main.py

import os
import re
import pandas as pd
import faiss
import torch
import numpy as np
from fastapi import FastAPI
from fastapi.middleware.cors import CORSMiddleware
from sentence_transformers import SentenceTransformer
from huggingface_hub import snapshot_download

REPO_ID = "abhinavsunil/kitchenelite-recipe-model"
MODEL_CACHE = "/tmp/model_cache"
TOP_K = 5

app = FastAPI(title="KitchenElite Recipe Search API")
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],  # Allows all origins
    allow_credentials=True,
    allow_methods=["*"],  # Allows all methods (GET, POST, etc.)
    allow_headers=["*"],  # Allows all headers
)
model = None
index = None
df = None


# ==============================
# STARTUP EVENT
# ==============================

@app.on_event("startup")
def load_assets():
    global model, index, df

    print("🚀 Downloading model repo snapshot...")

    local_dir = snapshot_download(
        repo_id=REPO_ID,
        local_dir=MODEL_CACHE,
        local_dir_use_symlinks=False
    )

    print("📦 Loading metadata...")
    df = pd.read_parquet(os.path.join(local_dir, "metadata.parquet"))

    print("📦 Loading FAISS index...")
    index = faiss.read_index(os.path.join(local_dir, "recipes.index"))

    print("📦 Loading SentenceTransformer model...")
    model = SentenceTransformer(local_dir, device="cpu")

    print("✅ All assets loaded successfully!")


# ==============================
# UTILITY FUNCTION
# ==============================

def clean_instructions(instruction_input):
    if isinstance(instruction_input, str) and instruction_input.startswith('c("'):
        content = re.search(r'c\("(.*)"\)', instruction_input)
        if content:
            return [
                step.strip().strip('"')
                for step in content.group(1).split('", "')
            ]

    if isinstance(instruction_input, (list, np.ndarray)):
        return list(instruction_input)

    return [str(instruction_input)]


# ==============================
# ROUTES
# ==============================

@app.get("/")
def home():
    return {"status": "KitchenElite API Running 🚀"}


@app.get("/search")
def search(query: str):
    query_vector = model.encode([query])
    faiss.normalize_L2(query_vector)

    distances, indices = index.search(
        query_vector.astype("float32"),
        TOP_K
    )

    results = df.iloc[indices[0]]

    output = []
    for _, row in results.iterrows():
        output.append({
            "name": str(row["name"]),
            "ingredients": (
                list(row["ingredients"])
                if isinstance(row["ingredients"], np.ndarray)
                else row["ingredients"]
            ),
            "calories": float(row["calories"]),
            "protein": float(row["protein"]),
            "instructions": clean_instructions(row["RecipeInstructions"])
        })

    return {
        "query": query,
        "results": output
    }