What the data for this category looks like
Data for this category primarily comes from customer drawings and special characteristic lists, ERP/MES systems, measuring and inspection tools, equipment process parameters, and supplier quality certificates. Daily processed documents include control plans, process flowcharts, work instructions, first article and patrol inspection records, SPC reports, PPAP materials, and 8D reports. Data is updated synchronously with each production batch and each shift. Key process parameters are collected continuously. Data fields include part numbers, special characteristic parameters, process requirements, inspection results, quality assurance grades, and other relevant items. Units include industry standard units such as millimeters, newton meters, and percentages.
Constraints imposed by these characteristics on model integration and configuration
Multi-source heterogeneous data structures, high-frequency update characteristics, and strict compliance requirements create multiple constraints for model integration and configuration. Multi-source data must be standardized to fit model input requirements. High-frequency collected real-time parameters require models to support incremental data processing. The precision requirements for special characteristic parameters must align with the model’s context length to prevent truncation of critical information. Under compliance requirements, models must link the version and approval status of corresponding documents to ensure bid documents and quotations comply with vehicle customers’ quality system specifications. Additionally, standardized high-volume manufacturing processes require model-generated quotations and documents to match process parameters, avoiding field inconsistencies.
How to Configure the System
This configuration is based on the interface and parameter logic of FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
maxContext | 8000–12000 token | Adapts to context needs for splicing multiple documents in this category, avoiding truncation of critical content such as special characteristics and process parameters |
maxTokens | 4000 token | Matches the conventional output length of bid documents and quotation sheets, ensuring complete generation of quotation details and compliance explanations |
| Workflow/API tool selection | auto | Supports automatic tool calling to extract structured information required for bids, including ERP data extraction and SPC report analysis |
vision | enabled | Supports parsing image attachments in customer drawings and PPAP materials, matching processing needs for image-based documents in this category |
temperature | 0.3–0.5 | Reduces output randomness, ensuring consistency of bid document and quotation content that complies with industry standards and customer requirements |
quoteMaxToken | 1500 token | Limits the output length of the quotation details section, focusing on core parameters and cost accounting content |
Three Common Configuration Mistakes
- Issue: The model returns empty content, and the interface shows no valid response. Cause:
AIPROXY_API_ENDPOINTis not configured correctly, or the local Ollama service is not started, resulting in failed delivery of model requests. - Issue: Parsing failure errors occur after uploading PPAP materials, SPC reports, or other documents. Cause: The
visionconfiguration is not enabled, which cannot process embedded drawing image attachments, leading to parsing interruption. - Issue: Tool calls cannot be triggered to extract ERP data after setting
toolChoiceto auto. Cause: The locally deployed deepseek-r1:70b model does not have pre-installed tool calling plugins, or the corresponding tool node is not associated in FastGPT.
How to Verify Successful Configuration
- Submit a test request, enter a question containing customer special characteristic parameters, and confirm that the model’s returned content includes the corresponding fields and units.
- Upload PPAP materials containing images, and confirm that the model correctly extracts annotated information from the images.
- Set
toolChoiceto auto, submit a test request including a data extraction requirement, and confirm that the corresponding tool call is triggered. - Adjust the
temperatureparameter, compare output consistency across different values, and confirm the value range that meets business requirements.
The values provided here are common starting points and should be evaluated against the reader’s own samples.