What Data Looks Like for This Category
Technical documents for PCB manufacturing come from multiple sources. These include Gerber files, drill tapes, and BOMs provided by customers. They also include CAM-generated production instructions, drilling and lamination parameters, chemical analysis records, and inspection reports from AOI, electrical test, and slice testing.
Documents update on a per-production-batch or per-shift schedule. Some device data is collected continuously. Most document structures include parameter values, process steps, and test results. Common fields cover professional parameters such as hole diameter, copper thickness, and offset. Supported units include millimeters, ounces, and percentages.
What Constraints Do These Characteristics Impose on Document Parsing and Chunking
PCB manufacturing document characteristics create multiple constraints for the parsing and chunking process. Multi-source heterogeneous formats, including vector Gerber files, Word process documents, and PDF inspection reports, require parsing tools to support distinct structure recognition logic for each format. Bulk uploaded production batch documents with high update frequency need processing timelines that match production rhythms. Documents with dense professional fields require retaining the association between parameters and their corresponding process descriptions during chunking, to avoid broken information after splitting. Long table documents such as lamination parameter summary sheets are easily damaged by traditional chunking logic. Table-adaptive chunking rules are needed to preserve integrity.
How to Configure
These configurations are adapted for FastGPT v4.17.0.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | PCB manufacturing documents contain many professional parameters and process descriptions. This range retains the association between parameters and their context, avoiding split information breaks |
chunkSplitter | Split by paragraph + table recognition | Adapts to the mixed structure of long tables and process paragraphs in documents, avoiding damage to table integrity |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Some large Gerber-related documents or bulk parsing tasks require longer processing time, matching the document processing rhythm of production batches |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Supports bundled uploads of multiple documents for a single production batch, adapting to bulk processing needs |
customPdfParse | Enable table-first parsing mode | For long tables in AOI and electrical test reports, prioritize retaining table structure to avoid data misalignment after chunking |
chunkSettingMode | Manual configuration | Flexibly adjust chunking parameters for different types of PCB documents such as production instructions and test reports, to adapt to diverse document structures |
Three Common Misconfigurations
- When uploading Gerber-related PDF documents larger than 10 MB, the interface shows a timeout error. The cause is failure to adjust the
PARSE_FILE_TIMEOUT_SECONDSparameter. The default duration is insufficient for parsing large PCB professional documents. - Some drill parameter documents in docx format cannot be parsed normally. The cause is failure to enable the industrial document-adaptive parsing mode. The default general parsing logic cannot recognize the parameter layout format unique to the PCB industry.
- AOI report tables in the knowledge base are split into multiple independent chunks, leading to disconnection between parameters and anomaly descriptions during subsequent retrieval. The cause is failure to enable the table-first recognition rule in chunking configuration. Traditional paragraph chunking logic damages table structure.
How to Confirm Proper Configuration
- Upload a single typical PCB document such as a lamination parameter PDF, and check the parsed chunk results. Confirm that long tables are not split.
- Upload batch production batch documents. Check that the completion time of parsing tasks matches the production rhythm, with no timeout errors.
- Search for professional parameter content in the knowledge base. Confirm that search results correspond to the original document paragraphs, with no separation between parameters and their context.
- Upload PCB documents in different formats such as Gerber, docx, and CSV. Confirm that all formats can be parsed normally.
The values listed here are common starting points. Adjust them based on your own samples.