What This Category’s Data Looks Like
Data for this category comes from daily business outputs of motor manufacturing winding, insulation, final assembly, and testing departments. It includes winding process cards, insulation treatment records, dynamic balance reports, voltage-resistant insulation test reports, and type and factory inspection test reports. Data sources cover electromagnetic design data, copper wire and insulating material batches, winding equipment, balancing machines, electrical test benches, and ERP/MES systems. Update frequency follows each unit/batch production run or each type change. Data fields include process parameters, test values, batch information, equipment numbers, and more, with clear unit markings such as turns, megohm-level insulation resistance values, vibration amplitude, and similar metrics.
Constraints for Document Parsing and Chunking
Documents including winding process cards and test reports contain mostly process and test parameters with attached units. Batch information and equipment numbers must be linked accurately, so chunking must preserve field associations and avoid splitting critical parameter groups. Document formats vary widely, including structured tables and plain text paragraphs. Parsing must support multiple formats. Documents for each unit or batch are updated frequently and tied to specific batches. Parsed results must retain the binding between batches and documents to prevent loss of traceability information after chunking. Test data such as voltage resistance and temperature rise has strict precision requirements. Chunking must ensure key values are not split across multiple chunks, which breaks contextual integrity for subsequent quality analysis.
Configuration Settings
The following configurations follow official rules for FastGPT v4.17.0:
| Configuration Item | Recommended Approach/Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the length of parameter groups with units in winding process cards and test reports, prevents splitting critical process or test data chunks |
chunkSplitter | Split by paragraph + table header | Adapts to mixed document formats of structured tables and plain text, preserves the link between process parameters and their descriptions |
customPdfParse | Enabled | Accurately extracts tables and fields from PDF test reports, avoids text misalignment |
PARSE_FILE_TIMEOUT_SECONDS | 120 seconds | Matches the parsing duration of large single type test reports, prevents timeout interruptions |
chunkSettingMode | Manual mode | Adjusts chunking rules individually for different document types (process cards, test reports) to ensure batch information is bound to associated parameters |
UPLOAD_FILE_MAX_SIZE | 100 MB | Matches upload limits for large single type test reports, prevents upload failure before parsing starts |
Three Common Errors
- Unrendered Markdown tags appear in knowledge base chunk displays. The cause is that
customPdfParseis not enabled, leading to abnormal structured table parsing and retention of original typesetting tags. - Document content fails to parse when passing in a link. The cause is that a reasonable value for
PARSE_FILE_TIMEOUT_SECONDSis not configured, or the linked document exceeds theUPLOAD_FILE_MAX_SIZElimit. - Chunk IDs cannot be copied from the chunk list. The cause is that the front-end rendering logic does not bind a copy event for chunk IDs, and no ID display optimization is implemented for batch-exported chunk data.
How to Verify Correct Configuration
- Upload a winding process card PDF, check that parsed chunks retain complete process parameter groups without splitting critical values and their attached units.
- Upload a type test report containing tables, confirm that parsed text has no typesetting misalignment or residual original Markdown tags.
- Pass in an external document link, verify that the parsing task completes within the preset timeout period without timeout interruptions.
- View the chunk list, confirm that the ID of each chunk can be selected and copied, with no display abnormalities.
The values listed here are common starting points. They should be validated against the samples relevant to the intended use.