This document is based on FastGPT v4.17.0.
Data Profile for This Category
Data for this category originates from ERP material master data, melting furnace thermometers, spectrometers, MES heat records, mold management systems, and customer technical drawings. Daily generated documents include casting process cards, batching lists, heat records, mold histories, first article inspection records, material reports, and nonconforming product review forms. Data updates occur per heat, per batch, or daily. Fields include heat number, mold number, alloy batching ratio, melting temperature, pouring parameters, defect type, and linked material, equipment, and production process information.
Constraints During Deployment and Upgrade
Scattered data sources cover production equipment, management systems, and business documents. Reserve configuration space for multi-source data access during deployment. Data updates happen at high frequency, with new records created per heat or batch. Adjust database connection parameters during deployment to ensure write performance. Data includes unique identifiers such as heat number and mold number. Retain field compatibility during cross-version upgrades to avoid losing production process linkages after migration. For offline upgrade scenarios, complete database directory migration in advance to ensure consistent data links before and after upgrade.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
DB_MAX_LINK | 100–150 connections | This category has frequent data updates, with multiple records generated per heat. Ensure the database connection pool sufficiently supports concurrent write operations |
MONGODB_URI | mongodb://[username]:[password]@[intranet IP]:27017/fastgpt?authSource=admin | Most foundries deploy intranet databases. Specify intranet access addresses and authentication information to ensure data security |
REDIS_URL | redis://[intranet IP]:6379/0 | Used to cache frequently accessed heat and mold data. Intranet deployment reduces network latency and improves retrieval efficiency |
| Maximum Context Window | 12000 token | Documents in this category include long text content such as melting parameters and composition reports. Match long context processing capabilities |
STORAGE_S3_ENDPOINT | Intranet object storage endpoint address | Used to store attachments such as heat records and spectrometer test reports. Intranet deployment ensures secure data transmission |
| Reranking | Select a suitable rerank model based on business scenarios | This category of data includes multi-dimensional parameters and long text reports. Use a rerank model to improve retrieval accuracy |
Three Common Misconfigurations
- Phenomenon: Historical heat linkage data is lost after cross-version upgrade, and some fields appear empty on the interface. Cause: Compatibility of core linkage fields such as heat number and mold number was not retained during upgrade, causing the migrated database to fail to match the original business linkage logic.
- Phenomenon: The system fails to start after offline upgrade, with an error prompt indicating database connection failure. Cause: The
MONGODB_URIconfiguration was not updated to the intranet access address during migration, or the database directory was not synchronously migrated to the new server, resulting in link interruption. - Phenomenon: Content truncation occurs when the system processes long text test reports, and complete composition data cannot be retrieved. Cause: The maximum context parameter set during deployment is lower than the actual document length, leading to early content truncation.
How to Verify Proper Configuration
- Log in to the system backend, view the
MONGODB_URIandREDIS_URLconfiguration items, and confirm they point to intranet-deployed database and cache service addresses. - Import a historical heat record, run a retrieval operation, and check that returned results include core fields such as heat number and mold number, with no abnormal truncation.
- Simulate multiple concurrent write operations, verify that no overflow errors occur in the database connection pool, and confirm that the
DB_MAX_LINKvalue matches the current business scale. - Upload a spectrometer test report attachment to the system, confirm that it can be downloaded and viewed normally, and verify that the
STORAGE_S3_ENDPOINTconfiguration is active.
The values provided are common starting points and should be measured against the reader's own samples.