What this category of data looks like
This category of data covers daily production support documents for manufacturing enterprise equipment departments, workshop maintenance teams, and public utility departments. Documents include equipment ledgers, inspection checklists, maintenance plans, work orders, fault analysis, spare parts lists, and special equipment inspection records. Data sources include EAM/CMMS systems, equipment sensor data, operator inspection records, maintenance history, spare parts inventory, and statutory inspection information. Update frequencies cover shifts, daily/weekly/monthly cycles, and real-time triggers when faults occur. The core associated field is equipment number. Inspection checklists include inspection points, normal/abnormal status, and inspection time fields. Work orders include fault descriptions, maintenance measures, and replaced spare part code fields.
What constraints these characteristics impose on the tool calling and plugins workflow
Multi-source data access requires tool calling to pull data across systems, rather than relying solely on static data from a single knowledge base. Frequently updated equipment sensor and inspection data requires tool calling triggers to be tied to production schedules and fault events. Fixed-cycle data pulling cannot adapt to real-time data requirements. Diverse document structures require tool calling parameter parsing to adapt to different field formats. For example, spare parts lists must match spare part code fields to quickly locate replacement parts. The timeliness of equipment fault response requires tool calling timeout times to be controlled within reasonable ranges, to avoid delaying maintenance scheduling. This page configuration was completed based on FastGPT v4.17.0.
How to set the configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Workflow/API tool selection | auto | Adapts to tool calling needs across multiple scenarios including equipment inspection, fault troubleshooting, and spare parts query |
| Workflow/API function-call configuration | enabled | Allows calling preset tools to obtain real-time EAM/CMMS data and inspection records |
| Single tool call timeout | 30–60 seconds | Matches the time requirements for equipment fault response in production scenarios, avoids delaying maintenance |
| Maximum number of tool call return results | Top 5–10 entries | Adapts to the standard number of information entries in inspection checklists and work orders, controls context length |
| Tool call retry count | 2 times | Avoids business interruptions caused by single call failures, while controlling resource consumption |
| Conversation context retention duration | 24–48 hours | Covers the work cycle of daily equipment inspection and maintenance for teams, avoids loss of key operation records |
Three common configuration errors
- Symptom: Plugin button clicks have no response, tool calling is not triggered. Cause:
functionCallis not configured asenabled, or API key binding verification for the corresponding tool is not completed. - Symptom: Tool calling is triggered continuously outside working hours, leading to excessive API balance consumption. Cause: No time filtering rules for tool calling are set, and triggering is not limited to production scheduling hours.
- Symptom: Plugin selection prompts pop up multiple times in a single conversation. Cause: The configuration for only calling one plugin per conversation is not enabled, or
toolChoiceis not set toautoand call times are not limited.
How to confirm the configuration is correct
- Access the test conversation interface, input equipment inspection-related questions, and verify if corresponding tool calling options pop up automatically or if tool query results are returned directly.
- Set time filtering rules, initiate test requests outside production hours, and confirm that tools are not triggered.
- Initiate a single-round conversation, then initiate another conversation on the same topic, and check if plugin selection prompts pop up only once.
- Switch to the standard conversation flow, input equipment-related questions, and confirm that content from the equipment inspection knowledge base is not returned.
The values provided are common starting points and should be evaluated against individual deployment samples.