What this category’s data looks like
The typical data sources for this category are business systems of natural resource departments, state-owned assets supervision institutions, property right trading institutions, and public resource trading centers. Daily documents include transfer announcements, trading notices, bid application documents, margin vouchers, transaction confirmation letters, listing result announcements, property right transfer specifications, and more. Data sources cover the national public resource trading platform, natural resource trading systems, property right exchange systems, and state-owned assets supervision systems. Data is updated daily. Most documents contain structured fields such as ownership descriptions, target details, and amount vouchers. Some documents have nested tables and long paragraph descriptions.
What constraints do these characteristics impose on the document parsing and chunking step
Documents from multiple systems have varying formats. The parsing module must support multiple export formats including PDF and Word. Business data is updated daily, requiring a stable and efficient parsing process to avoid delays that impact subsequent review steps. Nested tables in documents, such as target lists in property right transfer specifications, and long paragraph ownership descriptions may lose key information or cause cross-content mixing if chunked improperly. Structured fields such as margin amounts and target numbers must stay intact within chunks. Otherwise, this will lower the matching accuracy of subsequent file review. Batch-submitted bid application materials have large file sizes. The system must support large file upload and parsing.
How to configure the settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | This category’s documents include long paragraph ownership descriptions and nested tables. This range balances information integrity within chunks and retrieval precision, avoiding cross-application or cross-target content mixing |
chunkSplitter | Split by title, paragraph, and table boundaries | Adapts to the hierarchical structure of this category’s documents, preserving the title hierarchy of transfer announcements and transaction confirmation letters, as well as the row and column structure of tables |
customPdfParse | Enabled | Most documents in this category contain complex nested tables. The default parsing logic may lose table structure; enabling this option preserves complete table layout information |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Parsing large property right transfer specifications or batch bid application materials takes a long time. This duration prevents task failure due to mid-process timeout |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Supports upload of large files such as batch bid application materials and annual transaction archive packages, adapting to the batch submission needs of daily business in this category |
chunkSettingMode | Manual mode | Allows separate adjustment of chunking rules for different document types, such as transfer announcements and transaction confirmation letters, adapting to the diverse document formats of this category |
Three common misconfigurations
- Parsing tasks time out and return a 504 status code after uploading a large property right transfer specification: The
PARSE_FILE_TIMEOUT_SECONDSvalue was not adjusted to match this category’s documents. The default timeout setting is insufficient to complete parsing of large files. - Table structures are not preserved when uploading PDFs to the knowledge base, and complex target details are split into scattered text:
customPdfParsewas not enabled. The default parsing logic cannot identify the row and column relationships of nested tables, leading to lost or disorganized table content. - Margin amount fields from multiple bid applications are merged into the same chunk:
chunkSizewas set too large, and chunks were not split by application batch. This causes cross-application structured field mixing, which affects subsequent review matching.
How to verify correct configuration
- Upload a standard transaction confirmation letter PDF, and check if the parsed text chunks retain the row and column alignment information of the table, to confirm that the
customPdfParseconfiguration is active. - Upload a batch bid application package of around 100 MB, and check if the parsing task completes within 600 seconds without timeout errors, to confirm that the
PARSE_FILE_TIMEOUT_SECONDSconfiguration is correct. - Adjust
chunkSizeto 800 characters, upload a single bid application document, and check if chunks are split by each individual application, with no cross-application content merging. - Upload an archive package close to 1000 MB in size, confirm that the file can be uploaded and parsed normally without size limit errors.
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.