This configuration is adapted for FastGPT v4.17.0.
What the data for this category looks like
Data for this category comes from daily production and compliance documents of skincare, haircare, makeup, and special cosmetics manufacturers. Documents include formulation process sheets, weighing and feeding records, emulsification or filling records, cleaning and sanitization records, semi-finished and finished product inspection records, and label review documents. Data sources cover registered or filed formulations, raw material safety materials, production equipment, ERP/MES systems, laboratory data, and packaging material version files. Updates occur with each production batch or each line change. Documents include fields such as formulation components, production parameters, inspection results, and label claims. Fields and units adjust dynamically based on the requirements of each production batch.
Constraints imposed on tool calling and plugins
Data updates with each production batch or line change. Tool calling must support real-time pulling of the latest batch’s filed formulations and production records to avoid using expired data. Plugins must adapt parsing logic for different document formats, to ensure accurate extraction of fields like formulation components and inspection results. Data sources are scattered across ERP/MES, laboratory equipment, and other systems. Plugins must support cross-system data linkage. Strict compliance constraints require a consistency check for formulations and filings during the calling process. This prevents content that does not match registered filings.
How to set the configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Workflow/API tool selection | auto, or bind a specified workflow ID | Adapts to multi-workflow scenarios, can automatically match corresponding tasks for bid document or quotation preparation |
| Single tool call timeout | 600 seconds | Covers the full time required for formulation verification and cross-system data pulling, to avoid mid-process interruptions |
| Supported document parsing formats | Enable docx, pdf, xlsx parsing switches | Covers common document types for this category, such as formulation process sheets and inspection records |
| Tool call result cache duration | 24 hours | Matches the update cycle of a single production batch, to avoid reading expired filed or production data |
| Workflow/API function-call configuration | Enable mandatory verification mode | Ensures content returned by tool calls complies with cosmetics filing compliance requirements, to avoid unauthorized claims |
Three common configuration errors
- An error stating the workflow is not matched appears during a call. This happens when the corresponding workflow ID for bid or quotation preparation is not bound in
toolChoice. The system cannot identify the task type. - The interface call fails to read a docx document from a URL. This occurs when the URL remote pulling switch for the document parsing plugin is not enabled, or the allowed domain name whitelist is not configured.
- The tool connection status is abnormal, but Navicat can connect normally. This is caused by one of two issues: no correct database query statement template is specified in the plugin configuration, or the plugin is not authorized to access the corresponding data tables in ERP/MES.
How to confirm the configuration is complete
- Trigger a test call. Check if the returned result includes the formulation components and filing verification information required for this category. Confirm the data source matches the current production batch.
- View plugin logs. Confirm cross-system data pulling time stays within the preset range, with no timeout errors.
- Verify multi-workflow scenarios. When switching between different bid tasks, the tool automatically calls the corresponding workflow to complete processing.
- Upload a docx-format formulation process sheet for this category. Confirm the plugin correctly extracts raw materials, production parameters, and other fields.
The values provided here are common starting points, and should be evaluated against the reader’s own samples.