This page is written based on the configuration specifications for FastGPT v4.17.0.
What data for this category looks like
Process and standard data for pharmaceutical dosage manufacturing comes from approved formulation processes, MES electronic batch records, LIMS quality management systems, equipment and environmental monitoring data, and other sources. Daily documents include master production files, batch production records, batch packaging records, cleaning records, inspection records, deviation CAPA change documents, and batch release review forms.
Data is updated per production batch, each critical process step, and periodic monitoring plans. Fields include material batch numbers, process parameter ranges, test limits, equipment operating parameters, and more. Units mostly involve measurement standards for compliance requirements such as mass, volume, temperature, and time.
What constraints these characteristics impose on the "tool calling and plugins" workflow
Since data is updated per single production batch and linked to a unique batch identifier, tool calling must bind context tied to the batch dimension to avoid mixing data across batches. The document structure includes multiple types of compliance records, so tools must support targeted retrieval across multiple document types, rather than generic recall.
Data is updated frequently and linked to process execution nodes, so tool calling must support real-time synchronization of the latest batch records and deviation data to ensure that Q&A results match the latest status of the current production batch. Additionally, compliance requirements mandate that data sources are traceable, so tools must record calling links and data sources to meet audit needs.
How to set configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Workflow/API tool selection | Automatically select compliant tools | Must automatically match tools based on different scenarios such as process specification queries, compliance data retrieval, and anomaly notifications to avoid compliance risks caused by manual selection |
| Single Search Limit | Top 10 results | Compliance documents for pharmaceutical dosage manufacturing are grouped by batch. Too many retrieved results will exceed the context window and increase audit and compliance verification costs |
| Maximum Context | 8000 token | Must accommodate the complete batch record of a single production batch, associated deviations and inspection data, to ensure that Q&A covers all necessary information for the current batch |
| Plugin Timeout | 300 seconds | Database queries and LIMS data pulling require a certain amount of time. 300 seconds covers the normal duration of most compliance data retrieval tasks |
| Workflow/API function-call configuration | Triggered by binding production batch number | Must use the production batch number as a pre-call parameter to ensure that retrieved data matches the process and material information of the current production batch |
Three common mistakes
- Issue: When calling database connection tools, only three types of databases are supported, and Oracle database cannot be connected. Cause: Multi-database adaptation plugins are not configured. The current tool only has built-in basic database connection capabilities and does not extend Oracle driver support.
- Issue: The email sending plugin cannot be called normally, and no official usage guidelines can be found. Cause: The email plugin has not been downloaded and enabled in the plugin marketplace, and the platform's built-in plugin usage documentation has not been reviewed.
- Issue: An error
400 <400> InternalError.Algo.InvalidParameter: messages with role "tois reported when starting the project, but the Mongo visualization tool can connect normally. Cause: The parameter format of the email recipient is not configured correctly when calling the plugin, or database connection parameters are not correctly mapped in the FastGPT plugin configuration, resulting in parameter verification failure.
How to confirm that configurations are set correctly
- Submit a process specification query request for a single production batch, check whether the returned results match the approved formulation process information of the current batch, and confirm that the selection of retrieval tools and parameter configurations are correct.
- Trigger an email plugin test, check whether the email sending process can be completed normally, and confirm that the plugin is enabled and configured correctly.
- View the plugin calling link log, confirm that all database operations are bound to the production batch identifier, which meets the requirements of the pre-call conditions.
- Simulate a call request that does not comply with parameter rules, verify whether the corresponding error prompt is triggered, and confirm that the parameter verification configuration is in effect.
The values provided are common starting points and should be measured against the reader's own samples.