Data Profile for This Category
This category covers the full production process of liquid milk, yogurt, milk powder and dairy ingredients. Daily documents include raw milk acceptance records, standardized ingredient lists, pasteurization or UHT heat treatment records, CIP cleaning records, filling records, sample retention and microbial test reports. Data sources include milk source testing systems, production MES systems, homogenization and heat treatment equipment, online temperature and flow collection modules, CIP systems, and LIMS laboratory management systems. Updates occur in real time for each milk tank and production batch. Each document includes fields such as batch number, test values, and equipment parameters. Value units include mg/kg, g/100g, ℃, seconds, CFU/mL, and others.
Constraints on Tool Calling and Plugins
This category's data characteristics impose multiple constraints on the tool calling and plugins workflow. Multi-source heterogeneous data sources require configuring authentication and data pull rules for multiple tools, to connect separately to milk source testing, MES, equipment, and LIMS systems. Tightly batch-bound update rhythms require using production batch numbers as the core filter condition during tool calling, to prevent cross-batch data from being included. Inconsistent unit field structures require automatic unit conversion and field standardization after tool calling, to ensure consistency of summarized data. Real-time update characteristics require tool calling response times to match production tempo, to avoid data lag that disrupts ledger generation.
Configuration Guidelines
This document applies to FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Tool calling trigger condition | Trigger by production batch number | Matches the category's batch-bound data characteristics, avoids cross-batch data mixing |
| Maximum parallel tool calls | First 3 tools | Balances data pull efficiency and system load, adapts to multi-source data connection scenarios |
| Field standardization rules | Uniformly convert to standard units including ℃, seconds, mg/kg, CFU/mL | Adapts to unit differences across multi-source documents in this category |
| Workflow/API tool selection | auto | Allows the model to independently select tools matching the business scenario to complete data pulling and summary |
| Workflow/API function-call configuration | Enable only when a specific tool must be specified | Avoids the model from mistakenly calling unnecessary tools, simplifies calling logic |
| Response timeout threshold | 600 seconds | Matches real-time production response time requirements, avoids data lag affecting ledger generation |
| Result filtering rules | Only retain fields associated with the current batch | Ensures the generated ledger only corresponds to the target production batch |
Three Common Misconfigurations
- Calling logs show data acquisition exceptions, and model channels return error. Cause: Tool calling is not triggered by production batch number, leading to pulling redundant cross-batch data that exceeds interface data return limits.
- Text extraction tool returns 400 Invalid JSON payload received. Unknown name error. Cause: Undeclared custom fields are passed during tool calling, such as directly passing category-specific fields without declaring them in tool parameters.
- PgVector plugin fails to pull device data after upgrade. Cause: Plugin data source authentication configuration is not updated, leading to invalid connection permissions with original CIP systems and MES systems.
How to Confirm Proper Configuration
- Manually enter a production batch number, trigger tool calling, and verify returned data only includes records related to that batch.
- Review tool calling logs, confirm all called tools are pre-configured multi-source connection tools, and no undeclared fields are passed.
- Check the PgVector plugin version, confirm it is updated to the latest version compatible with the current FastGPT version, and authentication configurations are correct.
- Simulate real-time data updates, trigger ledger generation, and verify response time meets preset time requirements.
The values provided here are common starting points and should be evaluated against the reader's own samples.