⚙️ Logic-LM & Symbolic Solvers

# Logic-LM for drug interaction checking (simplified)
from anthropic import Anthropic

client = Anthropic()

def logic_lm_drug_checker(current_drugs, candidate_drug):
    """Check drug interactions using Logic-LM approach"""

    # Step 1: Problem Formulation (NL → Logic)
    formulation_prompt = f"""
Convert this drug interaction problem to logical constraints:

Current drugs: {', '.join(current_drugs)}
Candidate drug: {candidate_drug}

Output logical predicates:
- Drug(X) for each drug
- Contraindication(DrugA, DrugB) for known interactions
- Query: Can_add({candidate_drug})?

Formal constraint:
Can_add(D) :- NOT (exists C in CurrentDrugs where Contraindication(C, D))"""

    formulation = client.messages.create(
        model="claude-3-5-sonnet-20241022",
        max_tokens=300,
        messages=[{"role": "user", "content": formulation_prompt}]
    )
    logical_formulas = formulation.content[0].text

    # Step 2: Symbolic Solver
    # (In practice, would use Z3 or similar solver)
    solver_prompt = f"""
Using these logical constraints:
{logical_formulas}

Apply the constraint and determine:
- Is the query satisfiable? (Can we add the drug?)
- If not satisfiable, what contradictions exist?
- Provide the result as: SATISFIABLE / UNSATISFIABLE"""

    solver = client.messages.create(
        model="claude-3-5-sonnet-20241022",
        max_tokens=300,
        messages=[{"role": "user", "content": solver_prompt}]
    )
    solver_result = solver.content[0].text

    # Step 3: Result Interpretation (Logic → NL)
    interpretation_prompt = f"""
Given this solver result:
{solver_result}

Generate a clear clinical recommendation:
- Is it safe to add {candidate_drug}?
- Why or why not? (cite specific contraindications)
- What are the risks?
- What alternatives exist if contraindicated?

Format as a clinical summary for a physician."""

    interpretation = client.messages.create(
        model="claude-3-5-sonnet-20241022",
        max_tokens=400,
        messages=[{"role": "user", "content": interpretation_prompt}]
    )
    final_output = interpretation.content[0].text

    return {
        "formulation": logical_formulas,
        "solver_result": solver_result,
        "clinical_recommendation": final_output
    }


# Example usage
result = logic_lm_drug_checker(
    current_drugs=["warfarin", "aspirin"],
    candidate_drug="ibuprofen"
)

print("DRUG INTERACTION CHECK")
print("=" * 50)
print(result["clinical_recommendation"])