What data for this category looks like
Data for this category primarily comes from factory MES systems, equipment FDC monitoring modules, defect detection and metrology equipment, wafer map databases, process recipe libraries, and experimental datasets. Documents produced daily include process flow cards, batch operation records, equipment recipes, SPC control charts, exception handling sheets, defect maps, and yield reports. Data updates are triggered by each production batch. Equipment parameters are collected in real time. Summary reports are updated daily. Fields include unique wafer identifiers, nanoscale process parameters, batch yield rates, defect spatial distribution coordinates, and other relevant fields.
What constraints these characteristics impose on deployment and upgrades
Data for this category is multi-source, heterogeneous, and spans a wide range of update frequencies. During deployment, access protocols for different data sources such as MES and equipment FDC must be adapted. Custom field parsing must be supported to accommodate special fields like nanoscale process parameters and defect coordinates. High data continuity is required during production. During the upgrade phase, incremental updates must be supported to avoid interrupting data flow for active wafer batches. Exception report generation must align with production cycle requirements. During deployment, concurrent connection counts and response delays must be limited to prevent disruption to normal production line operations. A large volume of high-dimensional process and defect data also requires specialized storage configurations to avoid insufficient basic storage resources.
How to set configurations
The following configurations are formulated based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
DB_MAX_LINK | 100–150 connections | Adapts to multi-source data access and high-concurrency process data query requirements, and prevents database connection exhaustion |
MONGODB_URI | Exclusive cluster connection string with authentication information | Stores unstructured high-dimensional data such as wafer defect maps and batch operation records, and requires a dedicated cluster to ensure read and write performance |
PG_URL | Read-only instance address associated with the business database | Stores structured business data such as process flow cards and yield reports. A read-only instance avoids write conflicts |
REDIS_URL | Local cache cluster connection address | Caches real-time collected equipment FDC parameters and temporarily generated exception report drafts to reduce repeated query delays |
STORAGE_S3_ENDPOINT | Enterprise-specific object storage node address | Stores large files such as wafer maps and process recipe libraries to meet bulk data storage requirements |
ROOT_KEY | Randomly generated 32-character string | Ensures permission security for internal system interface calls, and prevents unauthorized access to process data configurations |
Three common configuration errors
- Symptom: Interface call failures occur after connecting to internal enterprise communication tools, with 403 status codes returned in logs. Cause:
FILE_TOKEN_KEYis not configured, causing request signature verification to fail and triggering platform security interception. - Symptom: Historical wafer batch data cannot be loaded normally after cross-version upgrade. Cause: Upgrade scripts for intermediate versions are not executed, and direct overwriting deployment leads to incompatible data structures.
- Symptom: Exception reports cannot be generated normally after configuring a local large model service. Cause: The model service address is not configured correctly, and forced binding to public interfaces causes restricted access to the local network.
How to confirm the configuration is complete
- Run the database connection test script to verify that the connection count set in
DB_MAX_LINKcan stably remain within the specified range. - Upload a wafer map file to verify that the storage node configured in
STORAGE_S3_ENDPOINTcan complete upload and download operations normally. - Trigger an exception report generation task and check whether the cache node configured in
REDIS_URLhas cached the temporarily generated draft data. - Log in to the system backend to view permission logs, and confirm that the verification rules configured in
ROOT_KEYhave not intercepted legitimate internal interface calls.
The values provided are common starting points and should be measured against the reader's own samples.