusion during Q&A.
Configuration settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Single clause length for housing construction policy documents mostly falls within this range. This retains clause integrity and avoids excessive splitting. |
chunkSplitter | Split by paragraph | Policy documents use paragraphs as logical expression units. Splitting by paragraph ensures each chunk contains a complete policy statement. |
customPdfParse | Enable header and footer filtering | Official documents often have redundant headers and footers. Enabling this extracts clean main text content. |
UPLOAD_FILE_MAX_SIZE | 100 MB | Single housing construction policy collection files typically do not exceed this size. This adapts to large documents for batch import. |
PARSE_FILE_TIMEOUT_SECONDS | 300 seconds | Parsing large PDF or Excel attachments takes extended time. This duration covers the parsing process for most files. |
chunkSettingMode | Manual configuration | Chunking rules must be adjusted for the special format of housing construction policies. Manual configuration allows flexible adaptation to document characteristics. |
Three common configuration mistakes
- Phenomenon: No content displays in the knowledge base after uploading an Excel file, or only two columns of data are retained after importing a CSV file. Cause: Parsing configuration for multi-column attachments is not enabled. The default CSV parsing only recognizes two columns of data, which does not adapt to the multi-column pricing fields in housing construction policy attachments.
- Phenomenon: When calling streaming return results, it is not possible to obtain both knowledge base call parameters and final return content at the same time. Cause: The parameter binding logic for external calls is not correctly configured, and knowledge base retrieval parameters are not associated and parsed with streaming return results.
- Phenomenon: A connection failure error appears after deploying marker-pdf, and PDF files cannot be parsed. Cause: Ports are not correctly mapped during Docker deployment, or the locally deployed parsing service is not started, preventing FastGPT from calling the parsing interface.
How to confirm configuration is correct
- Upload a PDF file of a housing construction policy, and check if the parsed text has removed redundant header and footer information.
- Import a CSV or Excel attachment with multiple columns of data, and verify that all fields are retained in the imported data.
- Initiate a policy Q&A request, and confirm that the returned result matches the original answer content stored in the knowledge base.
- Test the upload and parsing of large files, and confirm that no timeout error occurs during the parsing process.
The values provided 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.