What data for this category looks like
Data sources for this category include customer product models, mold flow or forming simulation files, CAD/CAM drawings, machining and measurement data, and records related to trial mold samples. 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 trigger at project milestones, and are synchronized after each mold trial or modification. Fields include engineering-specific parameters such as draft angle, shrinkage rate, assembly tolerance, and defect level.
Constraints for model access and configuration
This category’s data includes multi-format professional engineering files and structured parameters. Model access requires adaptation to non-pure text formats such as CAD files and simulation results. The high-frequency update rhythm of projects, mold trials, and modifications requires context windows to match the full data volume of a single project, preventing truncation of key parameters. Bid documents must strictly align with professional mold design terminology. During configuration, ensure the model recognizes exclusive engineering fields, and supports unified indexing of multi-source data.
The following configurations are based on FastGPT v4.17.0.
How to set configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
maxContext | 8000–12000 token | Matches the total volume of complete documents and parameters for a single project’s mold design, avoiding truncation of critical content such as mold trial reports and modification records |
maxTokens | 4000 token | Adapts to the output length of bid documents and quotation sheets, ensuring complete generation of quotation details and design descriptions |
quoteMaxToken | 1500 token | Limits the document length of a single reference, prioritizing extraction of core content such as DFM reports and part drawing parameters |
temperature | 0.2–0.4 | Reduces random output probability, ensuring that professional terminology and parameters in bid documents accurately match mold design standards |
vision | Enabled | Supports parsing visual data such as CAD part drawings and photos of trial mold samples, supplementing engineering information beyond text |
AIPROXY_API_ENDPOINT | Fill in according to the deployment environment | Adapts to the large model access address for local or cloud deployments, ensuring normal request routing |
AIPROXY_API_TOKEN | Fill in according to authorization requirements | Completes identity verification for model access, ensuring request permission compliance |
Three common configuration errors
- Symptom: GPU resources are not utilized when calling the large model, and single-core CPU usage reaches 100%. Cause:
AIPROXY_API_ENDPOINTis not correctly configured to point to a locally deployed large model service, resulting in requests being routed to CPU execution. - Symptom: Mold parameters in generated bid documents deviate from industry standards, such as draft angle and shrinkage rate values that do not meet industry norms. Cause: The
temperaturevalue is set too high, causing model output to stray from professional specifications. - Symptom: Rectification requirements for mold trial defects are missing from generated quotation sheets. Cause: The
quoteMaxTokenvalue is too small, truncating core rectification records from long documents.
How to verify successful configuration
- Submit test data including CAD part drawings and mold trial reports. Check if the model correctly recognizes visual content to confirm the
visionconfiguration is active. - Generate a simulated bid document. Cross-reference professional parameters against preset mold design standards to validate the
temperatureconfiguration. - Review the context content referenced by the model. Confirm that complete core documents such as DFM reports and mold modification notices are included to verify
maxContextandquoteMaxTokenvalues are appropriate. - Check API request logs. Confirm requests are correctly routed to the specified large model service to validate
AIPROXY_API_ENDPOINTandAIPROXY_API_TOKENconfigurations.
The values provided are common starting points and should be measured against the reader's own samples.