# main.py

import os
import asyncio
from fastapi import FastAPI, HTTPException
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel
import uvicorn
from serpapi import GoogleSearch
from openai import OpenAI
import json

# --- Securely get API keys from environment variables (provided by Cloud Run) ---
OPENROUTER_API_KEY = os.environ.get('OPENROUTER_API_KEY')
SERPAPI_API_KEY = os.environ.get('SERPAPI_API_KEY')

app = FastAPI(title="AI Shopping Assistant API")

app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"], # Allows all origins
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

class UserQuery(BaseModel):
    product_name: str
    description: str
    budget: float

# =========================================================================
# === PASTE YOUR 'search_with_serpapi' FUNCTION FROM COLAB HERE ===
# =========================================================================

def search_with_serpapi(search_query: str):
    """Performs a search on Google Shopping (India) using the SerpApi."""
    print(f"🕵️‍♂️ Searching Google Shopping with SerpApi for: '{search_query}'")
    try:
        search = GoogleSearch({
            "engine": "google_shopping",
            "q": search_query,
            "api_key": SERPAPI_API_KEY,
            "gl": "in",         # Geolocation: India
            "hl": "en",         # Host Language: English
            "currency": "INR"   # Currency: Indian Rupee
        })
        results = search.get_dict()
        shopping_results = results.get("shopping_results", [])
        if not shopping_results: return []

        # Extract the same detailed information your frontend expects
        top_results = []
        for item in shopping_results[:7]: # Get top 7 results for the AI to choose from
             price = item.get('extracted_price') # This field is usually just the number
             if not price: continue

             top_results.append({
                "title": item.get("title"),
                "price": price,
                "source": item.get("source"), # e.g., "Amazon.in", "Flipkart"
                "rating": item.get("rating"),
                "reviews_count": item.get("reviews"),
                "image_url": item.get("thumbnail"), # Use thumbnail for the image
                "link": item.get("link"),
                "features": [] # SerpApi doesn't provide feature bullets in the same way
            })
        print(f"✅ Found {len(top_results)} products from Google Shopping via SerpApi.")
        return top_results

    except Exception as e:
        print(f"❌ Error during SerpApi search: {e}")
        return None


# =========================================================================
# === PASTE YOUR 'get_llm_recommendations' FUNCTION FROM COLAB HERE ===
# =========================================================================

def get_llm_recommendations(user_query: UserQuery, product_data: list):
    """Generates a complex JSON object tailored for the React frontend."""
    print("🧠 Sending SerpApi data to LLM for analysis...")
    client = OpenAI(
      base_url="https://openrouter.ai/api/v1",
      api_key=OPENROUTER_API_KEY,
    )

    # --- Step 3: Updated Prompt ---
    # The prompt is updated to reflect the new, broader data source
    prompt = f"""
    You are an expert virtual shopping assistant. Your goal is to generate a structured JSON response to power a rich user interface.

    My Request:
    - Product: "{user_query.product_name}"
    - Needs: "{user_query.description}"
    - Budget: "Under ₹{user_query.budget} INR"

    Detailed Product Data from various online stores in India (via Google Shopping):
    {json.dumps(product_data, indent=2)}

    Based on my request and this data, return a single JSON object with the following exact keys: "summary", "comparison", "products", and "completeTheLook".
    1.  "summary": (String) A concise, friendly summary of your findings and top recommendation.
    2.  "comparison": (Object) A comparison between your top two recommendations. Must have "headers" and "rows" keys.
    3.  "products": (Array of Objects) A list of your top 2-3 recommended products. Each object must have these exact keys: "id", "name", "brand", "price", "rating", "image", and "link". Infer the brand from the title.
    4.  "completeTheLook": (Object) A suggestion for a complementary product. Must have "item" and "link" keys.

    Do not include any text or markdown outside of the main JSON object.
    """
    # -------------------------------

    try:
        response = client.chat.completions.create(
          model="google/gemini-flash-1.5",
          response_format={"type": "json_object"},
          messages=[{"role": "user", "content": prompt}],
          temperature=0.6,
        )
        print("✅ LLM response for React UI received.")
        return json.loads(response.choices[0].message.content)
    except Exception as e:
        print(f"❌ Error during OpenRouter API call: {e}")
        raise HTTPException(status_code=500, detail=f"Error communicating with the AI model: {e}")

# --- API Endpoint (This part is already complete) ---

@app.post("/search")
async def search_products_endpoint(query: UserQuery):
    search_query = f"{query.product_name} {query.description}"
    product_data = search_with_serpapi(search_query)
    if product_data is None: raise HTTPException(status_code=503, detail="The product search service (SerpApi) failed.")
    if not product_data: raise HTTPException(status_code=404, detail="Could not find any products matching your query.")
    recommendation_json = get_llm_recommendations(query, product_data)
    return recommendation_json

@app.get("/")
def read_root(): return {"message": "AI Shopping Assistant Backend is running!"}