What data for this category looks like
Data for this category includes packaging process cards, bonding and molding parameters, wafer incoming material maps, test program versions, yield reports, failure analysis reports, and other documents. Data sources include wafer test maps, MES systems, bonding and molding equipment, ATE test machines, X-ray/SAM detection equipment, and customer specification requirements. Data updates follow production batches, single-chip test results, and daily production progress. Fields include batch number, test bin value, failure type, equipment operating parameters, and other items. Format adjusts dynamically based on production links.
What constraints these characteristics impose on workflow orchestration
Since data updates follow production batches, single test results, and daily progress, workflows must support triggering execution by production batch, and associate multi-source heterogeneous production data. Differences in data formats across multiple devices and systems require built-in data format verification nodes in workflows, to prevent process interruptions caused by format mismatches. The need to align wafer incoming material maps with physical batches requires configuring cross-data source association nodes in workflows, to ensure consistency of batch information. The analysis requirements for test bin distribution and physical failures require workflows to support parallel processing of different types of detection data across multiple nodes, and reserve custom field mapping links to adapt to differences in customer specifications.
Configuration details
This configuration is based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale for This Setting |
|---|---|---|
Global variable variables | Set type to knowledge base selection, bind packaging process card and yield report knowledge bases, return top 5 results per single retrieval | Match the requirement to associate multiple types of production documents for this category, adapt to the analysis scale of single batch data volume |
| Conditional judgment node | Configured to match wafer incoming material maps and MES system data by batch number | Resolve the core time-consuming link of aligning wafer maps and physical batches |
nodeId | Assign a unique identifier to each workflow node | Facilitate quick location of abnormal process nodes, adapt to abnormal troubleshooting requirements |
| Node timeout setting | 600 seconds, retry 2 times | Adapt to data collection delays of ATE test machines and X-ray/SAM equipment, avoid unnecessary process interruptions |
| Failure handling strategy | Trigger an alert and record the batch number to the system log, log storage limit is 1000 MB | Correspond to the business requirement of quickly locating abnormal batches after an error, avoid storage resource overflow |
| Parallel node configuration | Start test bin analysis and physical failure analysis nodes simultaneously | Optimize execution efficiency of multi-dimensional analysis |
Three common configuration mistakes
- When configuring the custom
variablestype as knowledge base selection, failing to configure field matching rules in the conditional judgment node causes the judge to fail to trigger the corresponding branch, and the interface displays invalid conditional configuration. The cause is failure to map the knowledge base fields bound by the global variable to the matching items of the judgment node. - Not configuring a timeout fallback node for the workflow. When data collection by the ATE test machine times out, the workflow terminates directly, and the log shows the node status as incomplete with no subsequent execution records. The cause is failure to set timeout retry or exception jump logic, corresponding to the common problem of workflow execution ending halfway.
- Failing to verify the uniqueness of
nodeId. When publishing the workflow to a Feishu bot, multi-instance node conflicts cause failure to receive user requests, and the bot returns a connection failure prompt. The cause is failure to assign unique identifiers to each node according to specifications.
How to verify a complete configuration
- Enter the
variablesconfiguration page of the workflow, confirm that the knowledge base selection type variable has been bound to the corresponding production document knowledge bases such as packaging process cards and yield reports. - Trigger a simulated workflow execution with batch data, check whether the log displays successful batch number matching, and all node execution statuses are completed.
- Manually simulate a node timeout scenario, confirm that the workflow triggers an alert and records the batch number to the system log.
- Check the
nodeIdlist, confirm that there are no duplicate identifiers for all workflow nodes.
The values provided are common starting points and should be measured against the reader's own samples.