Data Characteristics for This Use Case
PCB manufacturing scenario data originates from customer Gerber files, drill files, BOM lists, CAM engineering files, equipment operating parameters, tank solution test records, AOI inspection reports, electrical testing reports, and sectioning reports. Data updates are generated per production batch, shift, or through continuous collection. Each document includes structured parameter fields and semi-structured inspection records. Fields cover professional parameters such as drilling rotation speed, lamination temperature, and hole copper thickness, with units including revolutions per minute, degrees Celsius, and micrometers, among others.
Constraints Imposed on Document Parsing and Chunking
PCB manufacturing documents include non-standard format engineering files and tabular inspection reports. Parameter fields are closely interconnected. The high-frequency update requirement means chunking must support incremental data processing. Format differences across multiple data sources require the parsing step to support multiple file types such as Gerber, PDF, and tabular files, while also preserving logical connections between parameters to avoid loss of professional information during splitting. The volume of frequently collected data grows with production scale, so chunking logic must balance parsing efficiency and content completeness.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | PCB documents contain closely linked parameter groups. Each chunk must cover complete parameter descriptions to avoid splitting key fields |
chunkSplitter | By paragraph + heading hierarchy | Engineering documents have chapter headings and paragraph-style parameter descriptions. Retaining hierarchy maintains logical connections between content |
customPdfParse | Enabled | PCB-specific PDFs contain tables, formulas, and non-standard layouts. Custom parsing preserves original structure and parameter associations |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Large Gerber parsing files or batch inspection report uploads require longer processing time, preventing task interruption from timeout |
UPLOAD_FILE_MAX_SIZE | 1000 MB | The total size of multiple inspection reports from a single production batch may be large, adapting to storage and parsing requirements for batch uploads |
Common Misconfigurations
- Phenomenon: Parsing API calls return
408 Request Timeout. Cause: ThePARSE_FILE_TIMEOUT_SECONDSparameter was not adjusted, and large file parsing time exceeded the default threshold. - Phenomenon: Parsed chunks lose the association between drilling parameters in tables and defect records. Cause:
customPdfParsewas not enabled, and the default parsing does not retain the table structure and layout logic of professional PDFs. - Phenomenon: After uploading multiple files, only chunked results for a single document are returned. Cause: Multi-file content was not spliced per API parameter specifications, causing the chunking logic to only recognize a single file.
How to Verify Correct Configuration
- Users may upload a test PDF containing a drilling parameter table and an AOI inspection report, then review the parsed chunks to confirm they retain the association between parameters and defect information within the table.
- Users may upload a single test file larger than 100 MB, then confirm the parsing process does not trigger a timeout error and that chunked results fully cover the document content.
- Users may call the API interface to upload multiple test documents, then check the returned chunked results to confirm they include valid content from all uploaded files.
- Users may review background configuration items in FastGPT v4.17.0, then confirm the values of parameters such as
customPdfParseandPARSE_FILE_TIMEOUT_SECONDSmatch the preset configuration.
The values provided are common starting points and should be evaluated against the reader's own samples.