What This Category of Data Looks Like
This category of data is maintained jointly by the R&D, process, quality, production, and procurement teams of manufacturing enterprises. Daily documents include ECN/ECR, change assessment forms, drawings and BOM version files, process verification reports, production switching plans, inventory disposal documents, and customer approval documents. Data sources cover PLM, ERP, MES systems, supplier materials, verification test records, and customer-specific requirements. Updates are triggered irregularly with each process or engineering change. The number of documents associated with a single change fluctuates based on change complexity. Fields include change number, version number, effective time, associated material/process nodes, approval status, and other relevant details.
Constraints for Citation Sources and Traceability
Change-related documents are tightly linked by version. Citation and traceability require precise matching of version numbers to prevent mixing of old and new versions. Data sources are scattered across multiple systems. Cross-data-source association and retrieval must be supported to ensure cited content comes from the latest archived versions of corresponding PLM, MES, and other systems. A single change is associated with a large number of documents. Retrieval scope must be limited to avoid context overload. Change effective times are clearly defined. Traceability must be tied to the effective node to ensure the audit process references the correct version before or after the switch. Customer approval documents are mandatory compliance basis. Traceability must retain complete approval chains and customer signing records, with no associated items omitted.
Configuration Settings
| Configuration Item | Recommended Approach | Rationale |
|---|---|---|
| Single Retrieval Limit | 6–10 items | The number of documents associated with a single change fluctuates with complexity; avoid excessive retrieval causing context overload |
| Maximum Context Token Count | 8000–12000 token | Covers the complete content of multiple types of associated documents such as ECN, BOM, and verification reports |
| Version Matching Rule | Enable precise version number matching | Avoid mixing old and new versions, ensure cited documents use the correct version before or after change takes effect |
| Data Source Binding | Bind specified PLM, ERP, MES data sources | Covers core data sources for this category, ensure cited content comes from official system archives |
quoteMaxToken | 1500–2000 token | Limits the output length of a single cited document, avoid long documents occupying excessive context space |
detail | Enabled | Display complete version number and data source information for cited documents, facilitating traceability verification |
Three Common Configuration Errors
- Phenomenon: Text content extraction component fails to extract fields from knowledge base citation results, returning empty results or non-target fields. Cause: The
responseDataparameter is not configured to enable structured data return, only unstructured text fragments are obtained. - Phenomenon: Citation results include documents not associated with the current change, with quantity exceeding expectations. Cause: Corresponding data sources are not bound, documents from other business modules are retrieved, and the change association scope is not limited.
- Phenomenon: Cited document versions do not match current change requirements, with old and new versions mixed. Cause: Version matching rules are not enabled, retrieval is only performed via keyword matching, and document version number fields are not identified.
How to Verify Proper Configuration
- Initiate a change-related query using the FastGPT v4.17.0 console, and verify that returned citation sources include documents from specified data sources such as PLM and ERP.
- Review the version number field in citation results to confirm it matches the effective version of the current change.
- Count the number of documents retrieved per query to confirm it does not exceed the preset single retrieval limit.
- Test the text content extraction component to confirm it can extract target fields including change number and version number from knowledge base citation results.
The values provided are common starting points and should be measured against the reader's own samples.