What the data for this category looks like
The data for this category primarily comes from wafer fab MES systems, equipment FDC collection modules, defect detection and metrology equipment, process recipe libraries, and experimental databases. Daily documentation includes process flow cards, batch run records, equipment recipes, SPC control charts, exception handling sheets, defect maps, and yield reports. Data updates follow the progress of wafer batches: key process parameters are collected in real time, and batch records and reports are generated daily. Document fields include wafer batch numbers, process parameters (units include nanometers, watts, etc.), defect spatial coordinates, equipment operating duration, yield percentage, and more.
What constraints these characteristics impose on the "HTTP Interfaces and External Systems" link
Data for this category is scattered across multiple heterogeneous external systems. HTTP interface integration must cover real-time collected equipment parameters, batch-generated batch records, and static process documents, so adaptation to the authentication rules and data formats of different interfaces is required. The strongly associated attribute of wafer batches requires that interface parameters must carry a unique batch identifier to avoid data confusion. Mixed calls for multiple types of documents need to support simultaneous transmission of structured parameters and unstructured files. At the same time, data anomalies will directly lead to wafer scrap, so interfaces must include idempotent processing and failure retry mechanisms to avoid production accidents caused by repeated submissions or missing data.
How to set up the configuration
This configuration applies to FastGPT v4.17.0.
| Configuration Item | Recommended Approach | Rationale |
|---|---|---|
Authorization | Carry a static token in Bearer format | Meets the general authentication requirements of industrial system APIs, avoids plaintext transmission of secrets |
| Timeout Threshold | 60–180 seconds | Covers the time required for real-time parameter pulling, large-size wafer map transmission, and batch record queries |
| File Upload Limit | 1000 MB | Adapts to the single-file size requirements of wafer maps and defect maps |
responseData | Extract structured business fields under the root node | Unifies the return formats of different external systems, reduces downstream data parsing costs |
| Batch Pull Limit | First 15 batch run records | Controls the data volume per request to avoid triggering interface rate limits |
variables | Preset core parameters such as wafer batch numbers and process nodes | Simplifies parameter passing during interface calls, ensures accurate data association |
Three common configuration errors
- An interface call returns
message: Invalid URL, code: 500. This occurs because the API root path of the external system is not configured correctly, or the carried parameters do not match the routing rules of the target interface. - An interface call triggers a truncated error prompt
Multimodal file size is. This occurs because a reasonable file transfer size limit is not set, and the single-file size of the wafer defect map exceeds the upload limit allowed by the interface. - MongoDB Change Streams disconnect and cannot reconnect in replica set mode. This occurs because automatic reconnection parameters are not configured, and the connection reset logic is not triggered after the primary node drifts, resulting in data stream interruption.
How to confirm the configuration is correct
- Call the test interface to pull a single batch record, check whether the
responseDatafield extracts the expected wafer batch number and process parameters. - After configuring
Authorizationauthentication, call a test interface without permissions, confirm that the correct authentication failure status code is returned. - Submit a large-size wafer map file, verify that the interface can receive it normally and return a valid response.
- Adjust the
variablesparameter, confirm that the preset business variables are automatically included in the interface call.
The values provided are common starting points and should be measured against the reader's own samples.