What the data for this category looks like
Daily documents for SMT mounting and electronic assembly include BOM and coordinate files, line change sheets, stencil and furnace temperature records, first article confirmation reports, SPI/AOI inspection reports, maintenance records, traceability labels, and more. Data sources include BOM, Gerber files, and coordinate data provided by customers, as well as factory MES system data, solder paste and component batch records, placement machines, SPI/AOI equipment, and reflow oven temperature data. Data is updated in real time for each work order and each line change. Documents combine structured tables and unstructured text, with fields including part number, reference designator, alternative part, coordinate value, furnace temperature parameters, and batch number. Corresponding units include millimeters, degrees Celsius, and batch numbers.
What Constraints These Characteristics Impose on the Document Parsing and Chunking Link
Documents in this category combine structured tables and unstructured text, with high data update frequency and large document volume per work order. These factors create multiple constraints for parsing and chunking. Frequently updated real-time work order documents require parsing services to support fast response, to avoid blocking subsequent production processes. Structured documents such as BOMs and line change sheets must retain the corresponding relationship of cell fields. The association between part numbers and reference designators must not be broken during chunking. Documents such as SPI/AOI reports that include defect locations and batch information must retain contextual connections. This ensures subsequent retrieval can accurately match defects with corresponding batches. Parsing large Gerber files and multi-batch reports requires sufficient time to avoid timeout interruptions.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale for This Setting |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the mixed table and text structure of SMT documents, avoiding splitting associated information for part numbers and corresponding reference designators |
chunkSplitter | Split by paragraph + table block | Retains structural integrity for BOM tables, furnace temperature record text, and SPI defect reports, preventing structured data from being incorrectly split |
customPdfParse | Enable table recognition mode | Accurately extract fields such as part numbers and reference designators from structured documents in Excel and PDF formats |
PARSE_FILE_TIMEOUT_SECONDS | 120–180 seconds | Adapts to parsing durations for large Gerber files and multi-batch SPI reports, avoiding parsing failures due to insufficient time |
UPLOAD_FILE_MAX_SIZE | 500 MB | Supports upload requirements for batch documents per work order, meeting storage and parsing needs for batch documents at EMS factories |
chunkSettingMode | Manual configuration | Customize chunking rules for special SMT document fields such as coordinate values and furnace temperature parameters, to avoid splitting key information |
Three Common Configuration Errors
- After uploading an Excel-format BOM file, parsed part number and reference designator fields are empty. Cause: The table recognition mode of
customPdfParseis not enabled, causing structured tables to be misjudged as plain text and split. - After uploading a large Gerber-related document, the interface displays a request failure with a timeout status code. Cause:
PARSE_FILE_TIMEOUT_SECONDSis set below 120 seconds, leaving insufficient parsing time. - Defect locations and corresponding batch numbers in SPI defect reports are split into different chunks in the chunking result. Cause:
chunkSplitteronly splits by character length, and the mixed splitting rule for paragraphs and table blocks is not enabled.
How to Verify Proper Configuration
- Upload a standard SMT BOM Excel file, check if parsed text retains complete corresponding relationships between part numbers, reference designators, and alternative parts.
- Upload a single-batch SPI/AOI report PDF, check if the chunking result integrates defect locations, inspection time, and batch number into the same chunk.
- Upload a batch line change sheet compressed package within 500 MB, confirm no timeout errors appear during parsing progress.
- Check the background parsing log of FastGPT v4.17.0, confirm the table recognition switch for
customPdfParseis enabled.
The values listed above are common starting points. Adjust them based on specific samples and use cases relevant to the deployment.