What the data for this category looks like
The data for this category comes from design BOMs, silicon steel sheet and insulating material procurement data, vacuum drying equipment logs, insulating oil test reports, and high-voltage test station records. Daily documents include winding and core assembly records, vacuum drying records, insulating oil performance reports, partial discharge and withstand voltage test reports, and product nameplate parameter tables. Data updates are triggered upon completion of a single product or batch of materials. Document fields include insulation resistance, partial discharge magnitude, drying temperature, etc., with corresponding units of megaohms, picocoulombs, and degrees Celsius.
What constraints these characteristics impose on the citation sources and traceability link
Each individual product corresponds to exclusive documents and data, requiring the traceability link to accurately bind to a single product batch to avoid mixing data across products. Frequently updated equipment and test data require traceability to support real-time association with newly completed records. Differences in fields across multiple document types require traceability to match the exclusive identifiers of corresponding documents. For example, insulating oil reports must be associated with oil sample numbers, and test reports must be bound to test dates. High-voltage test data has precision requirements, requiring traceability to retain complete references to original test values without arbitrary truncation or modification.
How to configure the settings
The following configurations are based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Maximum Context | Top 250 recall results | Matches the total data scale of multiple documents for a single product to avoid interference from irrelevant data |
quoteMaxToken | 7000–11000 token | Covers the total character count of complete test reports and assembly records for a single product |
responseData | Include original document ID and field values | Meets the requirement to directly associate original production and test records during traceability |
detail | Enable full field display | Retains complete information for high-precision values such as insulation resistance and partial discharge magnitude |
| Single Recall Limit | 300 entries | Adapts to the document count range for a single product in this category |
| Citation Display Threshold | Calibrated via actual measurement | Matches the citation priority of different test reports |
Three common configuration errors
- The issue is that the number of context entries displayed on the page does not match the actual number of recalled entries, and the number of context entries sent to downstream systems also does not match the actual number of recalled entries. The cause is failure to match the exclusive data range of a single product, resulting in recall of redundant data across batches.
- The issue is that the storage path of the original test report cannot be directly associated. The cause is failure to enable the original data inclusion item in the
responseDataconfiguration, and only summary content is returned. - The issue is that the default citation display cannot be customized to be turned off. The cause is failure to correctly configure the trigger rules for citation display, and failure to match the citation priority of documents in this category.
How to confirm the configuration is correct
- Using the FastGPT v4.17.0 interface, initiate a technical document review request for a single transformer, and check whether the original document paths and field values included in the returned results match the corresponding test reports.
- Check the total token count of the returned results to confirm that it falls within the configuration interval of
quoteMaxToken. - Verify that the number of context recall entries matches the document count range for a single product, and adjust the maximum context configuration as needed.
- Confirm that the cited fields include complete unit information, and verify that the
detailconfiguration is enabled.
The values given are common starting points and should be measured against the reader's own samples.