What this category of data looks like
Daily documents for SMT mounting and electronic assembly include BOMs and coordinate files, line change lists, stencil and furnace temperature records, first article confirmation reports, SPI/AOI inspection reports, maintenance records, and traceability labels. Data sources include customer-provided BOMs, Gerber files, coordinate data, factory MES systems, solder paste and component batch ledgers, mounters, SPI/AOI equipment, and reflow oven temperature data. Data update frequency follows each work order, each line change, or real-time generation. Documents contain structured fields such as part numbers, reference designators, alternative parts, temperature parameters, and coordinate values. Some documents have nested multiple tables and complex formatting.
What constraints these characteristics impose on the document parsing and chunking link
Documents in this category have strong associations between structured fields, high update frequency, and single-document volume that grows with work order batch size. Part numbers and reference designators in BOMs and AOI reports must retain their full associated relationships. Chunking must not cut field associations across paragraphs or tables. Real-time updated work order data requires efficient parsing processes to avoid production delays from timeouts. Multiple document types including PDF furnace temperature records, DOCX line change lists, and scanned stencil reports require different parsing logic. General chunking rules cannot cover all layout formats. Bulk-uploaded work order documents consume more parsing resources. Parameters must be adjusted to avoid upload or parsing failures.
How to set the configurations
| Configuration Item | Recommended Value | Basis for This Setting |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the length of single records in SMT document BOMs and AOI reports, avoiding splitting the association between part numbers and reference designators |
chunkSplitter | Split by paragraph + table | Adapts to structured data in line change lists and furnace temperature records, retaining field associations for each row within tables |
PARSE_FILE_TIMEOUT_SECONDS | 1200–1800 seconds | Addresses bulk work order documents larger than 10 MB, avoiding parsing failures caused by insufficient default timeout duration |
UPLOAD_FILE_MAX_SIZE | 10000 MB | Supports multi-document packaged uploads for single-batch work orders, avoiding uploads being blocked |
customPdfParse | Enable PDF Marker adaptation mode | Adapts to complex-layout PDFs or scanned documents such as stencils and furnace temperature records, improving parsing accuracy |
chunkSettingMode | Manual configuration | Adjust chunking rules for different document types such as BOMs and AOI reports, avoiding damage to business field associations from general chunking |
Three common mistakes
- Timeout errors occur when uploading work order documents larger than 10 MB. The cause is failure to adjust the
PARSE_FILE_TIMEOUT_SECONDSparameter; the default timeout duration is insufficient for parsing bulk structured documents. - Some DOCX-format line change lists cannot be parsed normally. The cause is failure to enable the
customPdfParseadaptation mode; general parsing cannot recognize reference designator and part number association fields within tables. - Only partial row content is extracted from AOI report tables in the knowledge base. The cause is an overly small
chunkSizesetting, which cuts cross-table row content for the same defect record during chunking.
How to confirm the configuration is correct
- In FastGPT v4.17.0, upload a single 10 MB BOM PDF document, check whether the parsing process completes normally, and confirm that the
PARSE_FILE_TIMEOUT_SECONDSparameter takes effect. - Import an AOI report containing multiple tables, check whether the chunking results retain complete defect records and corresponding reference designator information.
- Bulk upload five or more work order documents, confirm there are no upload failure prompts, and verify that the
UPLOAD_FILE_MAX_SIZEparameter does not restrict upload scale. - Perform parsing on a scanned stencil PDF document, check whether the parsing results retain key parameters of stencil openings, and confirm that the
customPdfParseadaptation mode takes effect.
The values provided are common starting points and should be measured against the reader's own samples.