Tool Calling and Plugins for Dose Adjustment in Rational

Data Characteristics in This Category

Dose adjustment data primarily originates from drug inserts, clinical guidelines, pharmacopoeias, and pharmacokinetic (PK)/pharmacodynamic (PD) databases. The update frequency for this data is relatively low, typically occurring with revisions to drug inserts or updates to clinical guidelines, which can range from several months to several years. Data document structures vary. Drug inserts often consist of unstructured text, including sections on dosage and administration, and use in special populations (e.g., hepatic/renal impairment, elderly, pediatric). Clinical guidelines may present information in structured tables, specifying recommended dosages for different physiological indicators. Key fields include drug name, indication, patient physiological indicators (e.g., creatinine clearance CrCl, body weight BW, liver function grading Child-Pugh Score), pre-adjustment dose, post-adjustment dose, adjustment basis, administration route, and frequency. Units involved include milligrams (mg), milliliters (mL), kilograms (kg), milliliters per minute (mL/min), and hours (h). Unit conversion is a common requirement in data processing.

Constraints Imposed by These Characteristics on "Tool Calling and Plugins"

The unstructured and semi-structured nature of dose adjustment data requires tool calling to flexibly handle text parsing and information extraction. For example, extracting dose adjustment rules for specific patient populations from drug inserts necessitates natural language processing (NLP) capabilities. Due to multiple physiological indicators and complex decision logic, tools or plugins must support multi-parameter input and conditional branching to simulate clinical decision-making processes. The low data update frequency means knowledge base construction and maintenance can adopt a periodic update strategy, without needing real-time synchronization. However, when data updates occur, tool calling logic may need adjustment, especially if the basis for dose adjustment or calculation formulas change. Furthermore, the diversity of units and conversion requirements necessitate that tools either include built-in unit conversion services or call external ones to ensure calculation accuracy. Dependencies on external databases, such as querying creatinine clearance calculation formulas or drug interactions, also impose requirements on the external interface capabilities of tool calling.

Configuration Settings

Configuration Item	Suggested Value	Rationale for This Value
`maxContext`	`2000 characters`	Ensures complete patient basic information and relevant drug information are passed, providing sufficient context for dose calculation.
`toolCallTimeout`	`60 seconds`	Allows ample execution time, considering potential multiple external API calls (e.g., PK/PD model calculations).
`knowledgeBaseIds`	Specific knowledge base ID	Binds knowledge bases containing drug inserts and clinical guidelines, ensuring the tool can access the latest medication information.
`functionSchema`	Dynamically generated or predefined	Function parameters may differ for various drugs or adjustment scenarios, requiring flexible definition or template generation.
`maxIterations`	`3`	Allows the tool to make a limited number of attempts in complex scenarios, such as retrying with adjusted parameters after an initial call failure.
`outputFormat`	`JSON`	Standardizes the output format for easier parsing and display in subsequent processes, including adjusted dose, basis, and precautions.

Three Common Mistakes

A tool call returns a 500 error or an empty result because the patient physiological indicator field names passed to the tool do not match the expected parameter names, leading to internal calculation failure.
The AI platform cannot reference knowledge base content in a conversation, even if the knowledge base is globally configured. This occurs because the knowledgeBaseIds parameter is not explicitly passed in the tool calling workflow, preventing the workflow from recognizing and utilizing the specified knowledge base.
A tool executes an SQL statement but does not return the expected result. This happens due to improper database connection pool configuration or a syntax error in the SQL query statement, leading to an unsuccessful database operation.

How to Confirm Correct Configuration

Construct a Q&A with typical patient physiological indicators (e.g., CrCl 30 mL/min, BW 60 kg) and a specific drug. Observe if the tool accurately returns the adjusted dose and the basis for adjustment.
Test edge cases, such as patients with extremely abnormal hepatic or renal function, to check if the tool provides reasonable dosage recommendations or contraindication warnings.
Verify that tool call results reflect the latest medication guidelines or drug insert information after knowledge base content updates. Compare outputs before and after updates to confirm.

The values provided are common starting points and should be measured against the reader's own samples.

Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-21.