app.py

import gradio as gr
from sentence_transformers import SentenceTransformer
import json

embedding_model = SentenceTransformer("all-MiniLM-L6-v2")

def get_embedding(texts_input):
    # Parse input nếu là JSON list
    if isinstance(texts_input, str):
        try:
            texts = json.loads(texts_input)
            if not isinstance(texts, list):
                texts = [texts_input]
        except:
            # fallback: mỗi dòng 1 chunk
            texts = [line.strip() for line in texts_input.split("\n") if line.strip()]
    else:
        texts = texts_input  # nếu client gửi list thật sự
    
    embeddings = embedding_model.encode(
        texts, convert_to_numpy=True, normalize_embeddings=True
    )
    return embeddings.tolist()

demo = gr.Interface(
    fn=get_embedding,
    inputs=gr.JSON(label="Input chunks"),  # <- quan trọng: JSON input
    outputs=gr.JSON(label="Embeddings")
)

demo.launch()