Reference only; verify against the target FastGPT release and deployment documentation.
What This Category of Data Looks Like
Data for document and material drafting in shipbuilding mainly comes from technical manuals accompanying ship design drawings, section construction inspection reports, classification society inspection reports, and bill of materials (BOM). Data updates follow construction milestones. Corresponding documents are updated weekly during the section construction phase. Final version updates are completed collectively before full ship delivery. Documents mostly have multi-chapter nested structures, including long technical paragraphs and nested tables. Fields include main dimensions (meters, tons), material grades such as EH36 steel, welding process parameters and other professional content. Supported formats include PDF, vector PDF exported from CAD, Excel spreadsheets and more.
What Constraints Do These Characteristics Impose on the Document Parsing and Chunking Step
The multi-format, nested structure and professional field characteristics of shipbuilding documents require the parsing link to adapt to different file formats. It must avoid damaging the integrity of technical paragraphs and tables. Long technical paragraphs and nested clauses must retain complete semantics. Professional parameters must not be split during chunking. Multi-format documents must be compatible with vector PDFs exported from CAD and Excel spreadsheets. This ensures accurate extraction of fields and units. Documents updated by construction milestones must support incremental parsing and batch processing. This avoids repeated parsing of full content.
How to Set the Configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Shipbuilding documents often contain long technical paragraphs. This range can fully retain descriptions of professional parameters such as welding processes and main dimensions |
chunkSplitter | Paragraph-based chunking + model recognition | Shipbuilding documents have nested technical clauses. Model recognition avoids splitting cross-page professional paragraphs |
customPdfParse | Enabled | Adapts to vector-layer PDFs exported from CAD, and fully extracts nested tables and formula formats |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Single full ship construction documents have large file sizes. Extending the parsing timeout allows complete parsing of all content |
chunkSettingMode | Manual configuration | Adjust chunking parameters for different document types such as BOM tables and inspection reports, to adapt to multi-format data for shipbuilding |
Three Common Mistakes
- Only some material entries appear in the ship BOM of the chunking result, with remaining fields empty. This occurs because
customPdfParseis not enabled. The default parsing cannot fully extract all content from nested tables. - A 504 timeout error occurs when calling the document parsing API. This is because
PARSE_FILE_TIMEOUT_SECONDSis not set correctly, or the configured value does not cover the parsing duration of a single large document. - When searching a CSV-format bill of materials, only about 30 entries are returned instead of all 50. This happens because the default chunking splits table rows into independent chunks. This breaks the complete parameter information of a single material entry, leading to missing associated content during retrieval.
How to Confirm the Configuration Is Correct
- Upload a PDF of a ship section construction inspection report. Check the parsed chunked content, and confirm that welding process paragraphs are not split across pages.
- Upload an Excel-format BOM table. Verify that the parsed result displays all material entries and corresponding parameters completely.
- Check the system logs of the parsing task. Confirm that the
customPdfParseconfiguration is active, and there are no prompt messages for failed format parsing. - Test the multi-file upload API. Confirm that the parameter transfer format meets requirements, and there are no cases of file parsing interruption.
The values provided here are common starting points, and should be evaluated against the user'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.