What this category's data looks like
Daily documents for SMT mounting and electronic assembly include BOM and coordinate files, changeover sheets, stencil and reflow oven temperature records, first article inspection reports, SPI/AOI inspection reports, maintenance records, traceability labels, and more. Data sources cover customer BOMs, Gerber files, coordinate files, MES systems, solder paste and component batch ledgers, placement machines, SPI/AOI equipment, and reflow oven temperature data. Update frequency follows per work order, each changeover, or real-time updates. Documents include specific fields such as part number, reference designator, component parameters, temperature, and coordinates. Some documents use structured table formats, while others use natural language process descriptions.
Constraints imposed on document parsing and chunking by these characteristics
This category includes diverse document types, with both structured tables and natural language descriptions. Associated fields such as part numbers and coordinates must have their contextual relationships preserved accurately. High-frequency, real-time updated work order documents require efficient parsing processes to prevent delays that disrupt production links. Field formats and units differ across documents, so parsing rules must be adapted to each type. Large batch-uploaded BOM and reflow temperature record files require extended parsing times, so timeout risks must be managed. Errors can lead to serious production issues such as incorrect component placement and cold solder joints. Parsing and chunking must therefore retain all core fields and their associated relationships as completely as possible.
Configuration settings
The following configurations are based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the information density per chunk of SMT documents, avoiding loss of associated fields such as part numbers and coordinates after splitting |
chunkSplitter | Mixed paragraph and fixed-length splitting | Adapts to document types that contain both natural language process descriptions and structured tables |
customPdfParse | Enable structured table extraction | Adapts to documents with tables such as SPI/AOI reports and reflow temperature records, retaining the original field alignment format |
PARSE_FILE_TIMEOUT_SECONDS | 120 seconds | Adapts to the parsing duration of batch work order documents, avoiding timeouts for large BOM files |
UPLOAD_FILE_MAX_SIZE | 200 MB | Supports batch uploading of large work order files, covering all categories of production documents |
chunkSettingMode | Manual configuration | Allows adjustment of chunking rules for different document types such as changeover sheets and first article inspection reports |
Three common configuration errors
- Symptom: After custom chunking configuration, the knowledge base automatically deletes duplicate document chunks, causing the custom index order to not match the actual storage. Cause: The system enables global document deduplication rules by default, and the corresponding verification is not disabled for high-frequency duplicate fields in SMT work orders.
- Symptom: When creating a knowledge base via API, real-time parsing status (such as parsing in progress, ready, failed) cannot be obtained, only the final result can be queried. Cause: Asynchronous parsing callbacks are not configured, or the correct status query interface is not called.
- Symptom: Large BOM or reflow temperature record files fail to parse, returning timeout-related errors. Cause: The
PARSE_FILE_TIMEOUT_SECONDSconfiguration value is lower than the actual parsing duration of large files, and the threshold is not adjusted for batch work order documents.
How to verify correct configuration
- Upload a typical changeover sheet document, check if the parsed chunks retain the contextual relationship of core fields such as part number, reference designator, and changeover time.
- Call the API to query the status of the knowledge base parsing task, confirm that real-time statuses of parsing in progress, ready, and failed can be obtained.
- Upload a BOM file containing duplicate part numbers, check if all custom chunked content is retained, and duplicate chunks are not automatically deleted.
- Upload a stencil parameter file close to 200 MB, confirm that parsing does not trigger timeout errors.
The values provided are common starting points and should be measured against the reader's own samples.