What this category's data looks like
This category of data originates from engineering, production, quality control, and laboratory departments of printed circuit board factories. It covers engineering production instructions, Gerber analysis records, drilling and lamination parameters, chemical solution analysis, AOI reports, electrical test reports, and slice test reports. Data sources include customer-provided Gerber files, drill tapes, BOM files, CAM output files, equipment operating parameters, bath test results, and AOI, electrical test, and laboratory detection data. Data is updated per production batch, per shift, or via continuous collection. Documents include structured parameter fields and unstructured report content. For example, drilling parameters include fields with clear units such as spindle speed and feed rate, while lamination parameters include fields such as pressure and temperature.
What constraints these characteristics impose on the knowledge base retrieval and recall link
The multi-source scattered data, high-frequency update rhythm, and mixed document types of PCB manufacturing create multiple constraints for retrieval and recall. Retrieval must support cross-collection association to link multi-source data with customer information and production link data. Recall logic must adapt to incremental synchronization to avoid delays caused by full-volume retrieval for high-frequency updated data. Mixed document types, including structured parameter tables and unstructured reports, require retrieval modes to balance semantic and keyword matching. Misjudgment of process abnormalities can lead to full batch scrapping, so recall results must have clear traceability capabilities to ensure content comes from specified process document collections.
How to configure the settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the precision requirements for PCB process parameters, preventing irrelevant lamination parameters from being recalled in drilling process Q&A scenarios |
limit | Top 6 entries | Single batch production data volume is large, limiting the number of recalled entries avoids context overload while covering reference information for core process links |
searchMode | mixedRecall | Adapts to retrieval of mixed structured and unstructured documents such as Gerber and CAM files, balancing semantic understanding and precise keyword matching |
usingReRank | Enabled | Distinguishes slice abnormalities in AOI reports from conventional process deviations, optimizing the ranking priority of recall results |
datasetSearchUsingExtensionQuery | Enabled | Supplementary retrieval of associated fields from customer BOMs and drill tapes, covering common issues in engineering data conversion links |
collectionIds | Bind exclusive collections to production workshops | Distinguishes process parameters and test data from different production lines, avoiding retrieval errors caused by cross-workshop data confusion |
Three common configuration mistakes
- Phenomenon: Retrieval results do not include knowledge base source identifiers, making it impossible to verify whether content comes from uploaded process documents. Cause: FastGPT's knowledge base traceability configuration is not enabled, and source information of document collections is not bound in the output.
- Phenomenon: Service logs display exceptions or security alerts related to TRACE requests. Cause: FastGPT's built-in TRACE request support is not disabled, exposing test interface risks from HTTP protocol implementation.
- Phenomenon: Newly uploaded drilling parameter or AOI report data cannot be recalled. Cause: Incremental synchronization rules triggered per production batch are not configured, and fixed-cycle full-volume synchronization is still used, failing to match the high-frequency update rhythm of PCB data.
How to confirm the configuration is successful
- Initiate a retrieval containing the keyword drilling speed, check whether the results include the source collection name of the uploaded document.
- Check the HTTP service configuration items of FastGPT v4.17.0, confirm that TRACE requests have been disabled.
- Upload a new lamination parameter document, initiate retrieval for the corresponding process, confirm that the new document content is normally recalled.
- Check the configuration of the retrieval request, confirm that only exclusive collections bound to PCB process documents are called, and no irrelevant datasets are mixed in.
The values provided are common starting points and should be measured against the reader's own samples.