What the Data for This Category Looks Like
Data for this category comes primarily from national public resource trading platforms, local trading platforms, engineering approval systems, and bidding supervision systems. The update frequency is daily. Documents processed daily include bidding announcements, qualification prequalification documents, bidding documents, bid documents, evaluation reports, and successful bid candidate announcements. Document structures include structured units such as section division, qualification requirements, bill of quantities, evaluation criteria, and construction periods. Some documents include tables, formulas, and attachments. Field units involved include quantity units and duration units.
Constraints on Document Parsing and Chunking
The above data characteristics create multiple constraints for the document parsing and chunking process. Daily updated data sources require incremental synchronization rules to avoid re-parsing full datasets. Bidding documents have numerous structured units. Chunking must accurately identify independent modules such as sections and qualification requirements, to prevent splitting core business content. Document formats differ across multiple sources. General parsing rules must be configured to adapt to export formats from different platforms. Some documents contain complex tables and formulas. Specialized parsing capabilities must be enabled to ensure key information is not lost.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the paragraph length of bidding documents, avoids splitting core business units such as sections or qualification requirements |
chunkSplitter | Split by headings and paragraphs | Matches the chapter structure of bidding documents, ensures chunks correspond to independent business modules such as announcements and qualification requirements |
customPdfParse | Enabled | Supports parsing tables, formulas, and structured content in bidding documents, avoids losing key information such as the bill of quantities |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Adapts to the parsing duration of large evaluation reports and bid documents, avoids parsing failure due to timeout |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Meets the upload requirements for large bid documents and batch bidding documents |
chunkSettingMode | Manual configuration | Adjusts chunking rules for the special structure of bidding documents, adapts to different types of business documents |
Three Common Configuration Mistakes
- After uploading a PDF document, the chunk preview shows that the file content cannot be read. The cause is that the
customPdfParseconfiguration is not enabled. The default parsing engine cannot handle complex tables and formulas in bidding documents. - After deploying Marker, the log displays an OCR error. The cause is that the OCR dependency package is not installed, or the
PARSE_FILE_TIMEOUT_SECONDSsetting is too short to complete OCR recognition for long documents. - The local version cannot return complex formulas normally. The cause is that the version is lower than FastGPT v4.17.0. Older versions of the PDF parsing engine do not have optimized structured formula recognition capabilities.
How to Confirm the Configuration Is Correct
- Upload a bidding announcement PDF that includes section division. Check the chunk preview to confirm that chunks do not split core content across sections.
- Upload a bidding document of the maximum allowed size. Check that the system does not trigger a file size limit prompt, confirming that the
UPLOAD_FILE_MAX_SIZEconfiguration takes effect. - Configure a knowledge base synchronization rule, add PPT and PDF file types. After executing the synchronization, check the synchronized file list to confirm that target files have been successfully synchronized.
- Upload an evaluation report that includes quantity formulas. Check whether the chunked content fully retains the formula and table structure, confirming that the
customPdfParseconfiguration takes effect.
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.