What this category of data looks like
This category of data comes from state-owned asset supervision information systems, enterprise asset management systems, property right exchange systems, valuation agency reports, and financial statements. The update frequency is monthly. Commonly processed documents include property right registration forms, asset evaluation reports, property right transfer plans, on-site transaction announcements, and asset disposal approval forms. These documents include both structured form files and long-text reports spanning dozens of pages. Some documents have custom page headers, encrypted forms, or non-standard layouts. Core business fields include ownership numbers, evaluation values, transaction base prices, and other key business information.
What constraints these characteristics impose on the "document parsing and chunking" link
Monthly batch updates include a large number of documents, and individual document volumes vary widely. Some documents are tens of KB in size, while others are long-text evaluation reports of dozens of MB. This places high requirements on parsing stability. Documents from multiple sources have inconsistent formats. Some PDF documents have encrypted forms or custom page headers, and conventional parsing may lose core fields. Core business fields must be fully retained within chunks to avoid splitting key logic such as ownership and transaction during retrieval. Batch uploads must adapt to different file size limits to avoid timeouts or parsing failures, which would otherwise affect subsequent process verification and audit work.
How to set configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
Chunk Size (chunkSize) | 800–1200 characters | This category of documents contains fixed form fields and long-text descriptions. This range balances context completeness and retrieval accuracy, avoiding overly long chunks that reduce recall or overly short chunks that disrupt business logic |
Chunk Splitter (chunkSplitter) | Split by fixed chapter titles of business documents | Property and asset documents often have fixed chapter titles such as "Evaluation Description", "Transfer Plan", and "Approval Opinion". Splitting by these titles preserves the integrity of business logic |
File Parsing Timeout (PARSE_FILE_TIMEOUT_SECONDS) | 1200 seconds | Large asset evaluation reports take a long time to parse. The default duration is insufficient for complete parsing, and this setting covers the parsing needs of most individual documents |
Single File Upload Limit (UPLOAD_FILE_MAX_SIZE) | 500 MB | Monthly batch uploads of collections of property right transaction announcements and packaged files of multiple evaluation reports often reach hundreds of MB. This limit adapts to conventional batch upload requirements |
Custom PDF Parsing (customPdfParse) | Enabled | Business PDFs often have encrypted forms and custom page headers and footers. Custom parsing can retain core business fields and layout structure |
Chunk Setting Mode (chunkSettingMode) | Manual Mode | Separate chunk rules need to be adjusted for form documents and long-text reports. Manual mode allows flexible adaptation to different document types |
Three common errors
- Uploading an asset evaluation report larger than 10 MB triggers the error
timeout of 360000ms exceeded. This happens because thePARSE_FILE_TIMEOUT_SECONDSparameter was not adjusted, and the default timeout duration is insufficient for parsing long documents. - Uploading an asset disposal approval form in docx format returns an empty parsing result. This happens because
customPdfParsewas not enabled to adapt to non-standard docx page headers, footers, and signature field areas. - Search results after chunking include irrelevant content from property right registration form attachments. This happens because
chunkSplitterwas not set to the fixed chapter separators of business documents, resulting in cross-chapter chunking that disrupts business context.
How to confirm configurations are correct
- Upload a single asset evaluation report of around 10 MB, check for no timeout errors during parsing, and confirm that the
PARSE_FILE_TIMEOUT_SECONDSsetting takes effect. - Upload an asset disposal approval form in docx format, check that the parsing result retains core business fields such as approval opinions and signature areas, and confirm that
customPdfParseis enabled and effective. - Upload a document containing a long evaluation description, check that chunking results do not split a single evaluation item into multiple unrelated chunks, and confirm that
chunkSizeandchunkSplittersettings are reasonable. - Upload packaged monthly update documents in batches, check that the
UPLOAD_FILE_MAX_SIZElimit is not triggered, and verify that the upload limit setting is correct.
The values provided above are common starting points, and should be evaluated against the reader's own samples.
Reference only; verify against the target FastGPT release and deployment documentation.
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.