app.py

import gradio as gr
from fastapi import FastAPI, Request
import uvicorn
import spaces
from sentence_transformers import SentenceTransformer

print("Loading embedding model");
Embedder = SentenceTransformer("mixedbread-ai/mxbai-embed-large-v1")

app = FastAPI()

@spaces.GPU
def embed(text):
    query_embedding = Embedder.encode(text)
    return query_embedding.tolist();

@app.post("/v1/embeddings")
async def openai_embed(req: Request):
    body = await req.json();
    print(body); 
    model = body['model'];
    text = body['input'];
    embeddings = embed(text)
    return {
		'object': "list"
		,'data': [{
			'object': "embeddings"
			,'embedding': embeddings
			,'index':0
		}]
		,'model': 'mixedbread-ai/mxbai-embed-large-v1'
		,'usage':{
			 'prompt_tokens': 0
			,'total_tokens': 0
		}
	}


with gr.Blocks() as demo:
    text = gr.Textbox();
    embeddings = gr.Textbox()
    
    text.submit(embed, [text], [embeddings]);


print("Demo run...");
(app2,url,other) = demo.launch(prevent_thread_lock=True, server_name=None, server_port=8000);

GradioApp = gr.mount_gradio_app(app, demo, path="", ssr_mode=False);


demo.close();


if __name__ == '__main__':
    print("Running uviconr...");
    uvicorn.run(GradioApp, host="0.0.0.0", port=7860)