app.py

import os
import streamlit as st
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, TranslationPipeline, pipeline

print("Loading the models...")

hf_token = os.getenv("HF_AUTH_TOKEN")
if not hf_token:
    raise ValueError("Hugging Face token not found. Please set the HF_AUTH_TOKEN environment variable.")

st.title("Translation and Summarization Web App")
st.write("""
### Powered by Hugging Face and Streamlit
This app uses pre-trained NLP models from Hugging Face to translate text between languages and summarize it.  
Enter text in the source language, select source and target languages, and see the translated and summarized text!
""")

@st.cache_resource
def load_translation_pipeline():
    print("Loading translation model...")
    model = AutoModelForSeq2SeqLM.from_pretrained(
        'issai/tilmash',
        use_auth_token=hf_token
    )
    tokenizer = AutoTokenizer.from_pretrained(
        "issai/tilmash",
        use_auth_token=hf_token
    )
    return TranslationPipeline(model=model, tokenizer=tokenizer, max_length=1000)

@st.cache_resource
def load_summarization_pipeline():
    print("Loading summarization model...")
    return pipeline("summarization", model="facebook/bart-large-cnn")

tilmash = load_translation_pipeline()
summarizer = load_summarization_pipeline()

languages = {
    "Kazakh (Cyrillic)": "kaz_Cyrl",
    "Russian (Cyrillic)": "rus_Cyrl",
    "English (Latin)": "eng_Latn",
    "Turkish (Latin)": "tur_Latn"
}

src_lang = st.selectbox("Select source language:", options=list(languages.keys()), index=0)
tgt_lang = st.selectbox("Select target language:", options=list(languages.keys()), index=2)

user_input = st.text_area("Enter text to translate:", "")

if st.button("Translate and Summarize Text"):
    if user_input.strip():
        translation_result = tilmash(user_input, src_lang=languages[src_lang], tgt_lang=languages[tgt_lang])
        translated_text = translation_result[0]['translation_text']

        st.subheader("Translation Result")
        st.write(f"**Translated Text:** {translated_text}")

        
        if len(translated_text) > 20: 
            summary_result = summarizer(translated_text, max_length=130, min_length=30, do_sample=False)
            summarized_text = summary_result[0]['summary_text']

            st.subheader("Summarization Result")
            st.write(f"**Summarized Text:** {summarized_text}")
        else:
            st.warning("Translated text is too short for summarization!")
    else:
        st.warning("Please enter some text to translate!")

st.sidebar.title("About")
st.sidebar.info("""
This app demonstrates the use of Hugging Face's NLP models with Streamlit.  
It uses the `issai/tilmash` model for translation and `facebook/bart-large-cnn` for summarization.
""")

print('After translation and summarization operation')