What data for this category looks like
Data for this category comes from electromagnetic design data, copper wire and insulating material batch records, winding equipment, balancing machines, electrical test benches, and ERP/MES systems. Daily documents include winding process cards, insulation treatment records, dynamic balance reports, withstand voltage insulation test reports, and type and factory test reports. Data is updated per individual motor, per product batch, or per type change. Core processing steps include checking winding turns and wiring, controlling paint dipping, drying, and insulation systems, and analyzing abnormal data for vibration, temperature rise, efficiency, and withstand voltage.
What constraints these characteristics impose on deployment and upgrade
Data sources for this category are scattered, covering multiple channels including equipment, materials, and systems. Deployment requires configuration to support multi-source data integration. Data update cycles are inconsistent, triggered by individual units, batches, or type changes. Upgrades require adjusting the incremental update logic of the vector database to match real-time requirements. There are diverse document types with specialized fields. Deployment requires preset extraction rules for specific fields such as winding turns and withstand voltage values. Upgrades must support new test report formats. Additionally, the special data formats of high-voltage and temperature rise tests require dedicated parsing templates to be configured during the deployment phase.
How to set the configuration
This configuration is based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
MONGODB_URI | Connection string pointing to the enterprise's dedicated MongoDB instance for electrical test data | Core test data such as withstand voltage tests and dynamic balance reports for this category are stored in a dedicated MongoDB database |
REDIS_URL | Local Redis service address, with 8 GB memory quota | Used to cache frequently accessed batch material and equipment operation data. 8 GB covers the daily data caching requirements of a single factory, calibrated via actual measurements |
STORAGE_S3_ENDPOINT | Standard endpoint address of the enterprise's internal object storage, with a 100 MB single file limit | Meets the storage requirements for unstructured documents such as winding process cards and type test reports |
DB_MAX_LINK | 200 | Matches the upper limit of concurrent data access from winding equipment and test benches, preventing connection overflow |
PG_URL | PostgreSQL database connection address for integrating the enterprise's ERP/MES system | Used to synchronize structured business data such as copper wire insulating material batches and production work orders |
FILE_TOKEN_KEY | Key string consistent with the enterprise's document encryption system | Ensures the security of parsed document data during transfer |
Three Common Errors
- A connection timeout error occurs every 30 minutes after deployment, and recovers after restarting the service. The cause is insufficient memory quota configured for
REDIS_URL, leading to overflow of frequently cached data and forced disconnection of connections. - The system version still shows the old version after running the upgrade script. The cause is that the local image cache was not cleared, so the old version image was pulled instead of executing the correct image update command.
- Documents have been parsed, but the model does not generate responses based on the document content. The cause is that
MONGODB_URIwas not correctly configured to point to the test data repository, so the vector database did not synchronize the test report data for the corresponding batch, and the model could not retrieve relevant context.
How to Confirm Proper Configuration
- Run the
docker pscommand to confirm that all FastGPT-related containers are running, with no abnormal exit records. - Access the domain specified by
FE_DOMAIN, upload a winding process card after entering the system, and confirm that the parsed document can be successfully retrieved in the vector database. - Connect to the database corresponding to
PG_URL, run a basic query statement, and confirm that material batch data from the ERP/MES system can be successfully synchronized to the FastGPT system. - View the Redis service monitoring panel corresponding to
REDIS_URL, and confirm that the number of cached connections does not exceed the 200 limit configured inDB_MAX_LINK.
The values provided are common starting points and should be measured against the reader's own samples.