What data for this category looks like
Data for this category includes equipment ledgers, inspection checklists, maintenance plans, work orders, failure analysis, spare parts lists, and special equipment inspection records. Data sources include EAM/CMMS systems, equipment sensor data, operator inspection records, maintenance history archives, spare parts inventory data, and statutory inspection information. Data update cycles cover shifts, daily/weekly/monthly intervals, and immediate triggers when a fault occurs. Individual records contain fields such as equipment number, inspection parameter thresholds, maintenance duration, spare part model, and corresponding units including millimeters, hours, counts, and others.
What constraints these characteristics impose on source citation and traceability
Multi-source data with uneven update rhythms requires the traceability link to bind data update timestamps, ensuring cited content matches current equipment status. Differences in field priority across document types require traceability to first match on-site execution documents such as inspection checklists and maintenance plans, then associate with maintenance history and spare parts data. The timeliness requirements for statutory inspection records require simultaneous verification of inspection cycles during traceability to avoid citing overdue compliance documents. Strongly associated fields for equipment ID and spare part model require traceability to bind cited content to the corresponding equipment via unique identifiers, preventing cross-equipment citation confusion.
How to set up configurations
This configuration is based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
quoteMaxToken | 1200–1600 token | Fits the single-block content length of equipment inspection checklists and work orders, avoids truncating critical parameters |
responseData | Include source document ID and update time | Used to match equipment ledgers and inspection records from EAM systems during traceability |
detail | Enable field-level display | Facilitates verification of core business fields such as spare part models and inspection thresholds |
| Single retrieval recall limit | Top 5–7 entries | Covers the number of commonly referenced documents used for daily team reference |
| Source document association rule | Bind by equipment ID | Matches the strongly associated field characteristics of data in this category |
Three common misconfigurations
- Issue: After setting
quoteMaxTokento 1500, blocks exceeding 1500 tokens in the knowledge base are still cited. Cause: This configuration item only limits the maximum token count for a single cited content block, and does not limit the initial storage length of single-block documents. Over-length blocks are automatically split into chunks before being included in citation candidates. - Issue: After retrieval results are passed to an HTTP request, the AI chat flow cannot generate citation information. Cause: The
responseDataconfiguration for returning source document metadata is not enabled, so results do not carry identifiers and time information usable for traceability. - Issue: Cited content is associated with a different device than the current work order. Cause: Field-level display via
detailis not enabled, and equipment ID fields are not verified, leading to documents from cross-equipment being incorrectly associated with the current task.
How to confirm configurations are correctly set
- Access the knowledge base configuration interface, verify that the value of
quoteMaxTokenfalls within the range suitable for single-block content length of equipment inspections and work orders. - Submit a test query for equipment fault troubleshooting, check that returned results include source document ID and update time metadata.
- Review the citation details in returned results, confirm that core business fields such as equipment number and inspection thresholds are displayed.
- Test retrieval of an over-length inspection checklist document, verify that it is correctly split into chunks and generates compliant citations.
The values provided are common starting points and should be evaluated against the reader's own samples.