This page is compatible with FastGPT v4.17.0.
Data Profile for This Category
This category covers daily documents from petroleum refining enterprise production operation, process, scheduling, equipment, and laboratory departments. These documents include operating procedures, production plans, material balance sheets, distillate export quality reports, interlock ledgers, shift handover records, and change management files. Data sources include DCS/APC systems, LIMS laboratory management systems, tank farm metering systems, crude oil evaluation reports, equipment management systems, and energy management systems. Update frequencies cover real-time, per-shift, and daily updates. Document fields include process parameters with clear units, such as degrees Celsius for distillation range, tons for material quantity, and timestamps for interlock triggers.
Constraints for Citation Sources and Traceability
Multi-source, heterogeneous data structures require the traceability workflow to use distinct paths for structured system export fields and unstructured document paragraphs, to avoid mixing information from different sources. Real-time, per-shift, and daily update frequencies require the associated knowledge base to support incremental synchronization and version tagging, to ensure cited content uses the latest valid version. The business characteristic of multiple continuous units coupled via logistics and energy networks requires traceability to link cross-unit related documents, to preserve complete context for Q&A content. Process fields with clear units require retaining original unit information during traceability, to prevent information deviation from unit conversion.
Configuration Settings
Three core parameters must be configured: quoteMaxToken, responseData, and detail. Combine these with Chinese UI settings to complete the overall configuration, as detailed below:
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Chunk Length | 800–1200 characters | Adapt to the paragraph structure of petroleum refining process documents, retain complete process steps |
| Overlap Length | 150–200 characters | Prevent cross-chunk process logic from being truncated, ensure traceability continuity |
| Single Recall Limit | Top 6–10 entries | Cover the scope of cross-unit coupled related documents |
quoteMaxToken | 1200–1800 token | Adapt to the long-text traceability requirements of petroleum refining process specifications |
responseData | Enabled | Retain structured fields and unit information from original documents |
detail | Full | Output the full path and update time of traceable documents |
Three Common Configuration Errors
- Symptom: A red numeric error pop-up appears during invocation. Cause: The knowledge base path bound to traceability is not correctly configured, causing the system to fail to match the storage node of the original document.
- Symptom: No traceability source information is attached to Q&A output. Cause: The full display of the
detailparameter is not enabled, or the configured chunk length is too small, resulting in the traceable paragraph not being fully recalled. - Symptom: Loading fails after associating multiple petroleum refining process knowledge bases. Cause: The single recall limit is not adjusted based on the total volume of enterprise documents, or a reasonable total knowledge base word count threshold is not set.
How to Verify Successful Configuration
- Submit a query about the operating procedures of a specific unit, check whether the output result includes the name and specific paragraph location of the original document.
- Access the knowledge base management page, confirm that the configured traceability binding path matches the actual storage path.
- Trigger a query, check whether the system log includes version and update time information of the traceable document.
- Adjust the single recall limit, test whether multiple unit knowledge bases load normally.
The values provided are common starting points and should be measured against the reader's own samples.