What data for this category looks like
Business data for metal cutting machining comes from product drawings and 3D models, CAPP/CAM systems, machine tool collection terminals, tool management systems, MES systems, and measurement and inspection equipment. Daily produced documents include process routes, operation cards, tool lists, CNC programs, first-piece inspection records, patrol inspection forms, and measuring tool calibration records. Core fields include tool life, cutting parameters, dimensional deviation, clamping references, and more. Data updates are triggered by work orders, individual products, sampling batches, or tool replacement actions. There is no fixed update cycle, but update frequency is closely tied to production pace.
What constraints these characteristics impose on the deployment and upgrade phase
The need to connect multiple sources of data for metal cutting machining requires the deployment phase to adapt to interface access from multiple systems, and requires configuration of dedicated storage and cache resources. Frequently updated process documents and exception reports require stable database connection pools to avoid resource exhaustion from concurrent writes. Different types of process documents have large format differences, requiring adaptation for parsing and storage of unstructured files such as CNC programs and 3D models. For scenarios deployed in the workshop intranet, external network access must be restricted, while ensuring connectivity between internal systems. This places high requirements on the precision of network configuration.
How to set the configurations
The following configurations apply to FastGPT v4.17.0
| Configuration Item | Recommended Setting | Rationale for This Setting |
|---|---|---|
DB_MAX_LINK | 10–20 connections | Metal cutting machining process documents and exception report data are written in batches with work orders. Limiting the number of database connections avoids server resource exhaustion |
MONGODB_URI | Intranet-accessible instance address | Stores unstructured business data such as operation cards and patrol records, adapting to batch storage requirements from multiple sources |
REDIS_URL | Locally deployed standalone instance address | Caches frequently accessed tool lists and reference plan data, adapting to business scenarios with high update frequency |
STORAGE_S3_ENDPOINT | Intranet object storage access address | Stores large-volume process files such as CNC programs and 3D models, meeting storage and reading requirements for large files |
FILE_TOKEN_KEY | Randomly generated 32-bit string | Encrypts access tokens for process files, preventing unauthorized file access |
FE_DOMAIN | Intranet domain name of the deployment node | Adapts to workshop intranet access scenarios, configures the access address for the frontend page |
Three common mistakes
- The symptom is that the FastGPT instance inside a Docker container cannot access intranet storage or external tools. The error shows connection timeout or 404 status code. The cause is that the container network mode is not configured as bridged, or the corresponding ports are not mapped, and the intranet access address for
STORAGE_S3_ENDPOINTis not set correctly. - The symptom is that the code running plugin returns an AxiosError 404 when executed. The cause is that the interface address configured in the plugin does not point to the locally deployed service, or the correct intranet access domain name is not configured in
FE_DOMAIN. - The symptom is that the intranet-deployed aiproxy cannot connect to external model services. The error shows connection timeout. The cause is that intranet proxy rules are not configured, or the proxy address is not pointed to the locally deployed model service.
How to confirm the configuration is complete
- Log in to the storage configuration page in the FastGPT backend, upload a CNC program file, and confirm that the file can be stored and downloaded to the local device normally.
- Upload an operation card document, run a parsing test, and confirm that preset fields such as tool number and cutting parameters can be extracted.
- Start the code running plugin, call the locally deployed tool interface, and confirm that the returned result has no 404 or timeout errors.
- Configure an advanced orchestration workflow, enter a test work order question, and confirm that the AI node after the judge can obtain the initial question content after running the workflow.
The values provided are common starting points and should be measured against the reader's own samples.