What Data Looks Like for This Category
Data sources for this category include customer-provided BOM, Gerber and coordinate files, factory MES systems, solder paste and component batch ledgers, and real-time collected data from placement machines, SPI/AOI detectors, and reflow ovens. Data update cycles cover per work order, each line change, and real-time collection scenarios. Common documents include line change work sheets, stencil maintenance records, furnace temperature curve reports, first article confirmation sheets, SPI/AOI defect reports, and maintenance traceability labels. Fields include part number, reference designator, component batch number, furnace temperature parameters, defect point coordinates, and more. Units include degrees Celsius, millimeters, and other measurement standards.
Constraints Imposed on Knowledge Base Retrieval and Recall
Real-time updated work order and line change data requires the retrieval pipeline to support frequently refreshed document recall. It must avoid using expired line change sheets or furnace temperature records. Multi-source heterogeneous data covering formats such as coordinate files and defect reports requires the retrieval model to prioritize exact matching of core fields like part number and reference designator. High-frequency collected SPI/AOI reports contain large volumes of point coordinate data. The retrieval recall context length must adapt to long text fragments, while filtering redundant information. The batch update feature by work order and line change requires binding the knowledge base to business sets. This prevents cross-work order data confusion, ensuring retrieval results only relate to the current production task.
Configuration Settings
These configurations are adapted for FastGPT v4.17.0:
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches core production fields such as part number and reference designator, filters low-relevance historical documents |
limit | Top 8–12 entries | Covers core documents for the current work order, avoids exceeding context window limits |
usingReRank | Enabled | Performs secondary sorting on recall results by fields such as defect points and furnace temperature parameters, improves matching accuracy |
datasetSearchUsingExtensionQuery | Enabled | Expands retrieval keywords based on production work order numbers, associates corresponding batch line change and furnace temperature records |
collectionIds | Bind the corresponding collection ID by production work order | Isolates knowledge base data for different work orders, prevents cross-task retrieval confusion |
Three Common Mistakes
- Phenomenon: Retrieved documents include table content, but tables do not display correctly in the generated production ledger report. Cause: The knowledge base retrieval did not properly adapt to table format parsing rules, or the context window truncated the complete row structure of the table.
- Phenomenon: The generated report includes line change sheets or SPI report data from other work orders. Cause: The
collectionIdsparameter was not used to bind the knowledge base collection corresponding to the current production task, leading to incorrect recall of cross-batch data. - Phenomenon: Large-size BOM or furnace temperature record documents take too long to upload, and upload priority cannot be adjusted or upload tasks cannot be paused. Cause: Knowledge base upload segment-first scheduling and manual control functions were not enabled, and upload resource order was not allocated based on file size.
How to Confirm Configuration Is Correct
- Upload documents such as BOM and line change sheets for the current work order, bind them to a dedicated knowledge base collection via
collectionIds, run a retrieval test, and confirm only documents within this collection are recalled. - Enter retrieval keywords containing part number and reference designator, check that the matching degree of recall results falls within the preset range, and ensure core production field associations are accurate.
- Generate a production ledger report, check that table content displays correctly, and confirm the retrieval pipeline properly parses document formats.
- Test the upload process for large-size documents, confirm that upload priority can be adjusted or upload tasks can be paused via the interface, meeting business scheduling requirements.
The values provided are common starting points and should be measured against the reader's own samples.