from transformers import pipeline
import requests
import json
import gradio as gr

pipe = pipeline("translation", "guymorlan/TokenizerLabeller")

# download json and open
# from https://huggingface.co/guymorlan/TokenizerLabeller/raw/main/playaling_words.json
r = requests.get("https://huggingface.co/guymorlan/TokenizerLabeller/raw/main/playaling_words.json")
data = json.loads(r.text)

# built gradio interface

def predict(input):

    out = pipe(input)[0]['translation_text']
    raw = out
    out = [x.strip() for x in out.split(" + ")]

    output =  f"{raw}<br><br><div style='direction: rtl; text-align: right; font-size: 20px; font-family: sans-serif; line-height: 1.5'>"
    for o in out:
        oo = [x.strip() for x in o.split("+")]
        newout = []
        for ooo in oo:
            if ooo in data:
                newout.append(f"<span style='color: green' title='{data[ooo]['translation']}\n{data[ooo]['features']}'>{data[ooo]['word']}</span>")
            else:
                newout.append(ooo)
                        
        output += "+".join(newout) + " | "

    return output[:-3] + "</div>"


gr.Interface(predict, "textbox", "html", title="Ammiya Tokenizer", description="Tokenize Ammiya text and show Playaling words").launch()