What Data for This Category Looks Like
Data for this category primarily comes from product drawings and 3D models, CAPP/CAM systems, machine tool collection terminals, tool management systems, MES systems, and measurement and inspection tool data. Daily documents include process routes, operation cards, tool lists, CNC programs, first-piece inspection records, patrol inspection forms, and measuring tool calibration records. Data updates are triggered by work order execution, part machining, sampling inspection, and tool replacement. Document fields cover processing parameters, tool life, inspection results, and other relevant details, with units including millimeters, revolutions per minute, hours, and other machining-specific units.
What Constraints Do These Characteristics Impose on Knowledge Base Retrieval and Recall?
The scattered data sources and numerous update trigger points for this category require the knowledge base to support batch import of multiple formats and incremental updates, to avoid retrieving outdated process parameters or inspection records. Documents contain multiple formats such as structured tables, CNC code, and inspection forms, and most fields use machining-specific terminology. This requires the retrieval link to adapt to professional vocabulary matching, while distinguishing the recall weights of different document types. A single machining work order is associated with multiple types of documents, so retrieval must accurately match core information such as the current process’s clamping scheme and tool parameters, to avoid retrieving irrelevant content that reduces drafting efficiency.
Configuration Settings
These configurations are based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the precision of metal cutting machining professional terminology, to avoid retrieving irrelevant process or tool information |
limit | Top 8 entries | Covers multiple types of documents such as current process routes and tool lists, while controlling context length |
searchMode | mixedRecall | Adapts to the retrieval needs of metal cutting documents in mixed formats including CNC program code, process tables, and inspection forms |
usingReRank | Enabled | Re-sorts retrieval results to prioritize core information such as the current work order’s clamping scheme and tool parameters |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands retrieval to historical documents such as tool wear, dimensional drift, and surface roughness abnormalities based on the user’s input exception report requirements |
collectionIds | Configured by workshop and process grouping | Distinguishes process data from different machining workshops, to avoid confusion between cross-workshop documents |
Three Common Misconfigurations
- Phenomenon: A parameter error is returned when calling the chat API with the
collectionIdsparameter. Cause: The knowledge base ID was not converted to a string format, or the passed ID does not correspond to the currently enabled dataset. - Phenomenon: The number of retrieval results is far fewer than expected, and cannot cover necessary documents such as the current process’s process route and tool list. Cause: The
limitparameter is set too low, failing to match multiple types of process-related documents. - Phenomenon: After modifying knowledge base content, the exported dataset still uses old version content. Cause: A full index update was not triggered on the knowledge base management page, and the exported data is cached old data.
How to Confirm Configuration Is Correct
- Enter the current process’s tool model and clamping scheme keywords, and check whether the corresponding tool list and cutting parameter documents are included in the retrieval results.
- Pass the preset
collectionIdsparameter, and check whether the retrieval results only include process data from the specified grouping. - Trigger a full knowledge base update, then export the dataset, and compare whether the modified documents appear in the exported file.
- After enabling
usingReRank, check whether the top result in the retrieval list is the core process file for the current work order.
The values given are common starting points and should be measured against the reader's own samples.