What Data Looks Like for This Category
Data sources for this category include DCS/APC systems, LIMS laboratory information management system, tank farm metering systems, crude oil evaluation database, equipment management systems, and energy management systems. Data update rhythms are divided into three categories: real-time (unit process parameters), shift-by-shift (shift handover records, material balance), and daily (distillate quality reports, change management documents). Daily documents cover operating procedures, production plans, material balance sheets, distillate quality reports, interlock ledgers, shift handover records, and change management documents. Data fields include crude oil density, distillation range, sulfur content, product octane number, equipment operating temperature, and others. Corresponding units include kilograms per cubic meter, degrees Celsius, percentage, and others.
Constraints Imposed on Tool Calling and Plugins
Scattered data sources and varying update rhythms require tool calling to adapt to different synchronization cycles. Real-time process parameters require high-frequency calling. Daily report-style data can be configured for scheduled triggering. Multiple types of documents correspond to different parsing requirements. Plugins must support parsing multiple formats including structured tables, PDF text, and time-series data. Specific business fields such as crude oil distillation range and product pour point have fixed unit and naming rules. Tool calling must strictly match data source field identifiers to avoid parameter misalignment. The high-risk nature of continuous production scenarios requires tool calling to be configured with failure retry and exception alert mechanisms. This ensures auditing processes are not interrupted.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Workflow/API tool selection | Fixed as functionCall | Adapt for precise calling of structured business data, match the standardized field requirements of refining operations |
| File upload parameters | Support local file and external link upload, allow binding business variables | Adapt to the upload requirements of multiple types of auditing documents in refining operations, support direct submission of shift handover records and material balance sheets |
| Call retry count | 3 times | Respond to network fluctuations during real-time process data calling, ensure continuity of multi-unit load coordinated auditing |
| Call timeout period | 600 seconds | Adapt to the time consumption requirements of multi-unit data integration, avoid interrupting material balance auditing processes due to overly long single calls |
| Plugin cache period | Configure according to data update frequency: 10 minutes for real-time data, 24 hours for daily data | Match the update rhythms of different data sources, reduce invalid calls, ensure timeliness of real-time data |
| Exception alert trigger | Push immediately after call failure | Adapt to the high-risk nature of refining operations, promptly respond to process fluctuations, interlock ledger auditing and other process exceptions |
Three Common Mistakes
- Phenomenon: When calling a custom workflow tool, uploaded file parameters cannot reference preset variables, and only external links are supported. Cause: Variable binding permission was not enabled in the plugin configuration for FastGPT v4.17.0. The current version only supports external link direct upload configuration.
- Phenomenon: When calling a database connection plugin, the interface returns an internal server error. Cause: Database connection parameters do not match the port number and access permissions of the refining business system, or environment variables relied on by the plugin were not correctly configured.
- Phenomenon: Auditing files stored in MinIO cannot be read normally. Cause: MinIO access keys and bucket permissions were not configured, or the correct storage path was not bound in the plugin, resulting in failure to obtain documents such as crude oil evaluation reports.
How to Confirm Configuration Is Complete
- Initiate a tool call, upload a local distillate quality report file, and verify that fields can be correctly parsed and structured results returned.
- View the plugin logs for FastGPT v4.17.0, confirm that the calling interval of real-time process data matches the configured cache period, with no abnormal errors.
- Test the database connection plugin, enter a query statement for the crude oil evaluation system, and verify that correct data such as crude oil density and distillation range can be returned.
- Disconnect the MinIO storage connection, trigger an auditing call for a material balance document, and check whether a preset exception alert is triggered.
The values given are common starting points and should be measured against the user’s own samples.