main.py

import os
import json
import faiss
import numpy as np
from fastapi import FastAPI, UploadFile, File, Form
from fastapi.middleware.cors import CORSMiddleware
from sentence_transformers import SentenceTransformer
from PIL import Image
import io
import requests

# Fix caching permissions for Hugging Face
os.environ["HF_HOME"] = "./cache"
os.environ["TRANSFORMERS_CACHE"] = "./cache"
os.environ["SENTENCE_TRANSFORMERS_HOME"] = "./cache"

app = FastAPI()

# Enable CORS (so Netlify frontend can call)
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],  # can restrict to Netlify domain later
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# Load products
with open("products.json", "r", encoding="utf-8") as f:
    products = json.load(f)

print(f"📦 Loaded {len(products)} products")

# Load FAISS index
index = faiss.read_index("products.index")

# Load CLIP model
print("🧠 Loading CLIP model...")
model = SentenceTransformer("sentence-transformers/clip-ViT-B-32", cache_folder="./cache")


@app.get("/")
def root():
    return {"message": "🚀 Visual Product Matcher API is running!"}


@app.post("/search_text")
def search_text(query: str = Form(...), top_k: int = 5, min_score: float = 0.0):
    """
    Search products using text query.
    """
    query_emb = model.encode([query], convert_to_numpy=True, normalize_embeddings=True)
    sims, indices = index.search(query_emb, top_k)

    results = []
    for sim, idx in zip(sims[0], indices[0]):
        score = float(sim)  # already cosine similarity (0–1)
        if score >= min_score:
            item = products[idx].copy()
            item["score"] = score
            results.append(item)

    return {"matches": results}


@app.post("/match")  # image search
async def search_image(
    file: UploadFile = File(None),
    image_url: str = Form(None),
    top_k: int = 5,
    min_score: float = 0.0
):
    """
    Search products using image query (upload or URL).
    """
    if file:
        image_bytes = await file.read()
        image = Image.open(io.BytesIO(image_bytes)).convert("RGB")
    elif image_url:
        response = requests.get(image_url)
        image = Image.open(io.BytesIO(response.content)).convert("RGB")
    else:
        return {"error": "No image provided"}

    image_emb = model.encode([image], convert_to_numpy=True, normalize_embeddings=True)
    sims, indices = index.search(image_emb, top_k)

    results = []
    for sim, idx in zip(sims[0], indices[0]):
        score = float(sim)  # cosine similarity
        if score >= min_score:
            item = products[idx].copy()
            item["score"] = score
            results.append(item)

    return {"matches": results}