What this type of data entails
Core data for semiconductor packaging and testing scenarios covers multiple categories of documents and real-time device data. Daily documents include packaging process cards, bonding and molding parameter tables, wafer incoming material drawings, test program version files, yield reports, and failure analysis reports. Data sources include wafer test maps, MES production systems, bonding and molding equipment, ATE test machines, X-ray/SAM detection equipment, and customer specification requirements. Data update frequency adjusts dynamically based on production batches, individual chip test results, and daily equipment operating status, with no fixed cycle. Document fields include batch numbers, process parameter thresholds, test bin codes, failure types, and some drawing documents include geometric dimension and position annotations.
What constraints these characteristics impose on workflow orchestration
Data in this scenario is distributed across multiple types of equipment and systems. Update frequency changes dynamically based on production nodes. Workflows must support real-time connection to and on-demand pulling of multiple data sources. Document structures vary significantly across different types. Some are plain text parameter tables, while others include wafer maps with visual annotations. Workflows must adapt to multi-format document parsing and retrieval. Process and test data for a single production batch have a strong correlation. This requires strictly limiting the retrieval scope to documents associated with the current batch, to avoid cross-batch data interfering with Q&A results. Additionally, incorrect process parameters or misaligned data can directly cause packaging failures or defective products to be released. Workflows must include verification and alert links to intercept abnormal data promptly.
Configuration settings
This configuration is based on FastGPT v4.17.0.
| Configuration Item | Recommended Approach | Rationale |
|---|---|---|
variables | Assign values grouped by scenario, using the format ["wafer_batch", "bond_temp", "test_program_version"] | Packaging and testing requires distinguishing three core variable categories: batch information, process parameters, and test programs, to avoid variable conflicts |
| Data source trigger node | Bind to each batch synchronization event of the MES system, start after a delay of 30 seconds | Production data for each batch must be synchronized promptly to avoid process delays caused by lag |
| Document retrieval scope | Limit to wafer incoming material drawings and packaging process cards associated with the current batch, return the top 3 matching results | Only documents corresponding to the current batch are needed for a single batch; excessive results will interfere with the AI's ability to generate accurate Q&A responses |
| Failure retry count | 2 retries, with a 10-second interval between retries | ATE test machines or X-ray equipment may occasionally go offline; retries can restore normal process flow |
nodeId | Bind to the unique identifier of the yield analysis node | Precise triggering of variable assignment for the failure analysis link is required to avoid confusion across nodes |
| Export configuration | Name exported files by batch number, bind the storage path to a specified directory | Packaging and testing requires retaining process Q&A and analysis results for each batch to facilitate traceability |
Three common errors
- Phenomenon: When calling the FastGPT API, the final variable value in the output overlaps with the result of the previous variable. Cause: The
variablesarray was not initialized, causing the workflow cache to not be cleared and historical variable data to be reused repeatedly. - Phenomenon: The workflow cannot be imported normally after export. Cause: Associated data source configuration items were not selected during export, resulting in missing device connection parameters required by the workflow.
- Phenomenon: When using the API link of the release channel to assign values to global variables, the interface returns empty fields. Cause: The array format of the
variablesparameter was not correctly passed in the API request, causing the variables to fail to be recognized.
How to confirm the configuration is complete
- Manually trigger the workflow once, check if the return value of the
variablesarray matches the current batch's process parameters and batch number. - Call the API of the release channel to access the workflow link, verify that the returned Q&A results only include document content associated with the current batch, with no cross-batch data included.
- Simulate an ATE test machine offline scenario, confirm that the workflow triggers 2 retries before entering an abnormal alert state.
- Export the workflow configuration file, check that all bound data sources and variable definitions are included, with no missing items.
The values provided here are common starting points and should be measured against the reader's own samples.