app.py

import gradio as gr
import numpy as np
from PIL import Image
from sentence_transformers import SentenceTransformer

model = SentenceTransformer('clip-ViT-B-32', cache_folder='/app/cache')


def image_to_embedding(img: np.ndarray = None, txt: str = None) -> np.ndarray:
    if img is None and not txt:
        return []
    
    if img is not None:
        embedding = model.encode(sentences=[Image.fromarray(img)], batch_size=128)
    else:
        embedding = model.encode(sentences=[txt], batch_size=128)

    return embedding


demo = gr.Interface(fn=image_to_embedding, inputs=["image", "textbox"], outputs="textbox", cache_examples=True)
demo.launch(server_name="0.0.0.0")