analysis.py

import os
import json
import io
import numpy as np
import tensorflow as tf
from PIL import Image
import google.generativeai as genai
from fastapi import APIRouter , File, UploadFile, Form
from fastapi.responses import JSONResponse


router = APIRouter()

genai.configure(api_key="") 
gemini_model = genai.GenerativeModel('gemini-2.5-flash')
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'


MODEL_PATH = 'phoenix_ai_model.h5'
CLASS_NAMES = ['gender_female', 'gender_male', 'kids', 'safe', 'unsafe']

try:
    mobilenet_model = tf.keras.models.load_model(MODEL_PATH, compile=False)
    print("AI System Online with FastAPI")
except Exception as e:
    print(f"Error loading model: {e}")
    mobilenet_model = None

async def get_dynamic_analysis_from_gemini(image_pil, detected_label, title="", description="", category=""):
    prompt = f"""
    The product is classified as '{detected_label}'.
    User Context:
    Title: {title}
    Description: {description}
    Category: {category}

    Tasks:
    1. Verify if the image matches the provided title, description, and category.
    2. Estimate suitable 'min_age' and 'max_age'.
    3. Confirm matching (True/False).

    Return ONLY JSON:
    {{"min_age": int, "max_age": int, "matching_confirmed": bool}}
    """
    try:
        response = gemini_model.generate_content([prompt, image_pil])
        raw_text = response.text.strip()
        
        
        start = raw_text.find("{")
        end = raw_text.rfind("}") + 1

        if start != -1 and end != -1:
            clean_json = raw_text[start:end]
            return json.loads(clean_json)
        else:
            raise ValueError("Invalid JSON from Gemini")

    except Exception:
        
        return {"min_age": 12, "max_age": 60, "matching_confirmed": True}

@router.post("/analyze-product")
async def analyze_endpoint(
    image: UploadFile = File(...),
    title: str = Form(""),
    description: str = Form(""),
    category: str = Form("")
):
    if mobilenet_model is None:
        return JSONResponse(content={"error": "AI Model not loaded"}, status_code=500)

    try:
        
        image_bytes = await image.read()
        pil_img = Image.open(io.BytesIO(image_bytes)).convert('RGB')
        
        
        img_resized = pil_img.resize((224, 224))
        img_array = np.array(img_resized) / 255.0
        img_array = np.expand_dims(img_array, axis=0).astype(np.float32)

        
        predictions = mobilenet_model.predict(img_array, verbose=0)
        confidence = float(np.max(predictions[0]))
        label = CLASS_NAMES[np.argmax(predictions[0])]

        
        if confidence < 0.6:
            return {
                "status": "rejected_low_confidence",
                "data": None,
                "error_msg": "Model confidence too low."
            }

        status = "rejected"
        final_data = {"target_gender": "unknown", "min_age": 0, "max_age": 0}

        if label != 'unsafe':
            
            dynamic_data = await get_dynamic_analysis_from_gemini(
                pil_img, label, title, description, category
            )
            
            if dynamic_data.get("matching_confirmed") is True:
                status = "approved"
                
                
                target_gender = "unisex"
                if label == 'gender_male': target_gender = "male"
                elif label == 'gender_female': target_gender = "female"
                elif label == 'kids': target_gender = "kids"

                final_data = {
                    "target_gender": target_gender,
                    "min_age": int(dynamic_data.get("min_age", 12)),
                    "max_age": int(dynamic_data.get("max_age", 60))
                }
            else:
                status = "rejected_content_mismatch"
        else:
            status = "rejected_unsafe"

        return {
            "status": status,
            "data": final_data if status == "approved" else None,
            "error_msg": "" if status == "approved" else f"Analysis failed: {status}"
        }

    except Exception as e:
        return JSONResponse(content={"status": "error", "message": str(e)}, status_code=500)