What the data for this category looks like
This category covers daily documents from refining enterprise production, process, scheduling, equipment, and laboratory departments. Documents include operating procedures, production plans, material balances, distillate export quality reports, interlock ledgers, shift handover records, and change management files. Data sources include DCS/APC, LIMS, tank farm metering, crude oil evaluation, equipment management systems, and energy systems. Update frequencies include real-time, per shift, and daily. Most documents contain fields such as process parameters, material quantities, and equipment status. Units include tons, cubic meters, degrees Celsius, flash point units, and others. The structure primarily combines structured tables and process descriptions.
What constraints these characteristics impose on the citation source and traceability link
Multiple continuous units in this category are coupled via material and hydrogen networks. Raw material properties have large fluctuations. Citation traceability must accurately associate with crude oil evaluation data and real-time unit parameters for the corresponding batch. Document types cover structured reports and unstructured procedures. Field formats vary significantly across different documents. The traceability link must match the field rules of the corresponding document to avoid incorrect cross-type document association. Data update frequencies include real-time, per shift, and daily. Traceability must filter valid sources based on data update cycles to ensure cited content matches current production status. Additionally, traceability for process fluctuations or interlock events requires quick location of corresponding ledgers and change records to support risk handling.
Configuration Settings
This document is compatible with FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
quoteMaxToken | 1500–2000 token | Matches the single-section content length of petroleum refining technical documents, avoids truncating key process parameters |
responseData | Return only structured document fields | Adapts to the characteristics of multiple structured reports in refining documents, simplifies traceability display |
detail | Enable full traceability | Supports displaying complete source information for distillate export quality reports and interlock ledgers |
| Traceability data source scope | Only associate data sources for the current unit | Matches the production scenario of multiple coupled units, avoids confusion of cross-unit data |
| Number of traceability entries displayed per single query | Top 3 entries | Controls display content density, aligns with the quick review habits of process personnel |
Three common mistakes
- Phenomenon: An error "No associated data source found" appears when viewing citation sources in chat responses. Cause: The data source association rule for the current unit is not configured, causing the traceability link to fail to match corresponding DCS or LIMS data.
- Phenomenon: After workflow invocation, the input and response citation content of knowledge base search are still displayed, and only the answer cannot be shown. Cause: The default switch for traceability display is not turned off, or the display rule for returning only answers is not configured.
- Phenomenon: Cited content truncates key process parameters, making it impossible to fully view distillate indicators. Cause: The value of
quoteMaxTokenis set too low, causing single-section document content to be truncated.
How to confirm the configuration is complete
- Initiate a query related to crude oil properties or unit load, check whether citation sources only associate data sources for the current production unit.
- Check the completeness of cited content, confirm that no key process parameters are truncated.
- Trigger workflow invocation, confirm that the response only displays the question answer, and does not include the input and response fields of knowledge base search.
- View system error logs, confirm that the prompt "No associated data source found" does not appear.
The values given are common starting points and should be measured against the reader's own samples.