What the data for this category looks like
Data related to inspection and process record audit for semiconductor wafer fabrication comes from MES systems, equipment FDC data, defect detection and metrology equipment, wafer map files, process recipe libraries and experimental datasets. Data update cadences include batch records generated per wafer lot, real-time synchronized equipment operating parameters, and daily aggregated yield reports. Daily documents include process flow cards, lot run logs, equipment recipe parameters, SPC statistical charts, exception handling sheets, defect maps, and more. Fields include lot identifiers, process node numbers, equipment operating values, defect coordinates, handling order numbers, and more. Some parameters use nanometer-scale units.
What constraints these characteristics impose during deployment and upgrade
Multi-source heterogeneous data sources require configuring parameters adapted to multi-system integration during deployment to avoid data synchronization interruptions. High-frequency update cadences require matching read/write performance of storage and cache to real-time and lot-level data write requirements to prevent data backlog. Diverse document types require presetting different parsing rules to handle unstructured documents such as wafer maps and SPC charts. The precision requirements of nanometer-scale process parameters require calibrating the matching degree of embedded models during deployment to ensure the accuracy of parameter interpretation. The upgrade phase must be compatible with historical lot record formats to avoid data format conflicts between old and new versions leading to record loss.
How to set configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
MONGODB_URI | mongodb://fabuser:SecurePass123@192.168.1.20:27017/fab_inspect?authSource=admin | Adapts to storage requirements for multi-lot, high-concurrency historical inspection records |
REDIS_URL | redis://:RedisPass456@192.168.1.21:6379/0 | Supports cache read/write for real-time synchronized equipment FDC data and SPC charts |
STORAGE_S3_ENDPOINT | http://192.168.1.22:9000 | Stores large-volume documents such as wafer maps and defect maps |
DB_MAX_LINK | 200–300 connections | Matches the connection upper limit for concurrent writes from multiple data sources |
FILE_TOKEN_KEY | Randomly generated 32-bit string | Ensures access permission security for wafer lot records and exception handling sheets |
Three common errors
- A
workflow error {"message":"Dangerous behavior"}error occurs during workflow debugging. The cause is failure to configure a security whitelist for local data sources, triggering risk interception for cross-system data access. - After upgrading to FastGPT v4.17.0, a rule trace symbol appears at the end of model conversation response results. The cause is failure to disable the system default context trace configuration item.
- Connection failure occurs when attempting to use a locally deployed m3e-large model as an embedding tool. The cause is failure to correctly specify the local model's API access address and port in the configuration.
How to confirm correct configuration
- Run a multi-source data synchronization test to verify that data from MES systems and equipment FDC can be properly imported into the system and generate complete lot records.
- Upload documents such as wafer maps and SPC charts to confirm that parsed content can be normally retrieved and associated with corresponding process parameters.
- Initiate an inspection record audit workflow to confirm that no
Dangerous behaviorerror messages appear. - Initiate a model conversation test to confirm that no additional rule trace symbols appear in response results.
The values provided are common starting points and should be assessed against the reader's own samples.