What data for this category looks like
Daily data sources for SMT mounting and electronic assembly include customer BOMs, Gerber files, mounting coordinate files, MES systems, mounters, SPI/AOI inspection equipment, reflow oven temperature records, and more. Data updates in real time per work order and each line change. Core documents include BOMs and coordinate files, line change sheets, stencil and oven temperature records, first article confirmation reports, SPI/AOI defect reports, repair records, traceability labels, and more. Data fields include part numbers, reference designators, component lot numbers, mounting coordinates, oven temperature parameters, and more. Units include millimeters, degrees Celsius, lot numbers, and more.
Constraints on source citation and traceability
Data for this category updates in real time per work order. Citation and traceability must match the latest data for the current work order, and cannot use historical redundant data. Multiple document types have large field differences. Dedicated field mapping rules must be configured for different documents such as BOMs, oven temperature records, and SPI reports. This ensures that part numbers, component lot numbers, and defect information required for audits are accurately associated. Traceability labels are bound to work order IDs. Source citations must associate all relevant documents via work order IDs to avoid mixing data across work orders. The audit link must also accurately match materials provided by suppliers and production records. The accuracy of citation and traceability directly affects the effectiveness of audit results.
Configuration settings
Configuration settings on this page are based on FastGPT v4.17.0.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
quoteMaxToken | 800–1200 token | Most single SMT-related documents are 500–1000 characters in length. This range retains all key parameters of cited content fully. |
responseData | Only return document data for associated work orders | This scenario requires precise matching of BOMs, oven temperature records, and SPI reports related to supplier audits. Irrelevant data is not needed. |
detail | Enabled | Subfields such as component lot numbers and mounting coordinates must be extracted for traceability to meet audit requirements. |
| Single recall limit | Top 6 entries | Core documents involved in SMT audits do not exceed 5 categories. Excessive recall will disrupt association accuracy. |
| Context window | 12000–16000 token | Total length of multiple work order documents must fit large model processing capabilities to avoid truncation of critical information. |
Three common misconfigurations
- Logs cannot distinguish records from different data sources. This occurs when association rules based on work order IDs or document types are not configured, leading to mixing of component lot data and oven temperature data from different batches.
- The phrase "Citation Marker: [1]" remains in large model output. This happens when automatic citation marker output is not disabled, and returned document content is not cleaned of formatting.
- Some audit documents have empty citation IDs. This occurs when the
detailconfiguration is not enabled, so unique identifier fields such as traceability labels and work order IDs are not extracted, making valid source citations impossible to generate.
How to confirm proper configuration
- Upload a real SMT work order document, initiate a supplier audit query, and check if returned source citations include core documents such as BOMs and oven temperature records for the corresponding work order.
- Review system logs to confirm that records from different data sources are correctly labeled, and data from different sources can be distinguished via work order IDs or document types.
- Check large model output content to confirm that phrases like "Citation Marker: [number]" do not appear.
- Verify that fields in source citations include subinformation such as component lot numbers and mounting coordinates to meet audit requirements.
The configuration values provided here are common starting points. Validate them against specific samples and use cases.