def get_recommendations(input_dict: dict, probability: float) -> list:
    """Generate retention recommendations based on the 5 key features."""
    recs = []

    service_calls = int(input_dict.get("Customer Service Calls", 0))
    monthly       = float(input_dict.get("Monthly Charge", 0))
    contract      = str(input_dict.get("Contract Type", "Month-to-Month"))
    account_len   = int(input_dict.get("Account Length", 0))
    gb_download   = float(input_dict.get("Avg Monthly GB Download", 0))

    # Critical risk — immediate action
    if probability >= 0.75:
        recs.append((
            "🎯", "Launch Immediate Retention Campaign",
            "Customer is at critical risk. Assign a dedicated account manager "
            "and offer a personalized retention package within 24 hours."
        ))

    # High service calls — #1 churn driver
    if service_calls >= 4:
        recs.append((
            "📞", "Resolve Service Issues Urgently",
            f"{service_calls} service calls detected — the top churn predictor. "
            "Escalate to senior support, identify root cause, and follow up proactively."
        ))
    elif service_calls >= 2:
        recs.append((
            "🛠", "Proactive Support Check-in",
            "Schedule a satisfaction call to address any unresolved issues "
            "before they escalate further."
        ))

    # High monthly charge
    if monthly > 90:
        recs.append((
            "💰", "Review & Optimize Pricing Plan",
            f"Monthly charge of ${monthly:.0f} is above average. "
            "Offer a bundled plan or loyalty discount to reduce cost by 15-20%."
        ))

    # Month-to-Month contract
    if contract in ["Month-to-Month", "0", 0]:
        recs.append((
            "📋", "Offer Annual Contract Upgrade",
            "Month-to-Month customers churn 3x more. "
            "Offer a 20% discount to switch to a 1 or 2-year contract."
        ))

    # New customer
    if account_len <= 12:
        recs.append((
            "🚀", "New Customer Onboarding Program",
            f"Only {account_len} months as a customer. "
            "Assign an onboarding specialist and schedule a 30-day satisfaction check-in."
        ))

    # High data usage — valuable customer
    if gb_download > 40:
        recs.append((
            "📡", "Offer Premium Data Plan Upgrade",
            f"Averaging {gb_download:.0f} GB/month — a high-value power user. "
            "Offer an unlimited data plan with priority network access."
        ))

    # Low usage — engagement risk
    elif gb_download < 10 and probability > 0.4:
        recs.append((
            "📊", "Re-engage Low Usage Customer",
            "Low data usage may indicate disengagement. "
            "Offer free data credits or a plan downgrade to retain the customer."
        ))

    if not recs:
        recs.append((
            "✅", "Customer in Good Standing",
            "No immediate action required. "
            "Monitor quarterly and consider a loyalty reward to strengthen retention."
        ))

    return recs