What data for this category looks like
Data for this category originates primarily from CAD/CAM software, order BOM systems, sheet material inventory ledgers, laser cutter and bending machine device logs, and MES manufacturing execution systems. Daily generated documents include 3D part unfolding drawings, CNC cutting programs, bending process cards, nesting planning drawings, first-piece inspection records, and process flow cards. Data updates align with order delivery timelines, daily inventory changes, and each production changeover. Document fields include sheet thickness, bending coefficient, hole position coordinates, bending angle, process number, and other metrics. Most units use millimeters, degrees, and square meters.
What constraints these characteristics impose on deployment and upgrade
Sheet metal cutting and bending data has scattered sources and varied formats. During deployment, teams must adapt to unstructured documents such as DXF/DWG files exported from CAD and G-code CNC programs, and connect structured data interfaces from multiple systems. Document update frequency is high, and trigger scenarios are unpredictable. During upgrade, incremental synchronization mechanisms must support on-demand pulling and real-time pushing. Document fields vary widely across different orders. Configure flexible field mapping rules to avoid knowledge base retrieval failures caused by missing fields. Production changeovers trigger bulk data changes. Deployed retrieval services require low-latency response capabilities, otherwise real-time query needs at production sites will be affected.
How to set configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
| REDIS_URL | redis://localhost:6379/0 | Used to cache frequently retrieved process parameters and document chunks. Local deployment default port meets lightweight deployment requirements |
| STORAGES3ENDPOINT | http://localhost:9000 | Adapts to local MinIO object storage, used to store large-volume documents such as unfolding drawings and bending process cards |
| DBMAXLINK | 100 | Limits database concurrent connection count to avoid exhausting connection resources of sheet material inventory and MES systems during simultaneous retrieval across multiple orders |
| FE_DOMAIN | http://fastgpt.example.com | Configures frontend access domain name for generating valid links for document previews |
| FILETOKENKEY | Random 32-character string | Used to encrypt document access tokens to prevent unauthorized retrieval of process documents |
| ROOT_KEY | Random 64-character string | Used to configure system administrator access keys to secure permissions after deployment |
Three common configuration mistakes
- Phenomenon: Cannot select an image understanding model when creating a general knowledge base. Cause: FastGPT v4.17.0 local deployment does not install multimodal model dependency packages, and the multimodal feature switch is not enabled in the configuration file.
- Phenomenon: Persistent errors occur during local deployment of SenseVoiceSmall. Cause: The operating environment dependencies for the speech model are not correctly configured, or model weight files are not placed in the storage path specified by
STORAGE_S3_ENDPOINT. - Phenomenon: Call to local MCP Server fails, returns a connection timeout error. Cause: The service address is not correctly filled in the config file, or network access permissions for the corresponding port are not enabled.
How to confirm configurations are correctly set
- Log in to the FastGPT frontend interface, upload a sheet metal unfolding drawing, and confirm normal parsing and generation of retrieval chunks.
- Navigate to the system configuration page, check that the configuration status of
REDIS_URLandSTORAGE_S3_ENDPOINTshows normal connection. - Submit a process-related query, such as "How to calculate bending coefficient", and confirm returned results include locally stored bending process card content.
- Test multi-order data synchronization, confirm newly added nesting drawings are automatically synchronized to the knowledge base.
The values provided are common starting points and should be measured against the reader's own samples.