This configuration guide applies to FastGPT v4.17.0.
What the Data for This Category Looks Like
Data sources for this category include cell grading equipment, production MES systems, welding and torque tools, BMS software, and EOL test benches. Updates occur per single battery pack, single energy storage cabinet, or each BMS software version change. A single record includes fields such as grouping rules, BOM lists, welding torque values, insulation and voltage withstand data, EOL test results, and system configuration parameters. Most fields are physical quantities or version identifiers. Torque units are N·m, and insulation resistance units are Ω.
Constraints Imposed by These Characteristics on Database and Operations & Maintenance
High-frequency updates at the single pack or single cabinet level require databases to support low-latency single-record writes and queries. Avoid batch operations that consume excessive resources. Multi-source heterogeneous data—including device collected values, software configurations, and test reports—requires separate structured and unstructured storage. Configure appropriate storage engines. Data traceability requirements mandate that complete version change records be retained. Operations links must support quick historical data traceback for single entities. Industrial device data access stability requires database connection pool configurations to adapt to concurrency fluctuations in industrial scenarios. Avoid connection overflow. Additionally, complex data association relationships require efficient cross-table queries. This supports multi-dimensional association during defect analysis.
How to Set Configurations
| Configuration Item | Recommended Approach | Rationale |
|---|---|---|
MONGODB_URI | Point to a replica set instance mounted on SSD, with read-write permission authentication | Store unstructured test reports and BOM documents. Replica sets ensure data reliability |
PG_URL | Configure a dedicated relational instance with SSL encryption | Store structured physical quantity data such as torque and insulation resistance. Ensure transmission security |
DB_MAX_LINK | 64–128 | Adapt to concurrent data reports from multiple industrial devices. Avoid connection overflow |
REDIS_URL | Point to an in-memory cache instance | Cache real-time BMS status and test progress. Shorten high-frequency query response times |
| Cache expiration time | 7200 seconds | Cache real-time test progress data. Avoid expired data affecting traceability accuracy |
| Data backup retention period | 30–90 days | Adapt to compliance requirements for quality traceability. Retain sufficient historical data for defect analysis |
Three Common Configuration Mistakes
- Phenomenon: A 503 status code appears when calling FastGPT application interfaces, or requests queue and time out. Cause: The
DB_MAX_LINKparameter is not adjusted. The connection pool is exhausted when industrial devices report data concurrently. - Phenomenon: Unable to query cell grading data from the SQL database. Returns empty results. Cause: The access permissions for
PG_URLare not configured correctly. Query permissions for the corresponding business database are not specified. - Phenomenon: Database backup exports cannot be restored at the single pack granularity. Only full knowledge base data can be exported. Cause: Backup strategies for structured business data and unstructured documents are not differentiated. The knowledge base dimension is incorrectly used as the only export granularity.
How to Verify Correct Configuration
- Execute single battery pack data write and query operations. Confirm response times meet business requirements. Adjust the
DB_MAX_LINKparameter based on actual operational scenarios. - Trigger a database backup task. Check if exported files include single pack data within the specified period. Confirm backup granularity meets traceability requirements.
- Simulate concurrent data reports from multiple devices. Monitor database connection count status. Confirm no connection overflow occurs.
- Test querying specified cell grading data from the SQL database. Confirm returned results match device collected values.
The values provided are common starting points. They should be evaluated against the samples relevant to the specific deployment.