What this category of data looks like
Data sources include ERP material master data, melting furnace temperature measurement data, spectrometer test results, MES heat records, mold management system data, and customer drawings. Update frequency is per heat, per batch, or daily. Daily documents include foundry process cards, batching lists, heat records, mold history records, first article inspection records, material test reports, and nonconforming product review forms. Fields include heat number, melting temperature, pouring time, mold number, material grade, defect type, and others. Units include ℃, seconds, pieces, and others.
Constraints imposed by these characteristics on the "tool calling and plugins" workflow
Multi-source heterogeneous data sources require plugins to support interface formats of different systems, preventing data parsing errors. High-frequency update cycles require tool calling triggers to align with production rhythms, avoiding use of outdated data. Mixed structured and unstructured document content requires tools to support both field extraction and text parsing. Cross-departmental data flow requires permission configurations for plugin calls to match business roles, ensuring data security.
How to configure settings
The following configurations are adapted for FastGPT v4.17.0.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
| Workflow/API tool selection | Auto | Adapt to the need for parallel calls of multiple data sources in foundry scenarios, automatically matching corresponding plugins |
| Workflow/API function-call configuration | Enabled | Support structured parameter transfer, suitable for standardized entry of alloy batching and temperature parameters |
| Document parsing trigger timing | After each heat production is completed | Match data update frequency, avoid calling outdated heat records |
| Plugin timeout threshold | 600 seconds | Cover the average time range of melting and spectrometer testing |
| Number of data items per single plugin call | Top 10 | Match the conventional data volume of daily production batches |
| Global variable initial value | Calibrated based on actual measurements | Adapt to heat number rules of different foundry workshops |
Three common configuration mistakes
- When a foundry process card link is passed to the document parsing tool, only empty text is returned. This occurs because parsing rules for structured foundry documents are not enabled. Using only general parsing logic cannot extract process parameter fields.
- When a company-deployed vector model plugin is called, access is denied. Curl testing operates normally. This occurs because the whitelist IP or identity token for the company's model was not added in the FastGPT plugin configuration, causing interface verification to fail.
- After setting the global variable initial value for a plugin, the variable does not update during calls. This occurs because the global variable was not bound to the plugin's input parameter fields. Setting only the initial value in the plugin configuration does not take effect.
How to verify a successful configuration
- Manually trigger the document parsing tool, pass a foundry process card link, and check if fields including heat number and melting temperature can be extracted.
- Call the plugin connected to spectrometer data, and check if the returned component data matches actual heat records.
- Trigger a plugin call, and check if the global variable changes as the heat number updates.
- Simulate a single plugin call timeout scenario, and check if the preset retry logic is triggered.
The values provided are common starting points and should be measured against the reader's own samples.