app.py

import base64
import io
import json
import os
import re
import secrets
from functools import wraps

import matplotlib
import matplotlib.cm as cm
import numpy as np
import requests as req
import tensorflow as tf
from flask import Flask, g, jsonify, request
from flask_cors import CORS
from PIL import Image
from dotenv import load_dotenv
from pymongo import MongoClient
from pymongo.server_api import ServerApi
from werkzeug.security import check_password_hash, generate_password_hash

matplotlib.use("Agg")

BASE_DIR = os.path.dirname(__file__)
load_dotenv(os.path.join(BASE_DIR, ".env"))
MODEL_PATH = os.path.join(BASE_DIR, "densenet121_covid_final.keras")
OPENROUTER_API_KEY = os.getenv("OPENROUTER_API_KEY", "")
OPENROUTER_MODEL = os.getenv("OPENROUTER_MODEL", "google/gemma-3-4b-it:free")
OPENROUTER_FALLBACK_MODELS = [
    model_name.strip()
    for model_name in os.getenv(
        "OPENROUTER_FALLBACK_MODELS",
        "meta-llama/llama-3.1-8b-instruct:free,mistralai/mistral-7b-instruct:free",
    ).split(",")
    if model_name.strip()
]
MONGODB_URI = os.getenv("MONGODB_URI", "")
MONGODB_DB_NAME = os.getenv("MONGODB_DB_NAME", "pulmoai")
FRONTEND_ORIGIN = os.getenv("FRONTEND_ORIGIN", "*")
PORT = int(os.getenv("PORT", "5000"))
DEFAULT_FRONTEND_ORIGINS = [
    "https://pulomonary-ai-care-system.vercel.app",
    "https://pulmonary-ai-care-system.vercel.app",
    "http://localhost:5173",
]


def parse_frontend_origins(origin_value):
    if origin_value == "*":
        return "*"
    origins = [origin.strip().rstrip("/") for origin in origin_value.split(",") if origin.strip()]
    return list(dict.fromkeys([*origins, *DEFAULT_FRONTEND_ORIGINS]))


FRONTEND_ORIGINS = parse_frontend_origins(FRONTEND_ORIGIN)


def is_allowed_origin(origin):
    return FRONTEND_ORIGINS == "*" or origin in FRONTEND_ORIGINS

app = Flask(__name__)
CORS(
    app,
    supports_credentials=True,
    resources={r"/*": {"origins": FRONTEND_ORIGINS}}
    if FRONTEND_ORIGINS != "*"
    else {r"/*": {"origins": "*"}},
)


@app.after_request
def add_cors_headers(response):
    origin = (request.headers.get("Origin") or "").rstrip("/")
    if origin and is_allowed_origin(origin):
        response.headers["Access-Control-Allow-Origin"] = origin
        response.headers["Vary"] = "Origin"
        response.headers["Access-Control-Allow-Credentials"] = "true"
        response.headers["Access-Control-Allow-Headers"] = "Authorization, Content-Type"
        response.headers["Access-Control-Allow-Methods"] = "GET, POST, OPTIONS"
    return response


CLASSES = ["COVID", "Lung_Opacity", "Normal", "Viral Pneumonia"]
DISPLAY_NAMES = {
    "COVID": "COVID-19",
    "Lung_Opacity": "Lung Opacity",
    "Normal": "Normal",
    "Viral Pneumonia": "Viral Pneumonia",
}
SEVERITY = {
    "COVID": "High",
    "Lung_Opacity": "Moderate",
    "Normal": "Low",
    "Viral Pneumonia": "High",
}

LOCAL_ADVICE_BY_DISEASE = {
    "Normal": {
        "summary": "The X-ray pattern appears within normal limits based on the model output. Clinical correlation is still important if symptoms such as fever, chest pain, or shortness of breath are present.",
        "urgency": "Low",
        "precautions": [
            "Continue routine health monitoring and seek care if new respiratory symptoms develop.",
            "Maintain good hydration and adequate sleep to support overall recovery and wellness.",
            "Avoid smoking and secondhand smoke exposure.",
            "Use a mask and hand hygiene in crowded settings if respiratory infections are circulating.",
            "Keep preventive vaccinations and routine checkups up to date.",
        ],
        "medications": [
            {"name": "No routine medication", "dosage": "As advised", "frequency": "Not routinely needed", "purpose": "No medication is typically needed for a normal study without symptoms."}
        ],
        "lifestyle": [
            "Stay physically active as tolerated.",
            "Maintain a balanced diet rich in fluids, fruits, and protein.",
            "Follow up with a clinician if symptoms persist despite a normal image result.",
        ],
        "followUp": "Routine follow-up only if symptoms continue or worsen.",
    },
    "Lung Opacity": {
        "summary": "Lung opacity can reflect infection, inflammation, fluid, or other causes and should be interpreted with symptoms and examination findings. A clinician should review the result to determine whether further imaging, oxygen assessment, or treatment is needed.",
        "urgency": "Moderate",
        "precautions": [
            "Seek prompt medical review, especially if you have fever, cough, chest pain, or breathlessness.",
            "Monitor oxygen saturation if available and seek urgent care for worsening shortness of breath.",
            "Avoid smoking, vaping, and dusty or polluted environments.",
            "Rest adequately and limit strenuous activity until reviewed.",
            "Maintain hydration unless a clinician has advised fluid restriction.",
        ],
        "medications": [
            {"name": "Symptom-directed treatment", "dosage": "As prescribed", "frequency": "As directed", "purpose": "Management depends on whether the opacity is due to infection, inflammation, or another cause."}
        ],
        "lifestyle": [
            "Prioritize rest and breathing comfort.",
            "Use steam inhalation or humidified air only if comfortable and clinician-approved.",
            "Keep follow-up imaging or lab appointments if recommended.",
        ],
        "followUp": "Arrange clinical review within 24-48 hours, sooner if symptoms are significant or worsening.",
    },
    "Viral Pneumonia": {
        "summary": "The model suggests viral pneumonia, which may cause cough, fever, fatigue, and reduced oxygen levels. Clinical assessment is important to determine severity and whether home care or hospital treatment is needed.",
        "urgency": "High",
        "precautions": [
            "Seek medical evaluation promptly, especially for shortness of breath or persistent fever.",
            "Check oxygen saturation if available and seek urgent care for low readings or breathing difficulty.",
            "Rest, isolate if infection is suspected, and use a mask around others.",
            "Drink fluids regularly unless you have been told to restrict intake.",
            "Go to emergency care immediately for confusion, bluish lips, or rapidly worsening breathing.",
        ],
        "medications": [
            {"name": "Supportive care", "dosage": "As prescribed", "frequency": "As directed", "purpose": "Treatment often focuses on fever control, hydration, and condition-specific therapy after clinician review."}
        ],
        "lifestyle": [
            "Reduce exertion and allow time for recovery.",
            "Avoid smoking and alcohol during illness.",
            "Track temperature, breathing symptoms, and oxygen levels if possible.",
        ],
        "followUp": "Same-day clinical review is recommended if symptoms are active; emergency care is needed for worsening breathlessness or low oxygen.",
    },
    "COVID-19": {
        "summary": "The model suggests a pattern concerning for COVID-19-related lung involvement. Severity can vary, so symptoms, oxygen level, and risk factors should guide whether home care, urgent review, or hospital evaluation is needed.",
        "urgency": "High",
        "precautions": [
            "Isolate according to local guidance and wear a mask around others.",
            "Seek prompt medical care for worsening cough, persistent fever, chest pain, or breathlessness.",
            "Monitor oxygen saturation if available and seek urgent care for low readings.",
            "Rest well and maintain hydration unless a clinician advises otherwise.",
            "Get emergency help for severe shortness of breath, confusion, or bluish lips.",
        ],
        "medications": [
            {"name": "Supportive care", "dosage": "As prescribed", "frequency": "As directed", "purpose": "Clinical treatment depends on symptom severity, timing of illness, and patient risk factors."}
        ],
        "lifestyle": [
            "Rest and avoid strenuous physical activity while symptomatic.",
            "Use good ventilation at home if isolating.",
            "Follow clinician advice on testing, antiviral eligibility, and recovery monitoring.",
        ],
        "followUp": "Medical review should be arranged promptly if symptoms are present, with urgent care for breathing issues or low oxygen.",
    },
}


def build_local_advice(disease, confidence):
    advice = LOCAL_ADVICE_BY_DISEASE.get(disease) or LOCAL_ADVICE_BY_DISEASE["Lung Opacity"]
    summary_suffix = f" Model confidence was {float(confidence):.1f}%."
    return {
        **advice,
        "summary": f"{advice['summary']}{summary_suffix}",
        "source": "local-fallback",
        "disclaimer": "This guidance is generated from predefined clinical safety rules and is not a substitute for a licensed medical evaluation.",
    }


def parse_openrouter_advice(response):
    try:
        result = response.json()
    except ValueError:
        print(f"OpenRouter returned non-JSON response: {response.text[:500]}")
        return None, "OpenRouter returned an invalid response", 502

    if "choices" not in result or not result["choices"]:
        print(f"OpenRouter response missing choices: {result}")
        error_message = "Unable to generate AI advice"
        if isinstance(result.get("error"), dict):
            error_message = result["error"].get("message", error_message)
        elif result.get("error"):
            error_message = str(result["error"])
        return None, error_message, 502

    text = result["choices"][0].get("message", {}).get("content", "")
    text = re.sub(r"```json|```", "", text).strip()
    match = re.search(r"\{[\s\S]*\}", text)
    if not match:
        print(f"OpenRouter response did not contain JSON: {text[:500]}")
        return None, "AI advice response was not valid JSON", 502

    advice_json = match.group(0)
    try:
        parsed_advice = json.loads(advice_json)
    except ValueError:
        print(f"Failed to parse AI advice JSON: {advice_json[:500]}")
        return None, "AI advice response could not be parsed", 502

    return parsed_advice, None, 200


def create_mongo_database():
    if not MONGODB_URI:
        raise RuntimeError("MONGODB_URI is not set. Add your MongoDB Atlas connection string.")

    client = MongoClient(MONGODB_URI, server_api=ServerApi("1"))
    database = client[MONGODB_DB_NAME]
    database.users.create_index("email", unique=True)
    database.sessions.create_index("token", unique=True)
    database.sessions.create_index("user_id")
    return database


mongo_db = create_mongo_database()

print("Loading model...")
model = tf.keras.models.load_model(MODEL_PATH, compile=False)
model.compile(optimizer="adam", loss="categorical_crossentropy", metrics=["accuracy"])
print("Model loaded")


def serialize_user(user_document):
    return {
        "id": str(user_document["_id"]),
        "fullName": user_document["full_name"],
        "email": user_document["email"],
    }


def create_session(user_id):
    token = secrets.token_urlsafe(32)
    mongo_db.sessions.insert_one({"token": token, "user_id": user_id})
    return token


def get_authenticated_user():
    auth_header = request.headers.get("Authorization", "")
    if not auth_header.startswith("Bearer "):
        return None, None

    token = auth_header.split(" ", 1)[1].strip()
    if not token:
        return None, None

    session_document = mongo_db.sessions.find_one({"token": token})
    if not session_document:
        return None, None

    user_document = mongo_db.users.find_one({"_id": session_document["user_id"]})
    return (user_document, token) if user_document else (None, None)


def require_auth(route_handler):
    @wraps(route_handler)
    def wrapper(*args, **kwargs):
        if request.method == "OPTIONS":
            return app.make_default_options_response()

        user_document, token = get_authenticated_user()
        if not user_document:
            return jsonify({"error": "Authentication required"}), 401

        g.current_user = user_document
        g.current_token = token
        return route_handler(*args, **kwargs)

    return wrapper


def get_gradcam_heatmap(current_model, img_array, pred_index=None):
    try:
        last_conv_layer = None
        for layer in reversed(current_model.layers):
            if isinstance(layer, tf.keras.layers.Conv2D):
                last_conv_layer = layer.name
                break
        if last_conv_layer is None:
            return None

        grad_model = tf.keras.models.Model(
            inputs=current_model.input,
            outputs=[current_model.get_layer(last_conv_layer).output, current_model.output],
        )

        img_tensor = tf.cast(img_array, tf.float32)
        with tf.GradientTape() as tape:
            tape.watch(img_tensor)
            conv_outputs, predictions = grad_model(img_tensor)
            tape.watch(conv_outputs)
            class_channel = predictions[:, pred_index]

        grads = tape.gradient(class_channel, conv_outputs)
        if grads is None:
            return None

        pooled_grads = tf.reduce_mean(grads, axis=(0, 1, 2))
        conv_outputs = conv_outputs[0]
        heatmap = conv_outputs @ pooled_grads[..., tf.newaxis]
        heatmap = tf.squeeze(heatmap)
        heatmap = tf.maximum(heatmap, 0) / (tf.math.reduce_max(heatmap) + 1e-8)
        return heatmap.numpy()
    except Exception as exc:
        print(f"Grad-CAM error: {exc}")
        return None


def overlay_gradcam(original_img, heatmap, alpha=0.4):
    heatmap_resized = np.uint8(255 * heatmap)
    jet = cm.get_cmap("jet")
    jet_colors = jet(np.arange(256))[:, :3]
    jet_heatmap = jet_colors[heatmap_resized]
    jet_heatmap = tf.keras.preprocessing.image.array_to_img(jet_heatmap)
    jet_heatmap = jet_heatmap.resize((original_img.shape[1], original_img.shape[0]))
    jet_heatmap = tf.keras.preprocessing.image.img_to_array(jet_heatmap)
    superimposed = jet_heatmap * alpha + original_img
    return np.clip(superimposed, 0, 255).astype(np.uint8)


@app.route("/auth/signup", methods=["POST"])
def sign_up():
    data = request.get_json() or {}
    full_name = (data.get("fullName") or "").strip()
    email = (data.get("email") or "").strip().lower()
    password = data.get("password") or ""

    if len(full_name) < 2:
        return jsonify({"error": "Full name must be at least 2 characters"}), 400
    if not re.fullmatch(r"[^@\s]+@[^@\s]+\.[^@\s]+", email):
        return jsonify({"error": "Please enter a valid email address"}), 400
    if len(password) < 6:
        return jsonify({"error": "Password must be at least 6 characters"}), 400
    if mongo_db.users.find_one({"email": email}):
        return jsonify({"error": "An account with this email already exists"}), 409

    user_result = mongo_db.users.insert_one(
        {
            "full_name": full_name,
            "email": email,
            "password_hash": generate_password_hash(password),
        }
    )
    user_document = mongo_db.users.find_one({"_id": user_result.inserted_id})
    token = create_session(user_document["_id"])
    return jsonify({"token": token, "user": serialize_user(user_document)}), 201


@app.route("/auth/signin", methods=["POST"])
def sign_in():
    data = request.get_json() or {}
    email = (data.get("email") or "").strip().lower()
    password = data.get("password") or ""

    user_document = mongo_db.users.find_one({"email": email})
    if not user_document or not check_password_hash(user_document["password_hash"], password):
        return jsonify({"error": "Invalid email or password"}), 401

    token = create_session(user_document["_id"])
    return jsonify({"token": token, "user": serialize_user(user_document)}), 200


@app.route("/auth/me", methods=["GET"])
@require_auth
def auth_me():
    return jsonify({"user": serialize_user(g.current_user)}), 200


@app.route("/auth/logout", methods=["POST"])
@require_auth
def logout():
    mongo_db.sessions.delete_one({"token": g.current_token})
    return jsonify({"message": "Logged out successfully"}), 200


@app.route("/predict", methods=["POST", "OPTIONS"])
@require_auth
def predict():
    if "file" not in request.files:
        return jsonify({"error": "No file uploaded"}), 400

    file = request.files["file"]
    print(f"Processing file: {file.filename}")

    try:
        img = Image.open(file.stream).convert("RGB")
        img_resized = img.resize((224, 224))
        img_array = np.array(img_resized) / 255.0
        img_array = np.expand_dims(img_array, axis=0)

        print("Running model prediction...")
        preds = model.predict(img_array)[0]
        pred_index = int(np.argmax(preds))
        pred_class = CLASSES[pred_index]

        print(f"Prediction complete: {pred_class}")

        # Simplified response without Grad-CAM to reduce memory usage
        findings = []
        for index, class_name in enumerate(CLASSES):
            findings.append(
                {
                    "name": DISPLAY_NAMES[class_name],
                    "confidence": f"{float(preds[index]):.4f}",
                    "severity": SEVERITY[class_name],
                    "predicted": index == pred_index,
                }
            )

        return jsonify(
            {
                "predicted": DISPLAY_NAMES[pred_class],
                "predicted_class_display": DISPLAY_NAMES[pred_class],
                "confidence": float(preds[pred_index]),
                "findings": findings,
                "model_version": "DenseNet121-COVID-v1.0",
            }
        )
    except Exception as e:
        print(f"Prediction error: {str(e)}")
        return jsonify({"error": f"Prediction failed: {str(e)}"}), 500


@app.route("/ai-advice", methods=["POST"])
@require_auth
def ai_advice():
    data = request.get_json() or {}
    disease = data.get("disease", "")
    confidence = data.get("confidence", 0)
    findings = data.get("findings", [])

    predictions_text = ", ".join(
        [f"{finding['name']}: {float(finding['confidence']) * 100:.1f}%" for finding in findings]
    )

    prompt = f"""You are a clinical AI assistant helping a radiologist interpret a chest X-ray analysis result. A DenseNet121 model has analyzed the X-ray with the following results:



Primary Diagnosis: {disease} ({confidence:.1f}% confidence)

All predictions: {predictions_text}



Provide a structured clinical response in this exact JSON format only. No markdown, no text outside the JSON:

{{

  "summary": "2-3 sentence clinical summary of the condition",

  "urgency": "Low | Moderate | High | Critical",

  "precautions": [

    "Precaution 1",

    "Precaution 2",

    "Precaution 3",

    "Precaution 4",

    "Precaution 5"

  ],

  "medications": [

    {{"name": "Drug name", "dosage": "e.g. 500mg", "frequency": "e.g. twice daily", "purpose": "brief purpose"}},

    {{"name": "Drug name", "dosage": "...", "frequency": "...", "purpose": "..."}}

  ],

  "lifestyle": [

    "Lifestyle recommendation 1",

    "Lifestyle recommendation 2",

    "Lifestyle recommendation 3"

  ],

  "followUp": "When and what kind of follow-up is recommended"

}}



For Normal findings provide general wellness advice. For diseases provide clinically appropriate guidance."""

    headers = {
        "Authorization": f"Bearer {OPENROUTER_API_KEY}",
        "Content-Type": "application/json",
        "HTTP-Referer": FRONTEND_ORIGINS[0] if FRONTEND_ORIGINS != "*" and FRONTEND_ORIGINS else "http://localhost:5173",
        "X-Title": "PulmoAI",
    }
    fallback_advice = build_local_advice(disease, confidence)

    if not OPENROUTER_API_KEY:
        return jsonify(
            {
                "advice": fallback_advice,
                "warning": "OPENROUTER_API_KEY is not configured. Returned local fallback guidance.",
            }
        )

    model_candidates = []
    for model_name in [OPENROUTER_MODEL, *OPENROUTER_FALLBACK_MODELS]:
        if model_name and model_name not in model_candidates:
            model_candidates.append(model_name)

    last_error = "Unable to generate AI advice"
    last_status = 502

    for model_name in model_candidates:
        body = {
            "model": model_name,
            "messages": [{"role": "user", "content": prompt}],
        }
        try:
            response = req.post(
                "https://openrouter.ai/api/v1/chat/completions",
                json=body,
                headers=headers,
                timeout=60,
            )
            response.raise_for_status()
        except req.RequestException as exc:
            status_code = getattr(exc.response, "status_code", 502)
            response_text = ""
            if getattr(exc, "response", None) is not None:
                response_text = exc.response.text[:500]
            print(f"OpenRouter request failed for model {model_name}: {exc}. Response: {response_text}")

            if status_code == 401:
                return jsonify({"error": "OpenRouter API key is invalid, expired, or belongs to a deleted account"}), 401

            last_status = status_code
            if status_code == 429:
                last_error = f"OpenRouter model '{model_name}' is temporarily rate-limited."
                continue

            last_error = f"Failed to fetch AI advice from OpenRouter using model '{model_name}'"
            continue

        parsed_advice, parse_error, parse_status = parse_openrouter_advice(response)
        if parsed_advice is not None:
            parsed_advice.setdefault("source", "openrouter")
            parsed_advice.setdefault(
                "disclaimer",
                "AI-generated guidance should be reviewed by a licensed clinician before making medical decisions.",
            )
            return jsonify({"advice": parsed_advice, "model": model_name})

        last_error = parse_error or last_error
        last_status = parse_status

    return jsonify(
        {
            "advice": fallback_advice,
            "warning": (
                f"{last_error} Returned local fallback guidance after trying: {', '.join(model_candidates)}."
            ),
        }
    ), 200


@app.route("/health", methods=["GET"])
def health():
    return jsonify({"status": "ok", "model": os.path.basename(MODEL_PATH), "database": MONGODB_DB_NAME})


if __name__ == "__main__":
    app.run(debug=True, host="0.0.0.0", port=PORT)