What the data for this category looks like
Coal transportation and sales declaration materials and file review data comes from electronic and paper documents generated across the entire business process. These include purchase and sales approval forms, weighing settlement slips, transportation permission files, monthly transportation and sales ledgers, and more. Data updates align with individual transportation and sales business or periodic business progress. Documents are primarily structured tables, mixed with unstructured approval comments and stamped pages. Fields include coal type identifier, transportation tonnage, settlement unit price, transportation route, approval node, and more. Units include tons, yuan, vehicle trips, kilometers, and more. Some documents are scanned copies and require OCR to restore their content.
What constraints these characteristics impose on the document parsing and chunking link
Structured tables make up a large share of the documents and have dedicated units. Parsing steps must retain cell hierarchy and field correspondence. During chunking, fields, values, and units must be bound together to avoid semantic breaks. Some documents are scanned copies, so OCR parsing must be adapted to restore complete content. Otherwise, fields may be missing or misaligned. In bulk declaration scenarios, the number of documents per batch is high. A single file may contain multi-page cross-page tables. Parsing duration and upload limits must be controlled to prevent process blockages. Documents with mixed formats must support parsing logic for both structured tables and unstructured text, to avoid splitting cross-page business data.
How to set up the configuration
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Coal transportation and sales documents are mostly structured tables with fields and associated approval comments. This range retains the complete semantics of a single business entry, avoiding splitting content across fields |
chunkSplitter | Split by paragraph and table boundaries | Adapts to mixed structured tables and unstructured approval comments in documents, retaining the overall table structure and paragraph logic |
customPdfParse | Enable OCR enhancement mode | Some weighing slips and transportation files are scanned copies. This configuration restores structured fields from scanned documents |
UPLOAD_FILE_MAX_SIZE | 500 MB | A single coal transportation and sales file may contain multiple batches of weighing slips and ledgers. This limit covers document sizes for most business scenarios |
PARSE_FILE_TIMEOUT_SECONDS | 120–180 seconds | Parsing large or multi-page documents takes longer. This range prevents timeout interruptions |
chunkSettingMode | Manual configuration | Adjust chunking rules for the structured characteristics of coal transportation and sales documents, avoiding automatic chunking that breaks business semantic associations |
Three Common Errors
- Phenomenon: After uploading a PDF or scanned copy, the knowledge base parsing result is empty or has missing fields. Cause:
customPdfParseis not configured to enable OCR enhancement mode, so structured content in scanned copies or non-standard PDF files cannot be recognized. - Phenomenon: Calling the parsing interface returns a timeout error with status code 504. Cause:
PARSE_FILE_TIMEOUT_SECONDSis not adjusted to the range suitable for large-volume files, causing long document parsing to time out. - Phenomenon: After uploading a document, chunks split cross-page monthly transportation and sales ledgers, causing single business data to be scattered across multiple chunks. Cause:
chunkSizeis set too small, orchunkSplitteris not configured to split by table boundaries, destroying the integrity of structured business data.
How to Confirm the Configuration Is Correct
- Upload a single coal transportation and sales file that includes structured tables and scanned content. Check if the parsing result retains the complete cell structure of the table and corresponding fields.
- Call the parsing interface to upload a document with more than 100 pages. Confirm the interface does not return a timeout error. Check if the
PARSE_FILE_TIMEOUT_SECONDSconfiguration covers the parsing duration. - Check the size of the uploaded document. Confirm it does not exceed the
UPLOAD_FILE_MAX_SIZEsetting, to avoid upload interception. - Adjust
chunkSizeto different ranges. Upload the same document and compare chunk results. Confirm chunks do not split the complete semantics of a single business entry.
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.