clinical-ddi-check 0.1.0

Drug-drug interaction check tool for clinical AI agents. Curated JSON data, RxNorm normalization, LangChain-ready.

clinical-ddi-check

Drug-drug interaction (DDI) check tool for clinical AI agents. Curated JSON data, RxNorm normalization, LangChain-ready.

Features

• Curated DDI data – Bundled JSON with common interactions (aspirin/warfarin, metformin/contrast, etc.)
• RxNorm normalization – Resolves "Aspirin 81mg" → "aspirin" via NLM RxNorm API
• Structured output – Pydantic models with severity, description, clinical advice
• LangChain adapter – Drop-in tool for LangGraph / OpenAI function calling

Installation

pip install clinical-ddi-check

For LangChain/LangGraph integration:

pip install clinical-ddi-check[langchain]

Quick Start

Standalone

from clinical_ddi_check import drug_interaction_check

# Check interactions between current meds
result = drug_interaction_check(medications=["warfarin", "aspirin"])
# result["count"] == 1, result["has_major_or_contraindicated"] == True

# Check proposed drug against current meds
result = drug_interaction_check(
    medications=["metformin"],
    proposed_drug="contrast media",
)
# result["interactions"][0]["severity"] == "contraindicated"

LangChain / LangGraph

from clinical_ddi_check import create_langchain_tool, LANGCHAIN_TOOL_SCHEMA

# Create a LangChain tool
tool = create_langchain_tool()
# Use with model.bind_tools([tool]) or in a LangGraph node

# Or use the raw schema for OpenAI function calling
# LANGCHAIN_TOOL_SCHEMA matches the format expected by bind_tools

Custom DDI Data

result = drug_interaction_check(
    medications=["drug_a", "drug_b"],
    ddi_data_path="/path/to/your/ddi.json",
)

DDI JSON format:

{
  "interactions": [
    {
      "drug_a": "aspirin",
      "drug_b": "warfarin",
      "severity": "major",
      "description": "Increased bleeding risk.",
      "clinical_advice": "Monitor for signs of bleeding."
    }
  ]
}

Severities: contraindicated, major, moderate, minor.

API

drug_interaction_check(medications, proposed_drug=None, patient_id=None, ddi_data_path=None)

Returns a dict:

• Success: interactions, medications_checked, proposed_drug, patient_id, count, has_major_or_contraindicated
• Error: error_code, message, details

normalize_medication_names(medications)

Normalizes drug names via RxNorm. Returns {"normalized": [...]} or error dict.

License

Apache-2.0