Data profile for this category
Core data for this category comes from ERP, WMS systems, barcode/RFID devices, purchase work orders, storage location management records, AGV transport systems, and weighing equipment. Daily documents include receiving slips, putaway slips, material picking and returning slips, inventory transfer slips, inventory count sheets, lot ledgers, first-in first-out (FIFO) and expiration date records.
Data update frequencies fall into three categories: real-time (AGV position, weighing data), daily (storage location changes, work order picking records), and monthly inventory (full warehouse material status).
Core fields include material code, lot number, storage location number, expiration days, inspection status, and weighing value, with units such as pieces, kilograms, and days.
Constraints for tool calling and plugins
Multiple data sources require tool calling to support integration with interfaces from different systems, and to handle document data with varying formats. Different update frequencies require matching tool calling trigger timing and data pull frequency. For example, real-time data needs frequent pulls, while monthly inventory data can be triggered on demand.
Diverse document structures require plugins to support custom field extraction rules, to accurately match required fields for different documents. The importance of core fields requires tool calling to add field validation logic, to avoid subsequent business risks caused by missing or incorrect key information.
Configuration guidelines
This page is written for FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Workflow/API tool selection | Auto-select based on scenario | Adapt to tool calling needs for different documents such as receiving, inventory counting, and inventory transfer |
| Workflow/API function-call configuration | Enable mandatory field matching | Ensure extracted core fields including material code and lot number meet business requirements |
| Tool calling timeout | 300–600 seconds | Cover normal business time spent on multi-source data pulling and physical and system record matching |
| Plugin context limit | First 10 historical documents | Avoid excessive historical data interfering with field extraction for current documents |
| Field extraction rules | Calibrate based on actual testing | Adapt to custom field combinations and formats for different documents |
| Error retry count | 2–3 times | Address network fluctuations or temporary interface call failures |
Three common configuration errors
- Scenario: After an application is converted to a team plugin, tool calling fails, returning empty results or an error. Cause: The
toolChoiceparameter is not enabled in plugin configuration, leading to incorrect binding of tool calling permissions. - Scenario: CSV collections created via interface calls do not match data generated in the interface, with incorrect field order or value format. Cause: The field separator and encoding format for the CSV file are not specified, causing the parsing logic to mismatch the default interface configuration.
- Scenario: After the plugin preview retrieves content, the final output has no results. Cause: The
functionCallparameter is not enabled in plugin configuration, causing extracted content to not be properly integrated into the output logic.
How to verify correct configuration
- Trigger tool calling for the corresponding document, check if the returned results include core business fields such as material code, lot number, and storage location number.
- Simulate a multi-source data pulling scenario, verify that the tool calling timeout setting covers normal business time consumption.
- Trigger an abnormal plugin call, check if retry operations are performed according to the configured retry count.
- Compare the CSV collection created via the interface with data generated in the interface, confirm that the field parsing logic matches the configured requirements.
The values provided are common starting points and should be measured against the reader's own samples.