What This Category of Data Looks Like
Normally, daily documents for furniture and wood product manufacturing include order disassembly drawings, cutting lists, edge banding and drilling programs, coating processes, hardware lists, trial assembly and packaging records. Data typically comes from order design systems, sheet and wood batch records, CNC cutting and drilling equipment, ERP/MES systems, and moisture content testing equipment. Update frequency aligns with orders, daily cycles, or material batches. Documents include fields such as panel dimensions, wood moisture content, hardware model numbers, production batch numbers, with units including millimeters, percentages, piece counts, and others.
Constraints Imposed on Document Parsing and Chunking
Differences in structure across document types create parsing constraints. For example, CNC program PDFs contain embedded parameter text and graphics, order disassembly drawings include accompanying text descriptions, and hardware lists are structured tables. Frequently updated order-level documents require parsing workflows to support fast batch processing, to avoid delaying production cycles due to excessive single-document parsing time. The large number of panels and detailed field breakdowns mean chunks must not be too small, causing redundant retrieval, or too large, causing key parameters to be obscured. Documents from disparate systems have inconsistent formats, so the parsing process must support extraction logic for multiple file types.
How to Set Configurations
The following configurations are based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the typical length of panel parameters and process steps in furniture craft documentation, preventing single chunks from containing semantically unrelated information |
chunkSplitter | Split by paragraph + order number | Adapts to the structure of documents driven by custom orders, preventing chunks from splicing panel information across multiple orders |
customPdfParse | Enable text layer first parsing | Adapts to embedded parameter text in CNC program and coating process PDFs, preventing loss of key values caused by only extracting the graphic layer |
PARSE_FILE_TIMEOUT_SECONDS | 120 seconds | Prevents parsing timeout interruptions when processing coating process documents over 100 pages or bulk cutting lists |
UPLOAD_FILE_MAX_SIZE | 500 MB | Supports bulk upload requirements for full batch sheet batch records, preventing upload failures for large documents |
Three Common Mistakes
- A
net::ERR_INCOMPerror occurs when uploading cutting lists over 100 MB. This happens because theUPLOAD_FILE_MAX_SIZEconfiguration value is smaller than the actual size of a single document, causing file transfer to be interrupted. - Chunk results contain panel information across multiple orders. This happens because
chunkSplitteris not configured to split by order numbers within documents, causing semantic chunks to splice information across orders. - Hole position parameters extracted when parsing CNC program PDFs are empty. This happens because
customPdfParseis not enabled, so only the image layer of the PDF is parsed instead of extracting embedded text parameters.
How to Confirm Configurations Are Set Correctly
- Upload a typical cutting list document, and check if the number of parsed chunks matches the number of panel batches marked in the document.
- Check the
PARSE_FILE_TIMEOUT_SECONDSconfiguration, upload a 100-page coating process document, and confirm the parsing status does not show a timeout. - After enabling
customPdfParse, upload a CNC program PDF, and verify that extracted parameters such as hole positions and feed rates are complete. - Upload a single sheet batch record under 500 MB, and confirm there are no interrupted errors during upload.
The values provided are common starting points and should be measured against the reader's own samples.