What the data for this category looks like
Business data for the sheet metal cutting and bending scenario mainly comes from CAD/CAM design software, order BOMs, sheet metal inventory systems, laser cutter and bending machine equipment logs, and MES production management systems. Daily documents include 3D part unfolding drawings, CNC cutting programs, bending process cards, nesting diagrams, first-piece inspection records, and process transfer cards. Data updates follow this rhythm: exclusive process documents are created per order, sheet metal inventory and equipment operation data are synced daily, and bending compensation coefficients and nesting plans are updated each time production changes model. Document fields include engineering numerical values and text identifiers such as sheet thickness (mm), bending angle, unfolded dimensions (mm), nesting utilization rate, and process number.
What constraints do these characteristics impose on the knowledge base retrieval and recall link
The multi-variety, small-batch business nature of this category requires retrieval and recall to accurately match the exclusive process parameters of individual orders, and avoid interference from cross-category documents. Most documents are engineering-structured content, including numerical parameters and drawing-related information. General text retrieval cannot cover parameter matching requirements. Frequently updated business data requires the knowledge base to support incremental synchronization, to avoid retrieving expired bending coefficients or nesting plans. Bid document and quotation preparation needs to associate multiple types of documents. For example, unfolding drawings and bending process cards must be recalled at the same time. Single-document recall cannot support complete quotation calculation logic.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | The process parameter accuracy requirements for sheet metal cutting and bending are high. Low-match irrelevant documents must be filtered out to avoid incorrect reference of excessively deviated bending coefficients |
limit | Top 8 entries | The number of documents associated with a single order in this category is limited. Retrieving too many entries at one time will exceed the large model context window, while ensuring coverage of multiple reference documents for the same process |
searchMode | mixedRecall | Supports joint matching of text keywords (such as "electrical cabinet body bending") and numerical parameters (such as "2mm cold-rolled sheet"), adapting to the retrieval needs of engineering documents |
usingReRank | Enabled | Re-ranks initial retrieval results by business relevance, prioritizes returning core documents including bending compensation coefficients and nesting diagrams, and avoids non-critical content interfering with quotation calculations |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands retrieval terms. Expands "bending process" to "unfolding drawing + bending process card", covering multiple types of documents required in bid documents |
collectionIds | Bind exclusive knowledge base collection ID for sheet metal cutting and bending | Only retrieves exclusive documents in this business domain, avoids mixing process data from other categories, and ensures business relevance of retrieved content |
Three Common Misconfigurations
- Issue: The knowledge base can query data, but there is no output after passing it to the large model. Cause:
collectionIdsis not configured correctly, resulting in the retrieval scope including irrelevant content from non-exclusive knowledge bases, and the large model cannot identify valid context. - Issue: The retrieval result only returns a single document, and cannot obtain the associated content of the nesting diagram and bending process card. Cause:
datasetSearchUsingExtensionQueryis not enabled, cross-document associated recall is not implemented, and complete business information required for bidding cannot be covered. - Issue: Retrieval results include sheet metal process documents with inconsistent accuracy. Cause: A reasonable
similaritythreshold is not set, resulting in low-match incorrect parameter documents being retrieved, which may cause subsequent production and assembly errors.
How to Verify Correct Configuration
- Enter a retrieval term that includes specific sheet metal parameters and process names, and check whether the retrieved results include exclusive documents such as unfolding drawings and bending process cards for the corresponding order.
- View the retrieval logs in FastGPT v4.17.0 to confirm that the
usingReRankfunction is active, and core business documents are ranked at the top of the results. - Test cross-document associated retrieval to verify whether
datasetSearchUsingExtensionQuerycan simultaneously retrieve associated process transfer cards and nesting diagrams. - Check the knowledge base configuration interface to confirm that
collectionIdsis bound to the correct sheet metal cutting and bending business collection, and no irrelevant knowledge bases are included in the retrieval scope.
The values provided here are common starting points, and should be evaluated against appropriate samples.