What data belongs to this category
The primary data stakeholders for this category are engineering, production, quality control, and laboratory personnel at printed circuit board (PCB) factories. Daily documents include engineering production instructions, Gerber analysis records, drilling and lamination parameters, chemical solution analysis reports, AOI reports, electrical test reports, and sectioning reports. Data sources include customer Gerber files, drill tapes, BOM lists, CAM files, equipment operating parameters, bath solution test results, AOI, electrical test, and laboratory detection data. Data is updated and generated per production batch, per shift, or via continuous collection, covering the full process from customer data conversion to finished product inspection.
What constraints do these characteristics impose on the "citation sources and traceability" link
The multi-batch, high-frequency updated data characteristics of PCB manufacturing require citation traceability to accurately match documents to a single production batch, to avoid mixing data across batches. Significant field differences exist across different documents: for example, lamination parameters include pressure and temperature units, and sectioning reports include aperture deviation values. Original fields and units must be retained during traceability. Both continuously collected real-time data and batch-generated report data are present. The collection timestamp and generation node of data must be distinguished to ensure the latest valid data for the corresponding audit link is cited. Citing the wrong core document may lead to scrapping of an entire batch of boards, so the corresponding relationship between data sources and audit links must be strictly bound.
How to configure the system
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
quoteMaxToken | 8000–12000 token | PCB manufacturing audit documents include multiple professional parameter reports, so sufficient citation content must be accommodated to cover full-process audit information |
responseData | Enable and bind the original document ID | Unique identifiers for data sources must be retained to enable precise traceability to specific documents for the corresponding production batch |
detail | Full output | Audits require detailed data across multiple links including drilling, lamination, and inspection; full output ensures comprehensive information with no omissions |
| Similarity threshold | 0.75–0.85 | There are many professional terms in PCB documents; this range balances recall accuracy and coverage, avoiding omission of key parameters |
| Single search limit | Top 6 entries | The number of audit-related documents for a single production batch is limited; excessive citations will exceed context capacity limits |
| Reranking | Enable | Relevance judgment for PCB documents requires combining professional semantics; reranking optimizes the sorting accuracy of search results |
Three common configuration mistakes
- Phenomenon: After enabling the reranking function, search results are empty or citation fragments disappear. Cause: The similarity threshold is set too low, filtering out PCB professional document fragments that originally meet requirements.
- Phenomenon: Citation requests time out, and the parameter report content required for audits cannot be returned. Cause:
quoteMaxTokenis set too small, or context loading does not adapt to the loading requirements of long documents. - Phenomenon: Custom citation prompts or templates cannot be set, and only basic parameters can be adjusted. Cause: The FastGPT version in use is lower than v4.17.0. Upgrade to this version or later to access the advanced configuration function.
How to confirm the configuration is correct
- Upload typical documents for a single production batch, initiate an audit query, and verify that the returned results include the original identifiers and field content of the corresponding documents.
- Adjust the enabled status of the reranking function, compare the sorting and recall quantity of search results, and confirm that the reranking function operates correctly.
- Check the system version identifier, confirm that the current FastGPT version is v4.17.0 or higher, to ensure the advanced configuration function is available.
- Simulate cross-batch document retrieval, confirm that returned results only associate the production batch currently being audited, with no mixed data across batches.
The values provided are common starting points and should be measured against the reader's own samples.