What the data for this category looks like
Daily documents for this category include packaging process cards, bonding and molding parameters, wafer incoming material drawings, test program versions, yield reports, and failure analysis reports. Data sources include wafer test maps, manufacturing execution systems (MES), bonding and molding equipment, ATE test machines, X-ray/SAM inspection equipment, and customer specification requirements. Updates occur per production batch, per individual chip test cycle, and daily. Documents must link to fields such as wafer batch numbers, equipment parameter thresholds, test bin codes, and yield values. Some drawing documents require matching physical coordinate information for the physical batch.
What constraints these characteristics impose on workflow orchestration
Data for this category is multi-source and heterogeneous. It includes real-time data directly from equipment, structured records from MES systems, static drawings, and unstructured test reports. Workflows must support unified access to multiple data sources and format adaptation. The high-frequency update requirement means workflow trigger nodes must bind to precise timing points such as production batches and individual chip tests. This prevents redundant execution and expired data. Significant structural differences exist across document types. Dedicated parsing and storage rules must be configured for drawing, parameter, and report documents respectively. Strong field-associated business logic requires variable passing within workflows to strictly match core identifiers like batch numbers and test bin codes. This prevents cross-batch data mixing from causing production accidents.
The following configurations are based on the interface and parameter logic of FastGPT v4.17.0.
How to set the configurations
| Configuration Item | Recommended Approach | Rationale |
|---|---|---|
| Form input node default value | Read the global production batch number from variables | Match the core business identifier, automatically bring in the current production batch information, and reduce manual input errors |
| Tool and MCP call diversion rules | Static documents use knowledge base calls, real-time equipment data uses MCP calls | Adapt to the update frequency and storage characteristics of the two types of data, reuse knowledge base for static parameters, directly connect to equipment for real-time data |
| Single knowledge base recall upper limit | Top 8 entries | Cover the core bin code distribution data for a single batch test, avoid redundant information interfering with cross-review |
| Workflow node timeout setting | 600 seconds | Adapt to the data return cycle of ATE test machines and X-ray/SAM equipment, avoid workflow interruptions caused by response delays |
nodeId binding configuration | Bind the nodeId of the form input node and knowledge base call node | Ensure that variables such as batch numbers are correctly passed across nodes, preventing data mixing |
| Document upload supported formats | PDF, PNG, CSV, JSON | Cover common document types for this category, including packaging process cards, wafer incoming material drawings, test program versions, and yield reports |
Three common errors
- Phenomenon: The default value of the form input node fails to load the global variable correctly, and the interface displays empty. Cause: The global variable for the production batch number is not pre-configured in
variables, or the variable reference path is not correctly associated. - Phenomenon: The workflow frequently triggers redundant execution, creating duplicate document review tasks. Cause: The workflow trigger timing is not bound to precise nodes such as production batches and individual chip tests. Instead, it uses fixed-time polling, which does not align with the high-frequency update business characteristics.
- Phenomenon: Uploaded wafer incoming material drawings cannot be parsed correctly, and drawing comparison data is missing in the review link. Cause: The PNG format document parsing rule is not configured, and only text-based document formats are supported.
How to verify a complete configuration
- Submit test data for a simulated production batch, and confirm the form input node automatically pulls in the preset global batch number.
- Upload typical documents of this category, trigger knowledge base calls and MCP calls separately, and confirm the diversion rules take effect and corresponding nodes execute normally.
- View workflow running logs, and confirm variables such as batch numbers and test bin codes in
variablesare passed correctly across nodes. - Trigger a node timeout test, and confirm the workflow handles delays up to the set duration as expected.
The values listed above are common starting points, and should be evaluated against the reader's own samples.