What Data for This Category Looks Like
This category covers the full semiconductor packaging workflow. Data sources include wafer test maps, MES systems, bonding and molding equipment, ATE test machines, X-ray/SAM inspection equipment, and customer specification documents. Daily documents include packaging process cards, bonding and molding parameter tables, wafer incoming material maps, test program version files, yield reports, and failure analysis reports. Data updates occur once per production batch, once per individual chip test, and once daily. Fields cover batch number, bonding power, molding pressure, test bin code, wafer coordinates, and more. No unified fixed format exists, so deployments must adapt to multi-source heterogeneous data characteristics.
Constraints Imposed by These Characteristics on Tool Calling and Plugins
Multi-source heterogeneous data sources require plugins to support connection to multiple different API endpoints, and require configuration of unified field mapping rules to avoid data format incompatibility between different devices and systems. High update frequency requires tool calling to support both scheduled triggering and event triggering modes. These modes adapt to daily summary and real-time ledger updates after a single production batch is completed, respectively. Diverse document types require plugins to handle table, text, and image inputs, and adapt to unstructured data such as wafer incoming material maps. Severe consequences of quality errors require tool calling to configure failure retry and alert mechanisms, to avoid mixed batches, incorrect programs, and other issues caused by data anomalies.
How to Set Configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Workflow/API tool selection | auto | Adapts to multi-type plugin calls, automatically selects appropriate tool execution logic to cover calling requirements for different data sources |
| Workflow/API function-call configuration | {"name":"get_production_data","parameters":{"batch_id":"${var_batch_id}"}} | Explicitly specifies calling the production data acquisition plugin, binds batch variables passed in the workflow, and ensures precise calling targets |
| Plugin trigger timing | 2:00 AM daily + within 10 minutes after each production batch is completed | Matches the update frequency of category data once daily and per batch, ensuring real-time ledger data |
| Parameter fallback rule | Prioritize variables passed in the workflow; use plugin preset default values when variables are missing | Adapts to dynamic parameter transfer scenarios, avoids calling failures caused by missing variables |
| Timeout setting | 300 seconds | Covers the average time required to pull data from multiple devices, avoids long-term workflow blocking |
| Retry count | 2 times | Addresses temporary network fluctuations or short-term device offline status, reduces invalid calling attempts |
Three Common Misconfigurations
- Symptom: A cross-origin error with status code 403 is returned when calling the ATE test machine data plugin. Cause: The cross-origin whitelist for the corresponding device API was not added in the FastGPT plugin configuration.
- Symptom: Some fields are empty after calling the production ledger generation plugin in the workflow. Cause: The parameter fallback rule was not configured correctly, variables passed in the workflow are missing, and plugin default values are not enabled.
- Symptom: Plugin calling times out, and the workflow is marked as failed. Cause: The timeout setting duration is shorter than the total time required to pull data from multiple devices, and does not match the characteristics of multiple category data sources and multiple pulling steps.
How to Verify Correct Configuration
- Manually trigger a plugin call, check if the returned results include category-specific fields such as packaging process cards and test bin distribution.
- Check the cross-origin whitelist in the plugin configuration, confirm that API domains of all connected devices and systems have been added.
- Simulate a scenario where variables are missing, verify whether the plugin automatically uses preset default values to complete the call.
- View workflow logs in FastGPT v4.17.0, confirm that the plugin calling timing matches the data update frequency.
The values provided are common starting points and should be measured against samples from the actual deployment.