config.py

import os

BASE_DIR = os.path.abspath(os.path.dirname(__file__))


class Config:
    SECRET_KEY = os.environ.get("SECRET_KEY", os.urandom(32).hex())

    # Database
    SQLALCHEMY_DATABASE_URI = "sqlite:///" + os.path.join(BASE_DIR, "screenings.db")
    SQLALCHEMY_TRACK_MODIFICATIONS = False

    # Upload settings
    UPLOAD_FOLDER = os.path.join(BASE_DIR, "uploads")
    MAX_CONTENT_LENGTH = 10 * 1024 * 1024  # 10 MB
    ALLOWED_EXTENSIONS = {"png", "jpg", "jpeg"}

    # Model settings
    MODEL_PATH = os.path.join(BASE_DIR, "models", "malaria_detector.keras")
    IMAGE_SIZE = (300, 300)
    THRESHOLD_PATH = os.path.join(BASE_DIR, "models", "optimal_threshold.txt")

    # Grad-CAM settings
    GRADCAM_FOLDER = os.path.join(BASE_DIR, "static", "gradcam")
    GRADCAM_MAX_AGE = 3600  # seconds before auto-cleanup

    # -----------------------------------------------------------------------
    # Symptom scoring — clinically weighted
    # -----------------------------------------------------------------------
    # Individual symptom weights (based on WHO malaria diagnostic criteria)
    SYMPTOM_WEIGHTS = {
        # Primary indicators (classic malaria triad)
        "fever": 3.0,
        "chills": 2.5,
        "sweating": 2.5,
        # Secondary indicators
        "headache": 1.5,
        "nausea": 1.5,
        "fatigue": 1.0,
        "muscle_aches": 1.5,
        "joint_pain": 1.0,
        "diarrhea": 1.0,
        # Severe malaria indicators
        "jaundice": 3.0,
        # Risk context (not symptoms but increase prior probability)
        "travel": 2.5,
        "previous_malaria": 1.5,
    }
    MAX_SYMPTOM_SCORE = sum(SYMPTOM_WEIGHTS.values())

    # Bonus when classic malaria triad is present together
    # Fever + chills + sweating occurring together is highly specific to malaria
    TRIAD_BONUS = 3.0

    # -----------------------------------------------------------------------
    # Risk aggregation
    # -----------------------------------------------------------------------
    # Base weights for combining ML prediction with symptom score
    ML_WEIGHT = 0.65
    SYMPTOM_WEIGHT = 0.35

    # When ML is highly confident (above this threshold), increase ML weight
    ML_HIGH_CONFIDENCE = 0.85

    # When classic triad is present but ML says negative, add this override
    # (could be early-stage infection not yet visible in smear)
    TRIAD_OVERRIDE_BOOST = 0.15

    # Risk level thresholds
    RISK_LOW_THRESHOLD = 0.30
    RISK_HIGH_THRESHOLD = 0.60

    # Age-based risk multipliers (children <5 and elderly are most vulnerable)
    AGE_RISK_MULTIPLIERS = {
        "child_under_5": 1.25,
        "child_5_14": 1.10,
        "adult_15_60": 1.0,
        "elderly_over_60": 1.15,
    }