app.py

# Make sure to install dependencies: pip install sentence-transformers gradio torch spaces
# ZamAI Embeddings Inference Space with GPU support for Hugging Face Spaces

import gradio as gr
from sentence_transformers import SentenceTransformer
import torch
import spaces

MODEL_ID = "tasal9/Multilingual-ZamAI-Embeddings"

# Robust device selection
device = "cuda" if torch.cuda.is_available() and torch.cuda.device_count() > 0 else "cpu"
try:
    # Try loading model with trust_remote_code=True for custom models.
    model = SentenceTransformer(MODEL_ID, device=device, trust_remote_code=True)
    model_status = f"Loaded model on {device.upper()}"
except Exception as e:
    model = None
    model_status = f"Model load error: {e}"

@spaces.GPU
def embed_text(input_text):
    if not input_text or not input_text.strip():
        return "Please enter some text to embed."
    if model is None:
        return model_status
    try:
        with torch.no_grad():
            emb = model.encode(input_text, device=device, show_progress_bar=False)
        if hasattr(emb, "shape"):
            shape = emb.shape
        else:
            shape = (len(emb),)
        return f"Embedding shape: {shape}\n{emb}"
    except Exception as e:
        return f"Embedding error: {e}"

with gr.Blocks(title="Multilingual-ZamAI-Embeddings Inference") as demo:
    gr.Markdown("# 🚀 Multilingual-ZamAI-Embeddings Inference Space")
    gr.Markdown("Enter text to get its embedding using the ZamAI model.")
    gr.Markdown(f"### System Status: `{model_status}`")
    with gr.Row():
        with gr.Column():
            test_input = gr.Textbox(label="Input Text", lines=3, placeholder="Enter text to embed...")
            test_btn = gr.Button("Get Embedding", variant="primary")
        with gr.Column():
            test_output = gr.Textbox(label="Embedding Output", lines=8, interactive=False)
    test_btn.click(embed_text, inputs=test_input, outputs=test_output)

if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", server_port=7860)