What the data for this category looks like
Data for inspection and process record audits in this category comes from electromagnetic design databases, copper wire and insulation material batch ledgers, winding equipment, balancing machines, electrical test benches, and ERP/MES systems. Daily generated documents include winding process cards, insulation treatment records, dynamic balance reports, dielectric strength test reports, and type and factory inspection reports. Fields cover winding turns, wiring method, impregnation temperature, vibration amplitude, temperature rise value, dielectric strength value, and more. Units include turns, ℃, mm/s, MPa, and others. The update frequency is once per finished product or production batch. New records and documents are also generated when product type changes.
What constraints do these characteristics impose on deployment and upgrades
Data sources for this category cover industrial equipment, business systems, and unstructured documents. Deployment requires compatible multi-protocol data source access and unstructured document parsing. High-frequency single-product or batch-level updates require deployed databases and cache services to support high-concurrency connections and incremental synchronization. Multiple types of specialized documents and fields require custom field mapping rules to ensure the audit link can accurately extract key parameters. During upgrades, ensure the association relationships of historical batch data are not lost. Adapt to database structure changes between different versions to avoid inspection records becoming unqueryable after upgrade. Additionally, some users deploy on Huawei all-in-one machines, so adaptation to the all-in-one's internal network policies and built-in service connection configurations is required.
How to set the configuration
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
MONGODB_URI | When deploying on a Huawei all-in-one machine, use the connection address of the built-in MongoDB instance, formatted as mongodb://<user>:<pass>@<host>:27017/fastgpt?authSource=admin | Adapts to storage requirements for multi-batch inspection data. The Huawei all-in-one built-in instance ensures low-latency local access |
REDIS_URL | redis://<host>:6379/0, set connection timeout to 300 seconds | Supports high-frequency inspection record queries and session caching. A 300-second timeout adapts to the execution cycle of batch audit tasks |
DB_MAX_LINK | 100–150 | Matches concurrent connection requirements for multi-device data access, avoids database connection exhaustion, and complies with FastGPT v4.17.0 version requirements |
STORAGE_S3_ENDPOINT | Internal access address of the Huawei all-in-one machine object storage service | Enables local storage of large files such as inspection reports, reducing cross-network transmission latency |
FE_DOMAIN | Internal domain name or fixed IP of the deployment node | Ensures normal interface calls and file downloads between internal systems |
Three common errors
- Phenomenon: Database connection timeout occurs when deploying on a Huawei all-in-one machine, with logs showing
connection refused. Cause: The built-in database instance address of the all-in-one machine was not used, and a public network connection was directly used, resulting in interception by the internal network policy. - Phenomenon: Winding turns data for historical batches is lost after upgrading from v4.8.10 to v4.9.9. Cause: The mapping configuration of associated fields was not exported before the upgrade, and the corresponding data migration script was not executed after the upgrade.
- Phenomenon: Only single-threaded execution occurs when running audit tasks in batches, resulting in much slower than expected generation speed. Cause: Corresponding concurrency parameters were not configured, or the used FastGPT version does not support batch parallel functions.
How to confirm the configuration is correct
- Run a data source connectivity test to verify that connections from winding equipment, test benches, ERP/MES systems to FastGPT are normal, and check that parsed fields match preset mapping rules.
- Upload a winding process card and insulation treatment record, confirm that the parsed document content can be normally imported into the audit process, with no empty fields or format errors.
- Trigger a batch audit task, check that the number of concurrent threads in the task execution log meets the configuration requirements, with no connection timeout or error messages.
- Restart the FastGPT service, verify that all configuration parameter values have not been reset, and internal interface calls and file storage functions work normally.
The values provided here are common starting points and should be measured against the reader's own samples.