What the data for this category looks like
Daily documents from process, equipment, quality, and production departments of crystalline silicon cell manufacturers include process cards, diffusion and coating parameters, printing and sintering records, EL inspection images, IV grading data, and yield reports. Data sources cover silicon wafer incoming materials, MES systems, diffusion furnaces, coating equipment, printing presses, EL and IV testing equipment, and laboratory data. Update frequency is per production batch, per individual inspection, and daily synchronization. Documents include fields such as process parameter values, inspection result grades, and production batch numbers. Units include degrees Celsius, amperes, watts, and others. Some documents use a mixed format of structured tables and images.
Constraints imposed by these characteristics on deployment and upgrade
Photovoltaic cell manufacturing documents contain large-volume images and structured data, with high update frequencies. Multiple departments submit documents simultaneously. These factors impose clear constraints on storage, connection pool, and upload configurations during deployment. Concurrent access to multi-source data requires database connection configurations adapted to high-concurrency scenarios, to avoid upload failures caused by connection exhaustion. Batch uploads of large-volume documents require adjustments to storage and interface timeout configurations, to prevent false upload failure alerts in the interface. Cross-departmental document cross-review requires configurations adapted to local area network access scenarios, to avoid internal collaboration obstacles caused by only container IP addresses being accessible.
How to set the configurations
Configuration settings are based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
DB_MAX_LINK | 20–30 connections | Adapts to concurrent scenarios where multiple departments upload process and inspection data, avoiding upload failures caused by connection pool exhaustion |
STORAGE_S3_ENDPOINT | Local object storage endpoint address | Adapts to storage needs for batch uploads of large-volume files such as EL inspection images and IV data, reducing cross-cloud storage latency |
REDIS_URL | Local Redis instance address, with memory limit set to 2048 MB | Caches document chunks and retrieval results, supports high-frequency cross-review requests, and reduces repeated calculations |
FE_DOMAIN | Local area network-accessible domain name or IP:port | Adapts to internal local area network access scenarios, avoiding internal collaboration obstacles caused by only container IP addresses being accessible |
FILE_TOKEN_KEY | Custom 32-character random string | Secures upload interfaces, preventing unauthorized access to batch process and inspection data |
MONGODB_URI | Local MongoDB connection string, with connection timeout set to 600 seconds | Stores batch yield reports and process parameter documents, adapting to long-duration read and write requirements for large documents |
Three common misconfigurations
- When uploading process parameter documents or EL inspection images larger than 500 MB, the interface returns a timeout error, but the background upload process proceeds normally, and the file is displayed after a period of time. This occurs because
DB_MAX_LINKand interface timeout configurations are not adjusted, leading to connection pool exhaustion or an overly low timeout threshold. - After Docker deployment is complete, built-in model calls return not found or permission errors. This occurs because
MONGODB_URIand associated model service parameters are not correctly configured, causing an interruption in the link between the knowledge base and model services. - After modifying the authentication key, existing service processes cannot respond to new authentication requests, and the interface prompts insufficient permissions. This occurs because not all FastGPT-associated service processes are restarted; only some containers are restarted, so the configuration does not take effect.
How to verify the configuration is correct
- Upload a process card file larger than 100 MB, confirm that the interface normally displays the upload progress with no timeout errors.
- Use another device on the local area network to access
FE_DOMAIN, confirm that the login page loads normally and login can be completed. - Run the
docker pscommand, confirm that all FastGPT-related containers are in a running state with no abnormal exits. - Batch upload five EL inspection images, confirm that the chunked content of the corresponding documents can be retrieved in the knowledge base within 10 minutes.
The values provided are common starting points, and should be evaluated against the reader's own samples.