What the Data for This Category Looks Like
The data for this category typically comes from silicon wafer incoming material records, production MES systems, operation logs of diffusion furnaces, coating equipment, and printing presses, outputs from EL and IV testing equipment, and laboratory test data. Daily documents include process cards, diffusion and coating parameter tables, printing and sintering process records, EL inspection images, IV grading data, and yield analysis reports. Data update cycles cover each batch of production, per-product test results, and daily production summaries. Fields include equipment operating parameters, electrical performance test indicators, appearance defect types, and other relevant content.
Constraints Imposed by These Characteristics During Deployment and Upgrade
Photovoltaic cell manufacturing data includes multiple types of content such as structured parameters, images, and grading data. During deployment, the system must adapt to parsing and storing multi-format data. High-frequency updated per-piece test data requires the system to have high-concurrency write and query capabilities. Database and cache parameters must be configured appropriately. Unified access to multi-source data requires advance planning of data connection paths to avoid production data interruptions that affect the accuracy of Q&A content. During upgrades, historical associated data of original equipment parameters must be retained to ensure consistent context for maintenance Q&A. At the same time, the system must be compatible with newly added EL image parsing plugins to ensure normal operation of defect recognition functions.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale for This Setting |
|---|---|---|
DB_MAX_LINK | 100–150 connections | Adapts to high-frequency concurrent access from multiple device logs and test data, avoids exhausting database connections |
MONGODB_URI | Full connection string pointing to a dedicated storage node | Stores large-volume unstructured content such as EL inspection images and IV grading data |
REDIS_URL | Connection address of a dedicated cache instance with memory no less than 2048 MB | Caches frequently queried process parameters and equipment inspection standards to reduce database load |
FE_DOMAIN | Intranet or public network domain name of the deployment server | Unifies front-end access entry, adapts to intranet deployment scenarios in production workshops |
FILE_TOKEN_KEY | 32-bit random string | Encrypts file upload requests to prevent unauthorized import of knowledge base content |
STORAGE_S3_ENDPOINT | Endpoint address of the corresponding object storage service | Stores batch-uploaded process documents, test reports, and other files |
Three Common Misconfigurations
- Phenomenon: After enabling the enhanced PDF parsing function, the knowledge base fails to normally parse photovoltaic cell process documents. Cause: Environment parameters required by minerU are not correctly configured, causing the parsing engine to fail to start normally.
- Phenomenon: When importing files via a workflow, the backend interface cannot receive JSON format content from POST requests. Cause: The correct Content-Type request header is not configured in the HTTP request node, resulting in failed parameter parsing.
- Phenomenon: A model has been enabled in the account model configuration, but the model cannot be selected in the workflow node. Cause: The model cache of the workflow has not been refreshed in time, or there is a compatibility issue between the deployment version (such as versions below FastGPT v4.17.0) and the configured version.
How to Confirm Proper Configuration
- Access the configured
FE_DOMAINaddress, check whether the system interface loads normally, and confirm that the front-end domain name configuration is correct. - Upload a photovoltaic cell process card PDF file, check whether it can be normally parsed and generate a knowledge base entry, and verify that the enhanced PDF parsing function is effective.
- Create an HTTP request node in the workflow, send a test JSON request to the backend interface, and confirm that the interface can normally receive and process the content.
- Enter the account model configuration page, refresh the workflow node's model list, and confirm whether the configured model is displayed normally.
The values given are common starting points and should be measured against the reader's own samples.