import gradio as gr
import os
import pandas as pd

path_to_L_model = str(os.environ['path_to_L_model'])
path_to_S_model = str(os.environ['path_to_S_model'])
read_token      = str(os.environ['read_token'])


languages = pd.read_csv("model_lang.csv", names=["Lang_acr"])

def check_lang(lang_acronym):
    if lang_acronym in languages["Lang_acr"].to_list():
        return "True"
    else: 
        return "False"
    
title = "DSA II"

description_main = """
A set of models to perform sentiment analysis. Choose between Large-Multilingual or Small-En-only. 

Use the interface to check if a language is included in the multilingual model, using language acronyms (e.g. it for Italian).
Select one of the two pages to start querying one of the two models.
"""

description_L = """
XLM-R tuned model, EN-tuned, pre-trained with 94 languages available (see original model [card](https://huggingface.co/xlm-roberta-large) to see which are available)
"""

description_S = """
A tuned BERT-base-cased model.
"""

examples = [
    ["I was followed by the blue monster but was not scared. I was calm and relaxed."],
    ["Ero seguito dal mostro blu, ma non ero spaventato. Ero calmo e rilassato."],
    ["Śledził mnie niebieski potwór, ale się nie bałem. Byłem spokojny i zrelaksowany."],
]

interface_words = gr.Interface(
            fn=check_lang,
            inputs="text", 
            outputs="text",
            description=description_main,
)

interface_model_L = gr.Interface.load(
            name=path_to_L_model,
            description=description_L,
            examples=examples,
            title=title,
            api_key=read_token,
)

interface_model_S = gr.Interface.load(
            name=path_to_S_model,
            description=description_S,
            examples=examples[0],
            title=title,
            api_key=read_token,
)

gr.TabbedInterface(
    [interface_words, interface_model_L, interface_model_S], 
    ["Intro", "Large Multilingual", "Base En"]
).launch()