What Data for This Category Looks Like
Data for this category typically comes from customer product models, mold flow or forming simulation files, CAD/CAM source files, machining and measurement records, and test sample piece data. Daily documents include mold design task sheets, DFM reports, part drawings, machining process cards, mold trial reports, mold modification notices, and mold history records. Data updates are managed per project, with new content added or existing content revised after each mold trial or modification. Document fields may include mold flow analysis parameters, machining tolerances, assembly accuracy requirements, mold trial defect records, and similar items. There are no unified fixed fields or unit requirements, and corresponding rules must align with project document specifications.
What Constraints These Characteristics Impose on Deployment and Upgrade
Mold project data includes CAD models and multi-format documents that can reach hundreds of megabytes per file, which creates requirements for storage and parsing capabilities. Data is updated independently per project, with update schedules adjusted based on project progress, so incremental synchronization and version management must be supported. Documents from different projects have strict permission isolation requirements, so deployment must ensure data access security. In addition, process Q&A results are directly linked to mass production milestones, so deployment must maintain service stability to avoid Q&A delays or errors caused by improper configuration that disrupt production progress.
How to Set Configurations
This page applies to FastGPT v4.17.0. The following are officially supported standard configuration items:
| Config Item | Recommended Setting | Rationale |
|---|---|---|
DB_MAX_LINK | 20–30 connections | Mold project data includes associated drawings, process cards and mold trial reports. A stable database connection pool must be maintained to avoid request queuing |
MONGODB_URI | Cluster connection string or standalone instance address | Store multi-project mold documents and Q&A contexts, to ensure data synchronization during multi-node deployment |
REDIS_URL | Local instance on port 6379 or cloud-hosted Redis cluster with memory configuration no less than 1000 MB | Cache Q&A sessions and embedding vector results to reduce resource consumption from repeated calculations |
STORAGE_S3_ENDPOINT | Endpoint address adapted to enterprise object storage | Store CAD models and mold flow simulation files that can exceed 500 MB per file, compatible with S3 protocol storage services |
FILE_TOKEN_KEY | 32-character random string | Generate access tokens for mold project documents, to prevent unauthorized cross-project document access |
Three Common Misconfigurations
- Symptom: Slow Q&A responses after container deployment, with 504 status code and response time exceeding the preset threshold. Cause: Failure to adjust the
DB_MAX_LINKparameter, with too low connection pool upper limit causing database request backlog. - Symptom: 404 error returned by embedding model calls. Cause: Failure to correctly bind the embedding model service address to FastGPT's configuration parameters, leading to inability to access the target model.
- Symptom: Cross-project access to multi-project data. Cause: Failure to configure
FILE_TOKEN_KEY, with no independent access permissions set for documents of different mold projects.
How to Confirm Configurations Are Correct
- Log in to the system monitoring page in the FastGPT backend, check the database connection count, and confirm that the connection number set by
DB_MAX_LINKdoes not remain fully utilized. - Upload a mold trial report to verify that the file can be parsed normally and the corresponding Q&A context is generated, with no parsing failure prompts.
- Initiate a query about mold demolding processes, confirm that the response meets expectations, with no timeout errors.
- For multi-node deployments, check the logs of each node, confirm that the cluster connection configured by
MONGODB_URIis normal, with no data synchronization failure prompts.
The values provided here are common starting points and should be evaluated against the target deployment's own samples.