What data in this category looks like
This category of data comes from PLM, ERP, MES, supplier materials, validation tests, and customer-specific requirements. Updates occur irregularly, triggered by each process or engineering change. Daily processed documents include ECN/ECR forms, change assessment reports, drawings and BOM version files, process validation reports, production switching plans, inventory disposal documents, and customer approval documents. Document fields include version numbers, material codes, process parameters, implementation standard numbers, etc. Some drawing documents include engineering unit information such as dimensions and tolerances.
What constraints these characteristics impose on document parsing and chunking
Documents in this category include multiple types such as structured forms, engineering drawings, and compliance documents, with scattered format sources including PDF, CAD exported documents, and Excel forms. There is no fixed unified template for the document structure of each change, so parsing must adapt to differentiated field layouts. Some documents include engineering units and precise parameters; chunking must retain the associated integrity of fields and parameters, avoiding splitting key information across fields. Approval identifiers in compliance documents must be fully retained and not truncated by chunking, while also supporting metadata extraction from multiple sources.
How to set the configuration
This configuration is based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | This category of documents includes structured fields and engineering parameters. This range can retain the integrity of information associated with a single change, avoiding splitting content across fields |
chunkSplitter | Split by paragraph + heading | Adapts to the hierarchical heading structure of documents such as ECNs and validation reports, retaining the logical association between change items and associated content |
customPdfParse | Enabled | Compatible with layer-enabled PDF documents exported from CAD, with higher accuracy when extracting drawing annotations and text content |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Large BOM or multi-drawing combined documents take longer to parse; this duration covers the parsing needs of most conventional change documents |
chunkSettingMode | Manual configuration | Allows adjustment of chunking parameters for different types of change documents (such as form-based, drawing-based) to adapt to the needs of multi-structure documents in this category |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Change documents in this category often include multiple drawings and BOMs; this size covers the upload needs of most conventional change documents |
Three common errors
- Phenomenon: Long delay with no response after calling the file parsing interface, eventually returning a timeout error. Cause: The
PARSE_FILE_TIMEOUT_SECONDSparameter was not adjusted; the default duration is insufficient for change documents that include multiple drawings. - Phenomenon: Parsing fails after enabling enhanced file parsing, and normalcy is restored after switching to the old version tool or Marker. Cause:
customPdfParsewas not configured to adapt to CAD-exported PDF formats, or an incompatible parsing engine version was used. - Phenomenon: The file parsing tool cannot be invoked during conversation, and uploaded change documents cannot be chunked. Cause:
chunkSettingModeis set to automatic mode, the parsing switch was not manually enabled, or the uploaded file size exceeds theUPLOAD_FILE_MAX_SIZElimit.
How to confirm the configuration is correct
- Upload a typical ECN document, view the parsed chunking results, and confirm that headings and corresponding change content are not split.
- Upload a PDF document including CAD-exported drawings, check whether the parsed results include the dimension and tolerance information annotated on the drawings.
- Upload a single change document that complies with the
UPLOAD_FILE_MAX_SIZElimit, and confirm that no timeout error is triggered during the parsing process. - Check the
customPdfParseswitch status on the configuration page, and confirm that it matches the document format currently being processed.
The values provided are common starting points and should be measured against the reader's own samples.