The following configurations are formulated based on the tool calling logic of FastGPT v4.17.0.
What Data for This Category Looks Like
The data for general mechanical equipment assembly scenarios mainly comes from PLM/ERP, part inspection data, assembly tool collection systems, test benches, MES, and customer technical agreements. Data updates are triggered per single equipment or per order, and include documents such as assembly process cards, multi-level BOMs, torque records (unit: N·m), gap adjustment records (unit: mm), commissioning reports, factory inspection sheets, packing lists, etc. Document fields include material codes, assembly processes, measured parameters, qualified judgment results, etc. Core fields vary across different documents.
Constraints Imposed on Tool Calling and Plugins
The multi-level BOM structure of this category requires tool calling to support layered traversal and associated queries, to avoid missing assembly requirements for lower-level materials. The demand for accessing multiple data sources requires tool configuration to be compatible with API formats of different systems, adapting to authentication and return rules of different interfaces such as PLM and MES. The high-frequency update rhythm per single equipment or order requires the trigger logic of tool calling to match the data pull cycle, to avoid synchronization delays. The field differences across different documents require parameter mapping by document type during tool calling, to prevent aggregation errors caused by field mismatches.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Workflow/API tool selection | Fixed selection of 2 preset tools | This scenario requires a fixed tool calling scope to avoid resource conflicts caused by concurrent calls to too many tools |
| Workflow/API function-call configuration | 10-second batch call interval | The assembly records and BOM data volume per order is large; a 10-second interval avoids interface rate limiting |
| Tool calling context length | 8000–12000 token | The total character count of multi-level BOMs and assembly documents is high, requiring adaptation for long context processing |
| Single tool call return limit | Top 5 results | Core ledger data is concentrated in the first 5 records; too many results increase aggregation time |
| Tool call failure retry count | 3 times | Interfaces for assembly data occasionally experience fluctuations; retries reduce failure rates |
Three Common Misconfigurations
- Phenomenon: Concurrent requests occur during tool calling, returning a 503 Service Unavailable status code. Cause: The tool selection scope is not fixed via the
toolChoiceconfiguration, allowing simultaneous calls to multiple tools and exceeding system concurrency limits. - Phenomenon: Ledger data returned after API calls is empty, and content from documents such as assembly process cards and torque records cannot be read. Cause: No data source mapping rules for tool calling are configured, so tools cannot associate with corresponding assembly documents in PLM or MES.
- Phenomenon: A connection error is reported when calling tools, and logs show failure to connect to the data source interface. Cause: API authentication parameters for tool calling are not correctly configured, leading to failure to connect to data source interfaces such as PLM and MES.
How to Confirm Configuration Completion
- Trigger a single-order ledger aggregation task, and check if only the 2 preset tools are called in the FastGPT tool calling logs.
- Verify the number of results returned by tool calls, confirming they meet the single tool call return limit configuration requirements.
- Simulate an interface connection failure scenario, checking if the preset retry mechanism is triggered, matching the configured failure retry count.
- Compare the fields returned by tool calls with the fields of actual assembly documents, confirming correct parameter mapping.
The values provided here are common starting points and should be evaluated against the reader's own samples.