What this category of data looks like
The data sources for this category cover ERP orders and inventory, BOM, process routes, equipment capacity, staff shifts, work in progress, and supplier arrivals. Daily involved documents include sales orders, master production schedules, material requirements plans, weekly and daily scheduling tables, work orders, material shortage lists, and delivery change notices. Fields include order number, process number, equipment ID, material code, planned capacity, actual working hours, delivery date, and others. Units include pieces, hours, days, and machine runs. Data update frequency is daily, weekly, or triggered when orders or exceptions occur.
What constraints these characteristics impose on the "tool calling and plugins" link
Since the data sources for this category cover multiple systems and update rhythms are not fixed, tool calling must support both on-demand triggering and scheduled synchronization modes. Data contains multi-dimensional resource constraint information, so plugins must accurately match field mappings to avoid data misalignment. A single call may involve a large number of work orders and scheduling data, so the number of returned entries per call must be limited to prevent timeouts. Cross-departmental business scenarios require that tool calling results be compatible with the permission ranges of different roles to ensure data access compliance.
How to set up the configuration
The following configurations are based on the FastGPT v4.17.0 plugin system:
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
| Tool selection mode | toolChoice | Allows the system to independently determine whether to call a tool, adapts to dynamic scheduling and inspection scenarios, and meets business flexibility requirements |
| Tool calling format | functionCall | Requires tool calling results to conform to standard function call formats, facilitating parsing of scheduling and equipment data and ensuring actionable results |
| Tool triggering timing | Triggered by order updates | Category data is updated when orders or exceptions occur; on-demand triggering reduces invalid calls and lowers system resource consumption |
| Number of data entries per call | First 20 entries | Scheduling and work order data volumes are large; limiting the number of returned entries avoids interface timeouts and ensures smooth conversation experience |
| Timeout wait duration | 600 seconds | Pulling equipment data and recalculating scheduling may require long processing times; this duration covers most business scenarios |
| Data source synchronization cycle | Every 12 hours | Daily cycle is used for daily category data updates; scheduled synchronization ensures data timeliness and aligns with business update rhythms |
Three common mistakes
- Phenomenon: Calling the equipment inspection interface returns empty fields or incorrect formats. Cause: No mapping relationship between the data source and category data fields is configured, resulting in failure to match key fields such as order number and equipment ID.
- Phenomenon: The first conversation wait exceeds 3 seconds, and subsequent conversations proceed at normal speed. Cause: The plugin data caching mechanism is not enabled; the first call needs to pull all scheduling and work order data, increasing response time.
- Phenomenon: Calling the internal ERP system via the HTTP plugin fails, and access via the proxy is not possible. Cause: No proxy parameters are added in the plugin configuration, making it impossible to penetrate the internal network to access enterprise data sources.
How to confirm the configuration is correct
- Initiate a test query containing the keywords "equipment inspection" and "shop scheduling", check whether the plugin returns matching work orders and equipment capacity data.
- View the plugin logs to confirm that the data source synchronization cycle conforms to the configured 12-hour triggering rule.
- Simulate an order insertion scenario, trigger tool calling, and check whether scheduling data can be updated in real time and corresponding results returned.
- Test the interface call of the internal ERP system, confirm that the proxy configuration takes effect and the data source can be accessed normally.
The values provided are common starting points and should be measured against the reader's own samples.