What This Category of Data Looks Like
Ledger data for semiconductor wafer fabrication comes from MES systems, equipment FDC monitoring tools, defect detection and metrology equipment, wafer map files, process recipe libraries, and experimental data. Daily ledger documents include process flow cards, batch operation records, equipment recipe parameters, SPC monitoring charts, exception handling sheets, defect atlases, and yield-related records.
Data update cadences cover each wafer batch transfer node, real-time equipment parameters, and daily summary statistics. Core fields include wafer batch ID, process temperature, equipment runtime duration, defect position coordinates, and more. Corresponding units include degrees Celsius, hours, nanometers, and more.
What Constraints These Characteristics Impose on the "Workflow Orchestration" Link
Because data sources are scattered and have varying formats, the workflow must be configured with multiple independent nodes to pull ledger data from different systems separately. Multiple update cadences require the workflow to support mixed trigger logic: batch-based triggering, real-time synchronization, and scheduled summary.
Ultra-long process chains and rework paths require the workflow to support tracing ledgers by process nodes and associating cross-batch exception handling records. Nanometer-level precision parameters require strict matching of field units between workflow nodes to avoid data confusion.
Cross-departmental ledger summary requires a unified output format, so the workflow must include a data format conversion node to adapt to report template requirements of different departments.
How to Set the Configurations
All configurations below are based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Trigger Timing Settings | Trigger after each wafer batch completes transfer | Matches the business logic that ledger updates for wafer manufacturing are centered on batch units |
variables Mapping Rules | Map wafer batch ID and process parameters to globally shared variables | Meets the requirement for sharing batch-level ledger data across multiple nodes |
nodeId Binding Rules | Bind unique identifiers corresponding to MES and equipment FDC for each data source node | Distinguishes ledger data from different sources to avoid data confusion between nodes |
| Timeout Threshold Setting | 600 seconds | Covers the reasonable time range for multi-process data pulling |
| Node Retry Count | 3 times | Addresses temporary disconnections or interface fluctuations when pulling equipment data |
| Log Output Configuration | Store to the workflow execution log directory | Facilitates troubleshooting of log issues in code execution modules |
Three Common Mistakes
- The console.log logs from the code execution module in the workflow cannot be viewed in the frontend inspection console. This occurs because FastGPT workflow logs are stored in a server-specified directory by default and not synchronized to the frontend inspection panel.
- The workflow reports an error during execution, prompting failure to pull ledger data. This occurs because access permissions were not correctly configured for the data source node bound to
nodeId, resulting in inability to connect to the MES or equipment FDC system. - The AI model dropdown list for the problem classification node is empty. This occurs because the API key for the corresponding large model was not configured on the platform, or the model call permission for the workflow was not enabled.
How to Confirm the Configuration Is Complete
- Manually trigger a workflow execution, confirm that all nodes have no errors and data pulling is completed.
- Verify that the fields mapped via
variablescover the core ledger items of the current batch, including wafer batch ID, process parameters, and more. - Check that the data source node bound to
nodeIdcan normally return ledger data from the corresponding system. - Review the report generated by the workflow to confirm it matches the structural requirements of the preset process flow cards and exception handling sheets.
The values provided here are common starting points and should be measured against the reader's own samples.