Data for This Category
The data for this category includes packaging flow cards, bonding and molding parameters, wafer incoming material maps, test program versions, yield reports, failure analysis reports, and more. Data sources include wafer test maps, factory MES systems, bonding and molding equipment logs, ATE tester outputs, X-ray/SAM inspection results, and customer specification requirements. Data update frequency adjusts based on production batches, single-unit test results, and daily operations. Different documents have dedicated fields: for example, wafer incoming material maps include wafer IDs and batch numbers, test program versions include version numbers and compatible chip models, yield reports include bin category data, and units are mostly physical dimensions, test voltage, time parameters, and similar metrics.
Constraints on Tool Calling and Plugins
The multi-source heterogeneous data sources, high-frequency update rhythm, and dedicated field characteristics of this category impose multiple constraints on the tool calling and plugins link. Multi-source data covers equipment logs, system reports, and external specifications. Plugins must support access to multi-format data sources to avoid information loss from format conversion. The high-frequency update rhythm requires binding plugin trigger logic to production nodes, ensuring calls retrieve the latest data for the current batch. Dedicated fields such as wafer IDs, bin categories, and test voltage parameters require configuring field mapping rules for tool calling, adapting to structural differences across documents. Plugins must also support non-standard log formats output by equipment.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Workflow/API tool selection | auto | Adapts to the demand for on-demand calling of multiple tools in semiconductor packaging and testing scenarios, no manual tool specification required |
| Tool timeout period | 300–600 seconds | Covers the average time range for equipment data pulling and report generation in this scenario, prevents workflow blocking |
| Field mapping rules | Match dedicated fields such as wafer IDs and bin categories | Ensures that parameters passed during tool calling match the field structure of category data |
| Workflow/API function-call configuration | enabled | Supports structured output of fixed fields for process documents and exception reports, complies with the format requirements of category documents |
| Plugin trigger condition | Bound to production batch completion node | Matches the data update rhythm of each batch in this category, ensures that the latest data for the current batch is called |
Three Common Configuration Mistakes
- After calling a workflow via the interface, running data in the conversation log is empty. The cause is that no field mapping rules are configured, and category-specific data fields are not passed correctly during tool calling, resulting in no valid information in the returned content.
- After a custom plugin runs, the output download address keeps jumping before showing final results. The cause is that the plugin trigger logic is not bound to production nodes, and repeated pulling of outdated data causes repeated rendering. This type of delayed rendering issue is fixed only in FastGPT v4.17.0 and later versions.
- When workflow A calls workflow B, workflow B cannot execute completely. The cause is that workflow B has no configured dependent data source trigger conditions, and no wafer data parameters for the corresponding batch are passed, resulting in execution interruption.
How to Verify Correct Configuration
- Check the tool calling log to confirm returned content includes category-specific fields such as wafer IDs and test bins, and verify fields match preset mapping rules.
- Trigger the plugin, then check if the output process document or exception report includes the latest data for the corresponding batch, and confirm trigger timing matches production node rhythm.
- Test the complete workflow for calling the plugin, confirm all configuration items take effect, and check for no timeout or data missing errors.
- Verify the configuration status of
toolChoiceandfunctionCall, confirm automatic structured output and multi-tool adaptation logic activate during tool calling.
The values provided are common starting points and should be measured against the reader's own samples.