What data for this category looks like
Daily documents for this category include fixture requirement documents, positioning datum schemes, fixture drawings, gauge operation manuals, acceptance reports, calibration records, and change histories. Data sources cover product digital models and GD&T, process routes, coordinate measuring machine data, on-site trial feedback, and metrology systems. Updates run per project milestones, each product change, and regular calibration cycles. Document structures mix text descriptions and professional annotations, with fields including GD&T tolerance zones, clamping force parameters, calibration cycles, and coordinate measuring machine measurement values. Units include millimeters, newtons, days, and other standard engineering units.
What constraints these characteristics impose on document parsing and chunking
Documents for this category mix text descriptions and professional annotations. Parsing must fully retain explanatory content for specialized fields such as GD&T to avoid loss of critical technical information. Cross-project change histories and fixture drawings must link to their corresponding versions. Chunking must not split positioning datum schemes that span multiple pages. Frequently updated calibration records and acceptance reports must split along time-based logic, so that corresponding batch detection data can be quickly located during Q&A. Large combined documents containing multiple drawings and records require parsing and chunking workflows adapted to large file processing needs.
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 documents mixing text and professional annotations for this category, avoids splitting multi-page positioning datum schemes |
chunkSplitter | Split by paragraph + filter headers and footers | Filters repeated project number headers from fixture documents, splits by natural paragraphs to retain integrity of specialized fields |
customPdfParse | Enabled | Extracts GD&T annotation text descriptions from PDF fixture drawings, avoids loss of critical technical information |
PARSE_FILE_TIMEOUT_SECONDS | 300–600 seconds | Adapts to large fixture documents containing multi-page coordinate measuring machine data, ensures complete parsing |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Supports uploading combined sets of multiple fixture documents per project, meets batch file processing requirements |
chunkSettingMode | Manual configuration | Adjusts chunking parameters for different document types such as calibration records and fixture drawings, adapts to specialized content retention needs |
Three common mistakes
- Symptom: After uploading a fixture drawing PDF, the parsing result lacks text descriptions of GD&T tolerance zones. Cause: The
customPdfParseconfiguration is not enabled, and text extraction for specialized drawings is not turned on. - Symptom: Multi-page positioning datum schemes are split into multiple independent chunks after chunking, making it impossible to associate complete datum logic during Q&A. Cause: The default general chunking strategy is used, and
chunkSizeand chunking rules are not adjusted for this category’s documents. - Symptom: Q&A results received from the front end are unsegmented full text, with line breaks and formatting of specialized fields not retained. Cause:
chunkSplitteris not configured to split by natural paragraphs, and Markdown format output is not enabled.
How to confirm configurations are correctly set
- Upload a test document containing fixture drawings and calibration records, check that the parsing result includes text descriptions of GD&T annotations, and confirm that
customPdfParseis enabled. - Review chunked text blocks, confirm that multi-page positioning datum schemes are not split, and verify that
chunkSizevalues match the configured settings. - Call the API to retrieve the text stream, check that segmented text and headings in Markdown format are included, and confirm that chunking rules are active.
- Upload multiple fixture documents from the same project, check that parallel processing for knowledge base creation operates normally, and confirm that parallel configurations meet requirements.
The values provided here are common starting points, and should be evaluated against your own samples.