What Data for This Category Looks Like
Typical entities in this category are panel furniture, solid wood furniture, custom home and door enterprises. Daily generated documents include order breakdown drawings, cutting lists, edge banding and drilling programs, coating process documents, hardware lists, assembly and packaging records. Data sources cover order design systems, sheet and wood batch ledgers, CNC processing equipment, ERP/MES systems, and moisture content testing equipment. Update frequency changes with orders, daily cycles, or individual material batches. Document structures vary widely: some are cutting Excel files with multiple sheets, some are scanned breakdown PDFs, and some are plain text process documents. Fields include panel dimensions, sheet thickness, hardware model numbers, moisture content values, and dedicated units such as millimeters, cubic meters, and percentage.
Constraints Imposed on Document Parsing and Chunking
The document characteristics of this category impose multiple constraints on the parsing and chunking process. Multi-sheet cutting Excel lists require accurate identification of the business attribute of each sheet to avoid mixing up cutting data for different panels. Scanned breakdown drawings and process documents require OCR recognition, otherwise text content cannot be extracted. Documents driven by custom orders with multiple panels have a large per-file volume, so chunking parameters must be set reasonably to avoid content fragmentation. Document formats vary significantly across different sources, so parsing rules for multiple file types must be adapted. Fields such as panel dimensions and moisture content have high precision requirements; parsing errors will directly affect subsequent production links.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the feature that single documents in this category contain multiple panel data, avoiding semantic breaks caused by over-splitting or overly long single chunks |
chunkSplitter | Switch based on document type | Enable OCR chunking for scanned PDFs, split Excel documents by sheet, split text documents by paragraph, to adapt to the diverse document formats in this category |
customPdfParse | Enabled | For scanned breakdown drawings and process documents, enable custom PDF parsing to ensure OCR recognition accuracy, adapting to the scenario where a large number of non-editable PDFs exist in this category |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Adapts to the large per-file volume of individual cutting lists or breakdown drawings, avoiding task interruptions due to parsing timeouts |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Allows uploading documents for single batches of sheet ledgers or combined multiple orders, matching the demand for batch document uploads in this category |
This page's configurations were tested on FastGPT v4.17.0
Three Common Configuration Mistakes
- After uploading a scanned breakdown PDF, the parsing result shows success but no extracted text. The cause is that the OCR function corresponding to
customPdfParseis not enabled, and only editable text PDFs are processed. - When parsing an Excel cutting list, only the data from the first sheet is extracted, and the content of other sheets is not recognized. The cause is that the chunking rule for splitting by sheet is not configured, and the multi-sheet structure is not recognized by default.
- After uploading a mixed-text-and-image process document, only text without the original image is displayed in the knowledge base. The cause is that the parsing configuration to retain original images is not enabled, and only OCR-processed text content is extracted.
How to Verify Correct Configuration
- Upload a scanned breakdown PDF, check if the parsed result contains recognizable panel dimension text, to confirm that the OCR function is active.
- Upload a cutting Excel file with multiple sheets, check if the parsed chunks are split separately for each sheet, and confirm that sheet names are correctly identified.
- Upload a process document with images, check if both text content and original images exist in the knowledge base, to confirm that the configuration to retain original images is active.
- Upload a large-volume cutting list, check if the parsing task completes within 600 seconds without timeout errors.
The configuration values provided here are common starting points. They should be evaluated against the user's own specific samples.