app.py

import streamlit as st
import requests
import chromadb
from sentence_transformers import SentenceTransformer
import json

# Initialize embedding model
model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")

# Connect to ChromaDB (Persistent)
DB_PATH = "./recipe_db"
client = chromadb.PersistentClient(path=DB_PATH)
collection = client.get_or_create_collection("recipes")

# Predefined Recipe Categories
recipe_categories = {
    "Desi": ["Nihari", "Karahi", "Biryani", "Haleem", "Saag"],
    "Fast Food": ["Burger", "Pizza", "Fries", "Shawarma"],
    "BBQ": ["Tikka", "Seekh Kebab", "Malai Boti"],
    "Seafood": ["Prawn Karahi", "Grilled Fish", "Fried Fish"]
}

# Check if ChromaDB has data, if not, insert sample data
if not collection.count():
    sample_recipes = [
        {"name": "Nihari", "city": "Lahore", "price": 800, "image": "https://example.com/nihari.jpg"},
        {"name": "Karahi", "city": "Lahore", "price": 1200, "image": "https://example.com/karahi.jpg"},
        {"name": "Biryani", "city": "Karachi", "price": 500, "image": "https://example.com/biryani.jpg"},
        {"name": "Chapli Kebab", "city": "Peshawar", "price": 400, "image": "https://example.com/chapli.jpg"},
        {"name": "Saag", "city": "Multan", "price": 600, "image": "https://example.com/saag.jpg"}
    ]

    for recipe in sample_recipes:
        embedding = model.encode(recipe["city"]).tolist()
        collection.add(
            ids=[recipe["name"]],
            embeddings=[embedding],
            documents=[json.dumps(recipe)]  # Convert dictionary to string
        )
    print("Sample data added to ChromaDB")


# Function to fetch restaurant data using Overpass API
def get_restaurants(city):
    overpass_url = "http://overpass-api.de/api/interpreter"
    query = f"""
    [out:json];
    area[name="{city}"]->.searchArea;
    node["amenity"="restaurant"](area.searchArea);
    out;
    """
    response = requests.get(overpass_url, params={'data': query})
    
    if response.status_code == 200:
        data = response.json()
        restaurants = []
        for element in data.get("elements", []):
            name = element.get("tags", {}).get("name", "Unknown Restaurant")
            restaurants.append(name)
        return restaurants[:5]  # Return top 5 results
    else:
        return ["No restaurant data found."]


# Streamlit UI
st.title("Pakistani Famous Recipes Finder 🍛")

# User inputs city
city = st.text_input("Enter a Pakistani City (e.g., Lahore, Karachi, Islamabad)").strip()

# User selects recipe type
recipe_type = st.selectbox("Select Recipe Type", options=list(recipe_categories.keys()))

# Optional: User inputs recipe (not mandatory)
query = st.selectbox("Select a Recipe (Optional)", ["Any"] + recipe_categories[recipe_type])

if st.button("Find Recipes & Restaurants"):
    if city:
        if query != "Any":
            # Retrieve specific recipe info from vector DB
            query_embedding = model.encode(query).tolist()
            results = collection.query(query_embedding, n_results=5)

            if results and "documents" in results and results["documents"]:
                st.subheader(f"Famous {query} in {city}")
                for doc in results["documents"]:
                    for recipe_json in doc:
                        recipe = json.loads(recipe_json)  # Convert back to dictionary
                        st.write(f"**Recipe:** {recipe['name']}")
                        st.image(recipe["image"], caption=recipe["name"], use_container_width=True)
                        st.write(f"Price: {recipe['price']} PKR")

                # Fetch restaurant data
                restaurants = get_restaurants(city)
                if restaurants:
                    st.subheader("Available at These Restaurants:")
                    for r in restaurants:
                        st.write(f"- {r}")
                else:
                    st.write("No restaurant data found.")
            else:
                st.write(f"No matching recipes found for '{query}' in {city}.")

        else:
            # Retrieve all famous recipes in the city
            city_embedding = model.encode(city).tolist()
            results = collection.query(city_embedding, n_results=5)

            if results and "documents" in results and results["documents"]:
                st.subheader(f"Famous Recipes in {city}")
                for doc in results["documents"]:
                    for recipe_json in doc:
                        recipe = json.loads(recipe_json)  # Convert back to dictionary
                        st.write(f"**Recipe:** {recipe['name']}")
                        st.image(recipe["image"], caption=recipe["name"], use_container_width=True)
                        st.write(f"Price: {recipe['price']} PKR")

            # Fetch restaurant data for multiple recipes
            restaurants = get_restaurants(city)
            if restaurants:
                st.subheader("Popular Restaurants in This City:")
                for r in restaurants:
                    st.write(f"- {r}")
    else:
        st.warning("Please enter a city name.")