What data this category includes
Daily documents for this category include cell bank or seed lot records, culture and purification records, aseptic filling records, environmental monitoring data, media fill simulation results, sterility test reports, batch release documents, and more. Data sources cover cell bank and seed lot systems, process analysis platforms, cleanroom monitoring systems, MES, LIMS, and equipment integrity testing tools. Update frequencies fall into three groups: real-time (cleanroom monitoring, process parameters), per monitoring plan (environmental monitoring), and per batch (culture and purification, filling records, and batch release documents). Each record contains fields such as batch ID, monitoring timestamp, key parameter thresholds, and test results. Some fields have clear units, such as culture temperature (℃) and colony-forming units (CFU).
What constraints these characteristics impose on tool calling and plugins
Data sources for this category are scattered, and the data includes both structured parameters and unstructured long text. Tool calling must support connecting to multiple types of system interfaces and parsing complex documents. Update frequencies vary widely across different data sources, so tool calling must adapt to multiple trigger logic types including real-time, per plan, and per batch. Production anomalies can lead to full batch scrap or serious patient safety risks, so tool calling must include built-in verification and failure handling mechanisms. This prevents the generation of incorrect process documents or exception reports. Category-specific fields and units require that tool calling strictly matches interface parameter formats, avoiding process interruptions caused by data incompatibility.
How to configure settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Workflow/API tool selection | auto | Automatically select the corresponding data source plugin based on the content of process documents and exception reports, no manual tool specification required |
| Workflow/API function-call configuration | Enabled | Structuredly call interfaces of systems such as LIMS and MES to retrieve structured production data related to batches |
| Tool call retry count | 2 times | Limited retries avoid invalid requests while reducing the risk of call failure from temporary network fluctuations |
| Tool call timeout threshold | 600 seconds | Adapt to the real-time pull duration of some cleanroom monitoring data, preventing timeout interruptions of batch-related processes |
| Tool call data source scope | Only cell bank, LIMS, and MES three data sources | Limit the call scope to avoid irrelevant data interfering with report compliance checks |
| Structured data parsing switch | Enabled | Extract key parameters from long text such as culture and purification records for compliance checks in exception reports |
Three common configuration mistakes
- Symptom: File format parameters cannot be passed correctly during HTTP requests, and the interface returns a
400 Bad Requesterror. Cause: The parameter type for tool calling is not configured per the structured attachment format of biologic aseptic records, so the incoming file format does not match the system interface requirements. - Symptom: Tool calling is not triggered by default, and the generated exception report lacks support from historical batch data. Cause:
toolChoiceis not set toautoand not associated with the knowledge base plugin, so the model does not automatically call the retrieval tool to obtain historical reference data. - Symptom: Incorrect global variable parameter assignment causes process document generation to fail, and the interface returns a
500 Internal Server Error. Cause: API call parameter mapping is not configured per category-specific fields such as batch number and monitoring timestamp, so incoming parameters do not match the required fields of the system interface.
How to confirm the configuration is correct
- Initiate a simulated call, pass a test batch ID, and verify that the tool automatically pulls structured production data from the corresponding LIMS and MES systems.
- View the tool call logs in FastGPT v4.17.0, and confirm that the parameter format when
functionCallis triggered matches the system interface requirements. - Simulate an exception scenario to trigger tool calling, and check that retries are performed as configured and failure alerts are triggered.
- Generate a process document or exception report, and verify that referenced data comes from the specified category data sources and that field units meet production requirements.
The values provided are common starting points and should be measured against the reader's own samples.