What the data for this category looks like
Data for this category typically originates from silicon wafer incoming material records, factory MES systems, operation logs of diffusion furnaces, coating equipment, and printing presses, EL and IV test results, and laboratory test data. Daily retained documents include process cards, diffusion and coating parameter configurations, printing and sintering process records, EL detection images, IV grading data, and yield reports. Data updates align with each production batch, each individual cell’s detection nodes, and daily process summary statistics. Data fields include process parameter values, detection images, electrical performance indicators, batch codes, and more. Formats and units adjust based on changes to different process parameters.
Constraints imposed on deployment and upgrade
Multi-source heterogeneous data structures, including structured parameters, images, and text reports, require configuring multimodal data storage and parsing capabilities during deployment to support access to business data in various formats. The high-frequency update cycle requires adjusting database connection pool parameters to support thousands of concurrent writes and queries, preventing data delays that disrupt real-time analysis. Cross-departmental differences in document sources require adapting log access protocols for different devices during upgrades, and unifying data association logic across disparate systems. The core requirement of cell-level traceability also requires configuring precise batch association rules to avoid data confusion across production batches, ensuring the accuracy of analysis results.
Configuration Settings
This page is written based on FastGPT v4.17.0.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
DB_MAX_LINK | 20–30 connections | Supports concurrent writes and queries of thousands of device logs and test data per batch, avoiding connection exhaustion |
MONGODB_URI | mongodb://[authentication info]@[host address]:[port]/fastgpt?authSource=admin | Stores semi-structured and unstructured business data such as EL detection images and IV grading data |
REDIS_URL | redis://[host address]:[port]/0 | Caches frequently accessed process parameters and batch association indexes, reducing query pressure on the core database |
STORAGE_S3_ENDPOINT | enterprise-built object storage node address | Stores large volumes of original EL detection images and process record files, supporting data sharing across multiple production nodes |
FILE_TOKEN_KEY | randomly generated 32-bit string | Secures access permissions for sensitive files such as detection images, preventing unauthorized access |
FE_DOMAIN | enterprise official deployed frontend access domain name | Configures the access address for the frontend interface, ensuring normal loading of knowledge base search and workflow pages |
Three Common Configuration Errors
- Symptom: After local deployment, process documents linked to the enterprise knowledge base cannot be correctly cited in responses. Only citation markers appear, and corresponding document content fails to load. Cause:
STORAGE_S3_ENDPOINTor the file storage path is not correctly configured, preventing normal loading of EL detection images, process cards, and other documents. - Symptom: After adding a custom quality analysis plugin to a workflow, the interface does not display input and output parameter configuration items, and no error logs are generated. Cause: Plugin parameter editing permissions are not enabled in the permission configuration corresponding to
ROOT_KEY, or the plugin configuration file is not correctly uploaded to the deployment directory. - Symptom: Calls to the embedding model continuously return errors, and text vectorization cannot be completed. Cause: The embedding-2 model is not correctly deployed and connected to the Open API channel, or the corresponding model call parameters are not updated in the system configuration.
How to Verify Successful Configuration
- Execute a batch import test task for process cards and EL detection images, and check whether the number of data entries in the file storage directory and database match.
- Start a workflow test, add a custom plugin, and view the parameter configuration interface to confirm that input and output fields can be edited normally.
- Call the embedding model interface with test text, and check whether vector results are returned normally without error messages.
- Access the frontend address corresponding to
FE_DOMAIN, and verify that the knowledge base search function can normally return matching process parameters and detection data.
The values provided are common starting points and should be measured against the reader's own samples.