What Data Looks Like for This Use Case
Photovoltaic module encapsulation scenario data sources include cell grading records, glass adhesive film backplane batch ledgers, workshop MES systems, string welding and lamination equipment operation logs, EL/IV tester test reports, and more. Data is generated synchronously per module and per material batch. Daily documents include layout drawings, material batch records, lamination process parameter logs, EL defect images, IV power test reports, insulation pressure resistance records, packing lists, and others. Each document includes fields such as cell current gear, lamination temperature and duration, crack location coordinates, power value, and more. Units include degrees Celsius, seconds, watts, megapascals, and others.
Constraints Imposed by These Characteristics on Deployment and Upgrade
High-frequency updated multi-source data in this scenario requires the deployment phase to support real-time vector synchronization and multi-format document parsing. It must be compatible with different data source types including equipment logs, test reports, and ledgers. Data association accuracy directly impacts the accuracy of process Q&A. Database and cache service stability must be ensured during deployment. Errors in this scenario are associated with customer claims and production losses. A verification link must be reserved after deployment. During upgrades, compatibility of the historical Q&A library must be maintained to avoid disrupting daily workshop operations.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
DB_MAX_LINK | 100–150 | Adapts to the multi-source data concurrent access requirements of photovoltaic module encapsulation scenarios, and avoids service interruptions caused by connection pool exhaustion |
MONGODB_URI | Connection string for the factory-exclusive cloud or local cluster | Used to store frequently updated component batch and test data, ensuring read-write performance and data security |
REDIS_URL | Local cache instance, cache duration 3600 seconds | Reduces overhead from repeated queries to MES and equipment data, improving Q&A response speed |
FE_DOMAIN | Factory intranet domain name or fixed intranet IP | Ensures stable internal access for departments such as string welding, lamination, and testing |
STORAGE_S3_ENDPOINT | Access address of the factory-built object storage | Stores large-volume documents such as EL images and IV test reports, reducing cross-network transmission latency |
Three Common Misconfigurations
- Phenomenon: Connection refused is returned when vLLM connects to OneAPI. Cause: Incorrect configuration of
MONGODB_URIorREDIS_URL, preventing the model service from completing initialization verification of dependent components. - Phenomenon: Containers run normally but the frontend page cannot be opened. Cause:
FE_DOMAINis not configured with the correct intranet access address, or port mapping rules are not correctly mapped to the application port. - Phenomenon: The downloaded yml file only displays 404 content during Docker deployment. Cause: No acceleration source was configured when pulling the image, causing the official image file pull to fail and only returning an error page.
How to Confirm Configuration Completion
- Run initialization using the FastGPT v4.17.0 deployment script, execute the
docker pscommand, and confirm that all dependent service containers are in the running state. - Access the configured
FE_DOMAINaddress, verify that the frontend page can load normally and display the initialization interface. - Call the model service interface, verify that the configured vector model and language model can be accessed normally, and return valid Q&A results.
- Upload a lamination process record document for a photovoltaic module, verify that the system can correctly parse and generate matching Q&A context.
The values provided are common starting points and should be measured against the reader's own samples.