What data for this category consists of
This category of data primarily comes from ERP orders and inventory, bills of materials (BOMs), process routes, equipment capacity, staff shifts, work-in-progress items, and supplier deliveries. Daily maintained documents include sales orders, master production schedules, material requirements plans, weekly production schedules, work orders, material shortage lists, and delivery change notices. Data update frequencies are daily, weekly, or real-time updates triggered by orders or abnormal events. Data fields include process number, standard working hours, equipment number, planned start and completion time, resource occupancy rate, and some documents require association with material codes and process parameters.
Constraints imposed on tool calling and plugins
Scattered data sources and varied update rhythms require tool calling to adapt to different data source synchronization cycles. For example, pull ERP order data in daily increments, and synchronize work-in-progress data in real time when abnormalities arise. Daily documents vary widely, with differing field formats and association logic. Configure toolChoice to a mode that adapts to multiple scenarios, and keep functionCall enabled to ensure accurate calls. Business processes include complex links such as order decomposition and resource-constrained scheduling. Pass precise parameters including process number, equipment number, and planned time for tool calls, to avoid results deviating from business requirements due to ambiguous parameters.
Configuration settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Workflow/API tool selection | auto | This category's business scenarios cover multiple links such as order decomposition and scheduling adjustment. Automatically selecting tools can adapt to parameter calls for different business needs |
| Workflow/API function-call configuration | enabled | Tool calls must be forcibly triggered to obtain accurate process specifications and scheduling-related data, avoiding deviations from plain text generation |
| Tool call timeout | 600 seconds | Data pulling for this category involves interactions with multiple systems such as ERP and equipment management, requiring sufficient time to complete cross-data source queries and integration |
| Knowledge base recall limit | Top 6–10 entries | Balance the comprehensiveness and response speed of process and scheduling data. Excessive recall results will increase the burden of subsequent logical processing |
| Document type filtering | Group by work orders, production schedules, and process routes | Different documents correspond to different business links. Grouped calls can improve the match between tool returned results and current questions |
Three common configuration errors
- Tool call return formats do not meet expectations, such as unclosed
<tool_call>tags or missing parameter fields. This occurs because fixed parameter mapping rules are not configured for documents such as work orders and production schedules in this category, leading to tool-generated call parameters that do not match target document fields. - Tool calls are not triggered based on business scenarios, such as not calling equipment capacity-related tools in emergency order insertion scenarios. This occurs because rules for triggering corresponding tools based on abnormal events are not configured, and tools are only called according to fixed processes, making it impossible to adapt to temporarily changed business needs.
- Tool call results in the workflow are not properly aggregated, and the final output contains redundant intermediate step content. This occurs because a result filtering link is not configured in the workflow, and intermediate outputs from previous tool calls are not blocked, leading to redundant content being mixed into the final reply.
How to verify a complete configuration
- Initiate a test query that includes order decomposition requirements, and verify that the tool call parameters include category-specific fields such as process number and planned time.
- View the tool call logs in FastGPT v4.17.0, and confirm that the timeout setting meets the actual response duration requirements of the business scenario.
- Trigger a simulated abnormal scenario, such as simulating an emergency order insertion or material shortage, and verify that the corresponding type of tool is called.
- Check the workflow configuration, confirm that the output of intermediate steps has been properly filtered, and only the final aggregated result is retained.
The values provided here are common starting points and should be evaluated against samples relevant to the specific deployment context.