from fastapi import FastAPI, HTTPException
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel, Field
import tensorflow as tf
import numpy as np

# =====================================================
# LOAD MODEL
# =====================================================

MODEL_PATH = "model_cnn_best.h5"

model = tf.keras.models.load_model(MODEL_PATH)

# =====================================================
# FASTAPI
# =====================================================

app = FastAPI(
    title="CNN Pose Classifier API",
    version="1.0.0",
    description="API klasifikasi aktivitas menggunakan CNN + MediaPipe Pose"
)

# =====================================================
# CORS
# =====================================================

app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],       # ganti dengan domain website jika production
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# =====================================================
# REQUEST MODEL
# =====================================================

class PoseInput(BaseModel):

    features: list[float] = Field(
        ...,
        min_length=66,
        max_length=66,
        description="66 pose features (x,y landmark)"
    )

# =====================================================
# ROOT
# =====================================================

@app.get("/")
def home():

    return {
        "status": "running",
        "message": "CNN Pose Classifier API",
        "input_shape": [66],
        "output": "binary classification"
    }

# =====================================================
# HEALTH CHECK
# =====================================================

@app.get("/health")

def health():

    return {
        "status": "healthy"
    }

# =====================================================
# PREDICTION
# =====================================================

@app.post("/predict")

def predict(data: PoseInput):

    try:

        # -----------------------------
        # Convert ke numpy
        # -----------------------------

        features = np.array(
            data.features,
            dtype=np.float32
        )

        # -----------------------------
        # Validasi jumlah fitur
        # -----------------------------

        if features.shape != (66,):

            raise HTTPException(
                status_code=400,
                detail="Input harus terdiri dari 66 fitur."
            )

        # -----------------------------
        # Validasi NaN dan Inf
        # -----------------------------

        if np.isnan(features).any():

            raise HTTPException(
                status_code=400,
                detail="Input mengandung NaN."
            )

        if np.isinf(features).any():

            raise HTTPException(
                status_code=400,
                detail="Input mengandung Infinity."
            )

        # -----------------------------
        # Reshape sesuai model
        # (batch,66,1)
        # -----------------------------

        features = features.reshape(
            1,
            66,
            1
        )
        # =============================
        # DEBUG INPUT
        # =============================
        print("=" * 60)
        print("DEBUG INPUT CNN")
        print("=" * 60)
        print("Shape :", features.shape)
        print("Min   :", np.min(features))
        print("Max   :", np.max(features))
        print("Mean  :", np.mean(features))
        flat = features.flatten()
        print("All 66 features:")
        for i, value in enumerate(flat):
            print(f"{i:02d}: {value:.6f}")
        # -----------------------------
        # Predict
        # -----------------------------

        prediction = model.predict(
            features,
            verbose=0
        )
        # =============================
        # DEBUG OUTPUT
        # =============================
        print("Raw prediction :", prediction)
        print("Probability    :", float(prediction[0][0]))
        print("=" * 60)

        probability = float(prediction[0][0])

        if probability >= 0.5:
            label = "berbahaya"
        else:
            label = "aman"

        return {

            "success": True,

            "prediction": label,

            "probability": round(probability, 4),

            "confidence": round(
                max(probability, 1 - probability),
                4
            )

        }

    except HTTPException:

        raise

    except Exception as e:

        raise HTTPException(
            status_code=500,
            detail=str(e)
        )