What data for this category looks like
Data for this category primarily comes from CAD/CAM software, order BOMs, sheet metal inventory systems, laser cutters, bending machines, and workshop MES systems. Daily outputs include unfolding drawings, CNC cutting programs, bending process cards, nesting diagrams, first-piece records, and process flow cards. Data updates are triggered by order generation, daily inventory checks, or equipment model changes. Fields include sheet metal thickness, bending compensation coefficient, cutting size, bending angle, process flow number, order association identifier, and others. Units are mostly millimeters, degrees, and square meters.
What constraints these characteristics impose on tool calling and plugins workflows
The multi-variety, small-batch business characteristics require tool calling to support dynamic loading of order-specific parameters, to prevent generic template adaptation failures. The need to connect multiple data sources requires plugins to be compatible with different data structures from CAD/CAM export formats, MES process fields, and sheet metal inventory ledgers. Data updates triggered by single model changes require tool calling contexts to support refreshing at model change nodes, to prevent old data from interfering with summary results. The drawing-driven business logic requires plugins to parse core fields from unfolding drawings and bending process cards, to ensure ledger fields match business requirements. Errors directly lead to sheet metal waste and assembly errors, so core field integrity must be verified before tool calling.
How to configure settings
This configuration is based on FastGPT v4.17.0. Specific configuration items are as follows:
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Workflow/API tool selection | Auto | Adapts to dynamic tool calling needs for sheet metal cutting and bending scenarios, no manual tool type specification required |
| Tool call timeout | 120 seconds | Matches typical time ranges for CAD drawing parsing and equipment data retrieval |
| Plugin data source mapping rules | Map separately by order BOM, MES process, and sheet metal inventory | Corresponds to business characteristics of multi-source data access |
| Single plugin call limit | First 8 entries | Matches process flow record count ranges for a single order |
| Workflow/API function-call configuration | Enable parameter validation | Prevents invalid field inputs from causing ledger summary errors |
Three common configuration mistakes
- Symptom: Ledger data generated after calling historical records is duplicated, including process records from multiple orders. Cause: Context is not configured to refresh by order or model change node, so old historical data is not cleared.
- Symptom: Retries after call timeout still fail, returning status code 504. Cause: Tool call timeout is set too short, not matching actual data retrieval times for laser cutters.
- Symptom: Tool calling cannot be triggered after local model deployment, returning the prompt "No available tools found". Cause: The
functionCallparameter is not enabled in the model configuration, so the model cannot recognize tool calling instructions.
How to verify a correct configuration
- Import a standard sheet metal unfolding drawing, trigger the ledger generation tool, and check whether the generated ledger includes the bending coefficient and cutting size fields.
- Simulate a model change operation, trigger tool calling, and confirm that the context refreshes automatically, and old order data is not mixed into the new ledger.
- Check system logs to confirm that the
toolChoiceparameter is called correctly, and the tool calling instruction is recognized and executed by the model. - Test local model access, confirm that after
functionCallis enabled, the tool calling prompt returns normally.
The values listed here are common starting points, and should be evaluated against samples specific to the deployment environment.